asterisk/main/stasis_message_router.c

353 lines
8.4 KiB
C
Raw Normal View History

/*
* Asterisk -- An open source telephony toolkit.
*
* Copyright (C) 2013, Digium, Inc.
*
* David M. Lee, II <dlee@digium.com>
*
* See http://www.asterisk.org for more information about
* the Asterisk project. Please do not directly contact
* any of the maintainers of this project for assistance;
* the project provides a web site, mailing lists and IRC
* channels for your use.
*
* This program is free software, distributed under the terms of
* the GNU General Public License Version 2. See the LICENSE file
* at the top of the source tree.
*/
/*! \file
*
* \brief Stasis message router implementation.
*
* \author David M. Lee, II <dlee@digium.com>
*/
/*** MODULEINFO
<support_level>core</support_level>
***/
#include "asterisk.h"
ASTERISK_FILE_VERSION(__FILE__, "$Revision$")
#include "asterisk/astobj2.h"
#include "asterisk/stasis_message_router.h"
#include "asterisk/vector.h"
/*! \internal */
struct stasis_message_route {
/*! Message type handle by this route. */
struct stasis_message_type *message_type;
/*! Callback function for incoming message processing. */
stasis_subscription_cb callback;
/*! Data pointer to be handed to the callback. */
void *data;
};
AST_VECTOR(route_table, struct stasis_message_route);
Multiple revisions 399887,400138,400178,400180-400181 ........ r399887 | dlee | 2013-09-26 10:41:47 -0500 (Thu, 26 Sep 2013) | 1 line Minor performance bump by not allocate manager variable struct if we don't need it ........ r400138 | dlee | 2013-09-30 10:24:00 -0500 (Mon, 30 Sep 2013) | 23 lines Stasis performance improvements This patch addresses several performance problems that were found in the initial performance testing of Asterisk 12. The Stasis dispatch object was allocated as an AO2 object, even though it has a very confined lifecycle. This was replaced with a straight ast_malloc(). The Stasis message router was spending an inordinate amount of time searching hash tables. In this case, most of our routers had 6 or fewer routes in them to begin with. This was replaced with an array that's searched linearly for the route. We more heavily rely on AO2 objects in Asterisk 12, and the memset() in ao2_ref() actually became noticeable on the profile. This was #ifdef'ed to only run when AO2_DEBUG was enabled. After being misled by an erroneous comment in taskprocessor.c during profiling, the wrong comment was removed. Review: https://reviewboard.asterisk.org/r/2873/ ........ r400178 | dlee | 2013-09-30 13:26:27 -0500 (Mon, 30 Sep 2013) | 24 lines Taskprocessor optimization; switch Stasis to use taskprocessors This patch optimizes taskprocessor to use a semaphore for signaling, which the OS can do a better job at managing contention and waiting that we can with a mutex and condition. The taskprocessor execution was also slightly optimized to reduce the number of locks taken. The only observable difference in the taskprocessor implementation is that when the final reference to the taskprocessor goes away, it will execute all tasks to completion instead of discarding the unexecuted tasks. For systems where unnamed semaphores are not supported, a really simple semaphore implementation is provided. (Which gives identical performance as the original taskprocessor implementation). The way we ended up implementing Stasis caused the threadpool to be a burden instead of a boost to performance. This was switched to just use taskprocessors directly for subscriptions. Review: https://reviewboard.asterisk.org/r/2881/ ........ r400180 | dlee | 2013-09-30 13:39:34 -0500 (Mon, 30 Sep 2013) | 28 lines Optimize how Stasis forwards are dispatched This patch optimizes how forwards are dispatched in Stasis. Originally, forwards were dispatched as subscriptions that are invoked on the publishing thread. This did not account for the vast number of forwards we would end up having in the system, and the amount of work it would take to walk though the forward subscriptions. This patch modifies Stasis so that rather than walking the tree of forwards on every dispatch, when forwards and subscriptions are changed, the subscriber list for every topic in the tree is changed. This has a couple of benefits. First, this reduces the workload of dispatching messages. It also reduces contention when dispatching to different topics that happen to forward to the same aggregation topic (as happens with all of the channel, bridge and endpoint topics). Since forwards are no longer subscriptions, the bulk of this patch is simply changing stasis_subscription objects to stasis_forward objects (which, admittedly, I should have done in the first place.) Since this required me to yet again put in a growing array, I finally abstracted that out into a set of ast_vector macros in asterisk/vector.h. Review: https://reviewboard.asterisk.org/r/2883/ ........ r400181 | dlee | 2013-09-30 13:48:57 -0500 (Mon, 30 Sep 2013) | 28 lines Remove dispatch object allocation from Stasis publishing While looking for areas for performance improvement, I realized that an unused feature in Stasis was negatively impacting performance. When a message is sent to a subscriber, a dispatch object is allocated for the dispatch, containing the topic the message was published to, the subscriber the message is being sent to, and the message itself. The topic is actually unused by any subscriber in Asterisk today. And the subscriber is associated with the taskprocessor the message is being dispatched to. First, this patch removes the unused topic parameter from Stasis subscription callbacks. Second, this patch introduces the concept of taskprocessor local data, data that may be set on a taskprocessor and provided along with the data pointer when a task is pushed using the ast_taskprocessor_push_local() call. This allows the task to have both data specific to that taskprocessor, in addition to data specific to that invocation. With those two changes, the dispatch object can be removed completely, and the message is simply refcounted and sent directly to the taskprocessor. Review: https://reviewboard.asterisk.org/r/2884/ ........ Merged revisions 399887,400138,400178,400180-400181 from http://svn.asterisk.org/svn/asterisk/branches/12 git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@400186 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-09-30 18:55:27 +00:00
static struct stasis_message_route *route_table_find(struct route_table *table,
Multiple revisions 399887,400138,400178,400180-400181 ........ r399887 | dlee | 2013-09-26 10:41:47 -0500 (Thu, 26 Sep 2013) | 1 line Minor performance bump by not allocate manager variable struct if we don't need it ........ r400138 | dlee | 2013-09-30 10:24:00 -0500 (Mon, 30 Sep 2013) | 23 lines Stasis performance improvements This patch addresses several performance problems that were found in the initial performance testing of Asterisk 12. The Stasis dispatch object was allocated as an AO2 object, even though it has a very confined lifecycle. This was replaced with a straight ast_malloc(). The Stasis message router was spending an inordinate amount of time searching hash tables. In this case, most of our routers had 6 or fewer routes in them to begin with. This was replaced with an array that's searched linearly for the route. We more heavily rely on AO2 objects in Asterisk 12, and the memset() in ao2_ref() actually became noticeable on the profile. This was #ifdef'ed to only run when AO2_DEBUG was enabled. After being misled by an erroneous comment in taskprocessor.c during profiling, the wrong comment was removed. Review: https://reviewboard.asterisk.org/r/2873/ ........ r400178 | dlee | 2013-09-30 13:26:27 -0500 (Mon, 30 Sep 2013) | 24 lines Taskprocessor optimization; switch Stasis to use taskprocessors This patch optimizes taskprocessor to use a semaphore for signaling, which the OS can do a better job at managing contention and waiting that we can with a mutex and condition. The taskprocessor execution was also slightly optimized to reduce the number of locks taken. The only observable difference in the taskprocessor implementation is that when the final reference to the taskprocessor goes away, it will execute all tasks to completion instead of discarding the unexecuted tasks. For systems where unnamed semaphores are not supported, a really simple semaphore implementation is provided. (Which gives identical performance as the original taskprocessor implementation). The way we ended up implementing Stasis caused the threadpool to be a burden instead of a boost to performance. This was switched to just use taskprocessors directly for subscriptions. Review: https://reviewboard.asterisk.org/r/2881/ ........ r400180 | dlee | 2013-09-30 13:39:34 -0500 (Mon, 30 Sep 2013) | 28 lines Optimize how Stasis forwards are dispatched This patch optimizes how forwards are dispatched in Stasis. Originally, forwards were dispatched as subscriptions that are invoked on the publishing thread. This did not account for the vast number of forwards we would end up having in the system, and the amount of work it would take to walk though the forward subscriptions. This patch modifies Stasis so that rather than walking the tree of forwards on every dispatch, when forwards and subscriptions are changed, the subscriber list for every topic in the tree is changed. This has a couple of benefits. First, this reduces the workload of dispatching messages. It also reduces contention when dispatching to different topics that happen to forward to the same aggregation topic (as happens with all of the channel, bridge and endpoint topics). Since forwards are no longer subscriptions, the bulk of this patch is simply changing stasis_subscription objects to stasis_forward objects (which, admittedly, I should have done in the first place.) Since this required me to yet again put in a growing array, I finally abstracted that out into a set of ast_vector macros in asterisk/vector.h. Review: https://reviewboard.asterisk.org/r/2883/ ........ r400181 | dlee | 2013-09-30 13:48:57 -0500 (Mon, 30 Sep 2013) | 28 lines Remove dispatch object allocation from Stasis publishing While looking for areas for performance improvement, I realized that an unused feature in Stasis was negatively impacting performance. When a message is sent to a subscriber, a dispatch object is allocated for the dispatch, containing the topic the message was published to, the subscriber the message is being sent to, and the message itself. The topic is actually unused by any subscriber in Asterisk today. And the subscriber is associated with the taskprocessor the message is being dispatched to. First, this patch removes the unused topic parameter from Stasis subscription callbacks. Second, this patch introduces the concept of taskprocessor local data, data that may be set on a taskprocessor and provided along with the data pointer when a task is pushed using the ast_taskprocessor_push_local() call. This allows the task to have both data specific to that taskprocessor, in addition to data specific to that invocation. With those two changes, the dispatch object can be removed completely, and the message is simply refcounted and sent directly to the taskprocessor. Review: https://reviewboard.asterisk.org/r/2884/ ........ Merged revisions 399887,400138,400178,400180-400181 from http://svn.asterisk.org/svn/asterisk/branches/12 git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@400186 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-09-30 18:55:27 +00:00
struct stasis_message_type *message_type)
{
Multiple revisions 399887,400138,400178,400180-400181 ........ r399887 | dlee | 2013-09-26 10:41:47 -0500 (Thu, 26 Sep 2013) | 1 line Minor performance bump by not allocate manager variable struct if we don't need it ........ r400138 | dlee | 2013-09-30 10:24:00 -0500 (Mon, 30 Sep 2013) | 23 lines Stasis performance improvements This patch addresses several performance problems that were found in the initial performance testing of Asterisk 12. The Stasis dispatch object was allocated as an AO2 object, even though it has a very confined lifecycle. This was replaced with a straight ast_malloc(). The Stasis message router was spending an inordinate amount of time searching hash tables. In this case, most of our routers had 6 or fewer routes in them to begin with. This was replaced with an array that's searched linearly for the route. We more heavily rely on AO2 objects in Asterisk 12, and the memset() in ao2_ref() actually became noticeable on the profile. This was #ifdef'ed to only run when AO2_DEBUG was enabled. After being misled by an erroneous comment in taskprocessor.c during profiling, the wrong comment was removed. Review: https://reviewboard.asterisk.org/r/2873/ ........ r400178 | dlee | 2013-09-30 13:26:27 -0500 (Mon, 30 Sep 2013) | 24 lines Taskprocessor optimization; switch Stasis to use taskprocessors This patch optimizes taskprocessor to use a semaphore for signaling, which the OS can do a better job at managing contention and waiting that we can with a mutex and condition. The taskprocessor execution was also slightly optimized to reduce the number of locks taken. The only observable difference in the taskprocessor implementation is that when the final reference to the taskprocessor goes away, it will execute all tasks to completion instead of discarding the unexecuted tasks. For systems where unnamed semaphores are not supported, a really simple semaphore implementation is provided. (Which gives identical performance as the original taskprocessor implementation). The way we ended up implementing Stasis caused the threadpool to be a burden instead of a boost to performance. This was switched to just use taskprocessors directly for subscriptions. Review: https://reviewboard.asterisk.org/r/2881/ ........ r400180 | dlee | 2013-09-30 13:39:34 -0500 (Mon, 30 Sep 2013) | 28 lines Optimize how Stasis forwards are dispatched This patch optimizes how forwards are dispatched in Stasis. Originally, forwards were dispatched as subscriptions that are invoked on the publishing thread. This did not account for the vast number of forwards we would end up having in the system, and the amount of work it would take to walk though the forward subscriptions. This patch modifies Stasis so that rather than walking the tree of forwards on every dispatch, when forwards and subscriptions are changed, the subscriber list for every topic in the tree is changed. This has a couple of benefits. First, this reduces the workload of dispatching messages. It also reduces contention when dispatching to different topics that happen to forward to the same aggregation topic (as happens with all of the channel, bridge and endpoint topics). Since forwards are no longer subscriptions, the bulk of this patch is simply changing stasis_subscription objects to stasis_forward objects (which, admittedly, I should have done in the first place.) Since this required me to yet again put in a growing array, I finally abstracted that out into a set of ast_vector macros in asterisk/vector.h. Review: https://reviewboard.asterisk.org/r/2883/ ........ r400181 | dlee | 2013-09-30 13:48:57 -0500 (Mon, 30 Sep 2013) | 28 lines Remove dispatch object allocation from Stasis publishing While looking for areas for performance improvement, I realized that an unused feature in Stasis was negatively impacting performance. When a message is sent to a subscriber, a dispatch object is allocated for the dispatch, containing the topic the message was published to, the subscriber the message is being sent to, and the message itself. The topic is actually unused by any subscriber in Asterisk today. And the subscriber is associated with the taskprocessor the message is being dispatched to. First, this patch removes the unused topic parameter from Stasis subscription callbacks. Second, this patch introduces the concept of taskprocessor local data, data that may be set on a taskprocessor and provided along with the data pointer when a task is pushed using the ast_taskprocessor_push_local() call. This allows the task to have both data specific to that taskprocessor, in addition to data specific to that invocation. With those two changes, the dispatch object can be removed completely, and the message is simply refcounted and sent directly to the taskprocessor. Review: https://reviewboard.asterisk.org/r/2884/ ........ Merged revisions 399887,400138,400178,400180-400181 from http://svn.asterisk.org/svn/asterisk/branches/12 git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@400186 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-09-30 18:55:27 +00:00
size_t idx;
struct stasis_message_route *route;
Multiple revisions 399887,400138,400178,400180-400181 ........ r399887 | dlee | 2013-09-26 10:41:47 -0500 (Thu, 26 Sep 2013) | 1 line Minor performance bump by not allocate manager variable struct if we don't need it ........ r400138 | dlee | 2013-09-30 10:24:00 -0500 (Mon, 30 Sep 2013) | 23 lines Stasis performance improvements This patch addresses several performance problems that were found in the initial performance testing of Asterisk 12. The Stasis dispatch object was allocated as an AO2 object, even though it has a very confined lifecycle. This was replaced with a straight ast_malloc(). The Stasis message router was spending an inordinate amount of time searching hash tables. In this case, most of our routers had 6 or fewer routes in them to begin with. This was replaced with an array that's searched linearly for the route. We more heavily rely on AO2 objects in Asterisk 12, and the memset() in ao2_ref() actually became noticeable on the profile. This was #ifdef'ed to only run when AO2_DEBUG was enabled. After being misled by an erroneous comment in taskprocessor.c during profiling, the wrong comment was removed. Review: https://reviewboard.asterisk.org/r/2873/ ........ r400178 | dlee | 2013-09-30 13:26:27 -0500 (Mon, 30 Sep 2013) | 24 lines Taskprocessor optimization; switch Stasis to use taskprocessors This patch optimizes taskprocessor to use a semaphore for signaling, which the OS can do a better job at managing contention and waiting that we can with a mutex and condition. The taskprocessor execution was also slightly optimized to reduce the number of locks taken. The only observable difference in the taskprocessor implementation is that when the final reference to the taskprocessor goes away, it will execute all tasks to completion instead of discarding the unexecuted tasks. For systems where unnamed semaphores are not supported, a really simple semaphore implementation is provided. (Which gives identical performance as the original taskprocessor implementation). The way we ended up implementing Stasis caused the threadpool to be a burden instead of a boost to performance. This was switched to just use taskprocessors directly for subscriptions. Review: https://reviewboard.asterisk.org/r/2881/ ........ r400180 | dlee | 2013-09-30 13:39:34 -0500 (Mon, 30 Sep 2013) | 28 lines Optimize how Stasis forwards are dispatched This patch optimizes how forwards are dispatched in Stasis. Originally, forwards were dispatched as subscriptions that are invoked on the publishing thread. This did not account for the vast number of forwards we would end up having in the system, and the amount of work it would take to walk though the forward subscriptions. This patch modifies Stasis so that rather than walking the tree of forwards on every dispatch, when forwards and subscriptions are changed, the subscriber list for every topic in the tree is changed. This has a couple of benefits. First, this reduces the workload of dispatching messages. It also reduces contention when dispatching to different topics that happen to forward to the same aggregation topic (as happens with all of the channel, bridge and endpoint topics). Since forwards are no longer subscriptions, the bulk of this patch is simply changing stasis_subscription objects to stasis_forward objects (which, admittedly, I should have done in the first place.) Since this required me to yet again put in a growing array, I finally abstracted that out into a set of ast_vector macros in asterisk/vector.h. Review: https://reviewboard.asterisk.org/r/2883/ ........ r400181 | dlee | 2013-09-30 13:48:57 -0500 (Mon, 30 Sep 2013) | 28 lines Remove dispatch object allocation from Stasis publishing While looking for areas for performance improvement, I realized that an unused feature in Stasis was negatively impacting performance. When a message is sent to a subscriber, a dispatch object is allocated for the dispatch, containing the topic the message was published to, the subscriber the message is being sent to, and the message itself. The topic is actually unused by any subscriber in Asterisk today. And the subscriber is associated with the taskprocessor the message is being dispatched to. First, this patch removes the unused topic parameter from Stasis subscription callbacks. Second, this patch introduces the concept of taskprocessor local data, data that may be set on a taskprocessor and provided along with the data pointer when a task is pushed using the ast_taskprocessor_push_local() call. This allows the task to have both data specific to that taskprocessor, in addition to data specific to that invocation. With those two changes, the dispatch object can be removed completely, and the message is simply refcounted and sent directly to the taskprocessor. Review: https://reviewboard.asterisk.org/r/2884/ ........ Merged revisions 399887,400138,400178,400180-400181 from http://svn.asterisk.org/svn/asterisk/branches/12 git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@400186 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-09-30 18:55:27 +00:00
/* While a linear search for routes may seem very inefficient, most
* route tables have six routes or less. For such small data, it's
* hard to beat a linear search. If we start having larger route
* tables, then we can look into containers with more efficient
* lookups.
*/
for (idx = 0; idx < AST_VECTOR_SIZE(table); ++idx) {
route = AST_VECTOR_GET_ADDR(table, idx);
if (route->message_type == message_type) {
return route;
Multiple revisions 399887,400138,400178,400180-400181 ........ r399887 | dlee | 2013-09-26 10:41:47 -0500 (Thu, 26 Sep 2013) | 1 line Minor performance bump by not allocate manager variable struct if we don't need it ........ r400138 | dlee | 2013-09-30 10:24:00 -0500 (Mon, 30 Sep 2013) | 23 lines Stasis performance improvements This patch addresses several performance problems that were found in the initial performance testing of Asterisk 12. The Stasis dispatch object was allocated as an AO2 object, even though it has a very confined lifecycle. This was replaced with a straight ast_malloc(). The Stasis message router was spending an inordinate amount of time searching hash tables. In this case, most of our routers had 6 or fewer routes in them to begin with. This was replaced with an array that's searched linearly for the route. We more heavily rely on AO2 objects in Asterisk 12, and the memset() in ao2_ref() actually became noticeable on the profile. This was #ifdef'ed to only run when AO2_DEBUG was enabled. After being misled by an erroneous comment in taskprocessor.c during profiling, the wrong comment was removed. Review: https://reviewboard.asterisk.org/r/2873/ ........ r400178 | dlee | 2013-09-30 13:26:27 -0500 (Mon, 30 Sep 2013) | 24 lines Taskprocessor optimization; switch Stasis to use taskprocessors This patch optimizes taskprocessor to use a semaphore for signaling, which the OS can do a better job at managing contention and waiting that we can with a mutex and condition. The taskprocessor execution was also slightly optimized to reduce the number of locks taken. The only observable difference in the taskprocessor implementation is that when the final reference to the taskprocessor goes away, it will execute all tasks to completion instead of discarding the unexecuted tasks. For systems where unnamed semaphores are not supported, a really simple semaphore implementation is provided. (Which gives identical performance as the original taskprocessor implementation). The way we ended up implementing Stasis caused the threadpool to be a burden instead of a boost to performance. This was switched to just use taskprocessors directly for subscriptions. Review: https://reviewboard.asterisk.org/r/2881/ ........ r400180 | dlee | 2013-09-30 13:39:34 -0500 (Mon, 30 Sep 2013) | 28 lines Optimize how Stasis forwards are dispatched This patch optimizes how forwards are dispatched in Stasis. Originally, forwards were dispatched as subscriptions that are invoked on the publishing thread. This did not account for the vast number of forwards we would end up having in the system, and the amount of work it would take to walk though the forward subscriptions. This patch modifies Stasis so that rather than walking the tree of forwards on every dispatch, when forwards and subscriptions are changed, the subscriber list for every topic in the tree is changed. This has a couple of benefits. First, this reduces the workload of dispatching messages. It also reduces contention when dispatching to different topics that happen to forward to the same aggregation topic (as happens with all of the channel, bridge and endpoint topics). Since forwards are no longer subscriptions, the bulk of this patch is simply changing stasis_subscription objects to stasis_forward objects (which, admittedly, I should have done in the first place.) Since this required me to yet again put in a growing array, I finally abstracted that out into a set of ast_vector macros in asterisk/vector.h. Review: https://reviewboard.asterisk.org/r/2883/ ........ r400181 | dlee | 2013-09-30 13:48:57 -0500 (Mon, 30 Sep 2013) | 28 lines Remove dispatch object allocation from Stasis publishing While looking for areas for performance improvement, I realized that an unused feature in Stasis was negatively impacting performance. When a message is sent to a subscriber, a dispatch object is allocated for the dispatch, containing the topic the message was published to, the subscriber the message is being sent to, and the message itself. The topic is actually unused by any subscriber in Asterisk today. And the subscriber is associated with the taskprocessor the message is being dispatched to. First, this patch removes the unused topic parameter from Stasis subscription callbacks. Second, this patch introduces the concept of taskprocessor local data, data that may be set on a taskprocessor and provided along with the data pointer when a task is pushed using the ast_taskprocessor_push_local() call. This allows the task to have both data specific to that taskprocessor, in addition to data specific to that invocation. With those two changes, the dispatch object can be removed completely, and the message is simply refcounted and sent directly to the taskprocessor. Review: https://reviewboard.asterisk.org/r/2884/ ........ Merged revisions 399887,400138,400178,400180-400181 from http://svn.asterisk.org/svn/asterisk/branches/12 git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@400186 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-09-30 18:55:27 +00:00
}
}
Multiple revisions 399887,400138,400178,400180-400181 ........ r399887 | dlee | 2013-09-26 10:41:47 -0500 (Thu, 26 Sep 2013) | 1 line Minor performance bump by not allocate manager variable struct if we don't need it ........ r400138 | dlee | 2013-09-30 10:24:00 -0500 (Mon, 30 Sep 2013) | 23 lines Stasis performance improvements This patch addresses several performance problems that were found in the initial performance testing of Asterisk 12. The Stasis dispatch object was allocated as an AO2 object, even though it has a very confined lifecycle. This was replaced with a straight ast_malloc(). The Stasis message router was spending an inordinate amount of time searching hash tables. In this case, most of our routers had 6 or fewer routes in them to begin with. This was replaced with an array that's searched linearly for the route. We more heavily rely on AO2 objects in Asterisk 12, and the memset() in ao2_ref() actually became noticeable on the profile. This was #ifdef'ed to only run when AO2_DEBUG was enabled. After being misled by an erroneous comment in taskprocessor.c during profiling, the wrong comment was removed. Review: https://reviewboard.asterisk.org/r/2873/ ........ r400178 | dlee | 2013-09-30 13:26:27 -0500 (Mon, 30 Sep 2013) | 24 lines Taskprocessor optimization; switch Stasis to use taskprocessors This patch optimizes taskprocessor to use a semaphore for signaling, which the OS can do a better job at managing contention and waiting that we can with a mutex and condition. The taskprocessor execution was also slightly optimized to reduce the number of locks taken. The only observable difference in the taskprocessor implementation is that when the final reference to the taskprocessor goes away, it will execute all tasks to completion instead of discarding the unexecuted tasks. For systems where unnamed semaphores are not supported, a really simple semaphore implementation is provided. (Which gives identical performance as the original taskprocessor implementation). The way we ended up implementing Stasis caused the threadpool to be a burden instead of a boost to performance. This was switched to just use taskprocessors directly for subscriptions. Review: https://reviewboard.asterisk.org/r/2881/ ........ r400180 | dlee | 2013-09-30 13:39:34 -0500 (Mon, 30 Sep 2013) | 28 lines Optimize how Stasis forwards are dispatched This patch optimizes how forwards are dispatched in Stasis. Originally, forwards were dispatched as subscriptions that are invoked on the publishing thread. This did not account for the vast number of forwards we would end up having in the system, and the amount of work it would take to walk though the forward subscriptions. This patch modifies Stasis so that rather than walking the tree of forwards on every dispatch, when forwards and subscriptions are changed, the subscriber list for every topic in the tree is changed. This has a couple of benefits. First, this reduces the workload of dispatching messages. It also reduces contention when dispatching to different topics that happen to forward to the same aggregation topic (as happens with all of the channel, bridge and endpoint topics). Since forwards are no longer subscriptions, the bulk of this patch is simply changing stasis_subscription objects to stasis_forward objects (which, admittedly, I should have done in the first place.) Since this required me to yet again put in a growing array, I finally abstracted that out into a set of ast_vector macros in asterisk/vector.h. Review: https://reviewboard.asterisk.org/r/2883/ ........ r400181 | dlee | 2013-09-30 13:48:57 -0500 (Mon, 30 Sep 2013) | 28 lines Remove dispatch object allocation from Stasis publishing While looking for areas for performance improvement, I realized that an unused feature in Stasis was negatively impacting performance. When a message is sent to a subscriber, a dispatch object is allocated for the dispatch, containing the topic the message was published to, the subscriber the message is being sent to, and the message itself. The topic is actually unused by any subscriber in Asterisk today. And the subscriber is associated with the taskprocessor the message is being dispatched to. First, this patch removes the unused topic parameter from Stasis subscription callbacks. Second, this patch introduces the concept of taskprocessor local data, data that may be set on a taskprocessor and provided along with the data pointer when a task is pushed using the ast_taskprocessor_push_local() call. This allows the task to have both data specific to that taskprocessor, in addition to data specific to that invocation. With those two changes, the dispatch object can be removed completely, and the message is simply refcounted and sent directly to the taskprocessor. Review: https://reviewboard.asterisk.org/r/2884/ ........ Merged revisions 399887,400138,400178,400180-400181 from http://svn.asterisk.org/svn/asterisk/branches/12 git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@400186 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-09-30 18:55:27 +00:00
return NULL;
}
/*!
* \brief route_table comparator for AST_VECTOR_REMOVE_CMP_UNORDERED()
*
* \param elem Element to compare against
* \param value Value to compare with the vector element.
*
* \return 0 if element does not match.
* \return Non-zero if element matches.
*/
#define ROUTE_TABLE_ELEM_CMP(elem, value) ((elem).message_type == (value))
/*!
* \brief route_table vector element cleanup.
*
* \param elem Element to cleanup
*
* \return Nothing
*/
#define ROUTE_TABLE_ELEM_CLEANUP(elem) ao2_cleanup((elem).message_type)
static int route_table_remove(struct route_table *table,
struct stasis_message_type *message_type)
{
return AST_VECTOR_REMOVE_CMP_UNORDERED(table, message_type, ROUTE_TABLE_ELEM_CMP,
ROUTE_TABLE_ELEM_CLEANUP);
}
static int route_table_add(struct route_table *table,
Multiple revisions 399887,400138,400178,400180-400181 ........ r399887 | dlee | 2013-09-26 10:41:47 -0500 (Thu, 26 Sep 2013) | 1 line Minor performance bump by not allocate manager variable struct if we don't need it ........ r400138 | dlee | 2013-09-30 10:24:00 -0500 (Mon, 30 Sep 2013) | 23 lines Stasis performance improvements This patch addresses several performance problems that were found in the initial performance testing of Asterisk 12. The Stasis dispatch object was allocated as an AO2 object, even though it has a very confined lifecycle. This was replaced with a straight ast_malloc(). The Stasis message router was spending an inordinate amount of time searching hash tables. In this case, most of our routers had 6 or fewer routes in them to begin with. This was replaced with an array that's searched linearly for the route. We more heavily rely on AO2 objects in Asterisk 12, and the memset() in ao2_ref() actually became noticeable on the profile. This was #ifdef'ed to only run when AO2_DEBUG was enabled. After being misled by an erroneous comment in taskprocessor.c during profiling, the wrong comment was removed. Review: https://reviewboard.asterisk.org/r/2873/ ........ r400178 | dlee | 2013-09-30 13:26:27 -0500 (Mon, 30 Sep 2013) | 24 lines Taskprocessor optimization; switch Stasis to use taskprocessors This patch optimizes taskprocessor to use a semaphore for signaling, which the OS can do a better job at managing contention and waiting that we can with a mutex and condition. The taskprocessor execution was also slightly optimized to reduce the number of locks taken. The only observable difference in the taskprocessor implementation is that when the final reference to the taskprocessor goes away, it will execute all tasks to completion instead of discarding the unexecuted tasks. For systems where unnamed semaphores are not supported, a really simple semaphore implementation is provided. (Which gives identical performance as the original taskprocessor implementation). The way we ended up implementing Stasis caused the threadpool to be a burden instead of a boost to performance. This was switched to just use taskprocessors directly for subscriptions. Review: https://reviewboard.asterisk.org/r/2881/ ........ r400180 | dlee | 2013-09-30 13:39:34 -0500 (Mon, 30 Sep 2013) | 28 lines Optimize how Stasis forwards are dispatched This patch optimizes how forwards are dispatched in Stasis. Originally, forwards were dispatched as subscriptions that are invoked on the publishing thread. This did not account for the vast number of forwards we would end up having in the system, and the amount of work it would take to walk though the forward subscriptions. This patch modifies Stasis so that rather than walking the tree of forwards on every dispatch, when forwards and subscriptions are changed, the subscriber list for every topic in the tree is changed. This has a couple of benefits. First, this reduces the workload of dispatching messages. It also reduces contention when dispatching to different topics that happen to forward to the same aggregation topic (as happens with all of the channel, bridge and endpoint topics). Since forwards are no longer subscriptions, the bulk of this patch is simply changing stasis_subscription objects to stasis_forward objects (which, admittedly, I should have done in the first place.) Since this required me to yet again put in a growing array, I finally abstracted that out into a set of ast_vector macros in asterisk/vector.h. Review: https://reviewboard.asterisk.org/r/2883/ ........ r400181 | dlee | 2013-09-30 13:48:57 -0500 (Mon, 30 Sep 2013) | 28 lines Remove dispatch object allocation from Stasis publishing While looking for areas for performance improvement, I realized that an unused feature in Stasis was negatively impacting performance. When a message is sent to a subscriber, a dispatch object is allocated for the dispatch, containing the topic the message was published to, the subscriber the message is being sent to, and the message itself. The topic is actually unused by any subscriber in Asterisk today. And the subscriber is associated with the taskprocessor the message is being dispatched to. First, this patch removes the unused topic parameter from Stasis subscription callbacks. Second, this patch introduces the concept of taskprocessor local data, data that may be set on a taskprocessor and provided along with the data pointer when a task is pushed using the ast_taskprocessor_push_local() call. This allows the task to have both data specific to that taskprocessor, in addition to data specific to that invocation. With those two changes, the dispatch object can be removed completely, and the message is simply refcounted and sent directly to the taskprocessor. Review: https://reviewboard.asterisk.org/r/2884/ ........ Merged revisions 399887,400138,400178,400180-400181 from http://svn.asterisk.org/svn/asterisk/branches/12 git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@400186 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-09-30 18:55:27 +00:00
struct stasis_message_type *message_type,
stasis_subscription_cb callback, void *data)
{
struct stasis_message_route route;
int res;
ast_assert(callback != NULL);
ast_assert(route_table_find(table, message_type) == NULL);
Multiple revisions 399887,400138,400178,400180-400181 ........ r399887 | dlee | 2013-09-26 10:41:47 -0500 (Thu, 26 Sep 2013) | 1 line Minor performance bump by not allocate manager variable struct if we don't need it ........ r400138 | dlee | 2013-09-30 10:24:00 -0500 (Mon, 30 Sep 2013) | 23 lines Stasis performance improvements This patch addresses several performance problems that were found in the initial performance testing of Asterisk 12. The Stasis dispatch object was allocated as an AO2 object, even though it has a very confined lifecycle. This was replaced with a straight ast_malloc(). The Stasis message router was spending an inordinate amount of time searching hash tables. In this case, most of our routers had 6 or fewer routes in them to begin with. This was replaced with an array that's searched linearly for the route. We more heavily rely on AO2 objects in Asterisk 12, and the memset() in ao2_ref() actually became noticeable on the profile. This was #ifdef'ed to only run when AO2_DEBUG was enabled. After being misled by an erroneous comment in taskprocessor.c during profiling, the wrong comment was removed. Review: https://reviewboard.asterisk.org/r/2873/ ........ r400178 | dlee | 2013-09-30 13:26:27 -0500 (Mon, 30 Sep 2013) | 24 lines Taskprocessor optimization; switch Stasis to use taskprocessors This patch optimizes taskprocessor to use a semaphore for signaling, which the OS can do a better job at managing contention and waiting that we can with a mutex and condition. The taskprocessor execution was also slightly optimized to reduce the number of locks taken. The only observable difference in the taskprocessor implementation is that when the final reference to the taskprocessor goes away, it will execute all tasks to completion instead of discarding the unexecuted tasks. For systems where unnamed semaphores are not supported, a really simple semaphore implementation is provided. (Which gives identical performance as the original taskprocessor implementation). The way we ended up implementing Stasis caused the threadpool to be a burden instead of a boost to performance. This was switched to just use taskprocessors directly for subscriptions. Review: https://reviewboard.asterisk.org/r/2881/ ........ r400180 | dlee | 2013-09-30 13:39:34 -0500 (Mon, 30 Sep 2013) | 28 lines Optimize how Stasis forwards are dispatched This patch optimizes how forwards are dispatched in Stasis. Originally, forwards were dispatched as subscriptions that are invoked on the publishing thread. This did not account for the vast number of forwards we would end up having in the system, and the amount of work it would take to walk though the forward subscriptions. This patch modifies Stasis so that rather than walking the tree of forwards on every dispatch, when forwards and subscriptions are changed, the subscriber list for every topic in the tree is changed. This has a couple of benefits. First, this reduces the workload of dispatching messages. It also reduces contention when dispatching to different topics that happen to forward to the same aggregation topic (as happens with all of the channel, bridge and endpoint topics). Since forwards are no longer subscriptions, the bulk of this patch is simply changing stasis_subscription objects to stasis_forward objects (which, admittedly, I should have done in the first place.) Since this required me to yet again put in a growing array, I finally abstracted that out into a set of ast_vector macros in asterisk/vector.h. Review: https://reviewboard.asterisk.org/r/2883/ ........ r400181 | dlee | 2013-09-30 13:48:57 -0500 (Mon, 30 Sep 2013) | 28 lines Remove dispatch object allocation from Stasis publishing While looking for areas for performance improvement, I realized that an unused feature in Stasis was negatively impacting performance. When a message is sent to a subscriber, a dispatch object is allocated for the dispatch, containing the topic the message was published to, the subscriber the message is being sent to, and the message itself. The topic is actually unused by any subscriber in Asterisk today. And the subscriber is associated with the taskprocessor the message is being dispatched to. First, this patch removes the unused topic parameter from Stasis subscription callbacks. Second, this patch introduces the concept of taskprocessor local data, data that may be set on a taskprocessor and provided along with the data pointer when a task is pushed using the ast_taskprocessor_push_local() call. This allows the task to have both data specific to that taskprocessor, in addition to data specific to that invocation. With those two changes, the dispatch object can be removed completely, and the message is simply refcounted and sent directly to the taskprocessor. Review: https://reviewboard.asterisk.org/r/2884/ ........ Merged revisions 399887,400138,400178,400180-400181 from http://svn.asterisk.org/svn/asterisk/branches/12 git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@400186 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-09-30 18:55:27 +00:00
route.message_type = ao2_bump(message_type);
route.callback = callback;
route.data = data;
Multiple revisions 399887,400138,400178,400180-400181 ........ r399887 | dlee | 2013-09-26 10:41:47 -0500 (Thu, 26 Sep 2013) | 1 line Minor performance bump by not allocate manager variable struct if we don't need it ........ r400138 | dlee | 2013-09-30 10:24:00 -0500 (Mon, 30 Sep 2013) | 23 lines Stasis performance improvements This patch addresses several performance problems that were found in the initial performance testing of Asterisk 12. The Stasis dispatch object was allocated as an AO2 object, even though it has a very confined lifecycle. This was replaced with a straight ast_malloc(). The Stasis message router was spending an inordinate amount of time searching hash tables. In this case, most of our routers had 6 or fewer routes in them to begin with. This was replaced with an array that's searched linearly for the route. We more heavily rely on AO2 objects in Asterisk 12, and the memset() in ao2_ref() actually became noticeable on the profile. This was #ifdef'ed to only run when AO2_DEBUG was enabled. After being misled by an erroneous comment in taskprocessor.c during profiling, the wrong comment was removed. Review: https://reviewboard.asterisk.org/r/2873/ ........ r400178 | dlee | 2013-09-30 13:26:27 -0500 (Mon, 30 Sep 2013) | 24 lines Taskprocessor optimization; switch Stasis to use taskprocessors This patch optimizes taskprocessor to use a semaphore for signaling, which the OS can do a better job at managing contention and waiting that we can with a mutex and condition. The taskprocessor execution was also slightly optimized to reduce the number of locks taken. The only observable difference in the taskprocessor implementation is that when the final reference to the taskprocessor goes away, it will execute all tasks to completion instead of discarding the unexecuted tasks. For systems where unnamed semaphores are not supported, a really simple semaphore implementation is provided. (Which gives identical performance as the original taskprocessor implementation). The way we ended up implementing Stasis caused the threadpool to be a burden instead of a boost to performance. This was switched to just use taskprocessors directly for subscriptions. Review: https://reviewboard.asterisk.org/r/2881/ ........ r400180 | dlee | 2013-09-30 13:39:34 -0500 (Mon, 30 Sep 2013) | 28 lines Optimize how Stasis forwards are dispatched This patch optimizes how forwards are dispatched in Stasis. Originally, forwards were dispatched as subscriptions that are invoked on the publishing thread. This did not account for the vast number of forwards we would end up having in the system, and the amount of work it would take to walk though the forward subscriptions. This patch modifies Stasis so that rather than walking the tree of forwards on every dispatch, when forwards and subscriptions are changed, the subscriber list for every topic in the tree is changed. This has a couple of benefits. First, this reduces the workload of dispatching messages. It also reduces contention when dispatching to different topics that happen to forward to the same aggregation topic (as happens with all of the channel, bridge and endpoint topics). Since forwards are no longer subscriptions, the bulk of this patch is simply changing stasis_subscription objects to stasis_forward objects (which, admittedly, I should have done in the first place.) Since this required me to yet again put in a growing array, I finally abstracted that out into a set of ast_vector macros in asterisk/vector.h. Review: https://reviewboard.asterisk.org/r/2883/ ........ r400181 | dlee | 2013-09-30 13:48:57 -0500 (Mon, 30 Sep 2013) | 28 lines Remove dispatch object allocation from Stasis publishing While looking for areas for performance improvement, I realized that an unused feature in Stasis was negatively impacting performance. When a message is sent to a subscriber, a dispatch object is allocated for the dispatch, containing the topic the message was published to, the subscriber the message is being sent to, and the message itself. The topic is actually unused by any subscriber in Asterisk today. And the subscriber is associated with the taskprocessor the message is being dispatched to. First, this patch removes the unused topic parameter from Stasis subscription callbacks. Second, this patch introduces the concept of taskprocessor local data, data that may be set on a taskprocessor and provided along with the data pointer when a task is pushed using the ast_taskprocessor_push_local() call. This allows the task to have both data specific to that taskprocessor, in addition to data specific to that invocation. With those two changes, the dispatch object can be removed completely, and the message is simply refcounted and sent directly to the taskprocessor. Review: https://reviewboard.asterisk.org/r/2884/ ........ Merged revisions 399887,400138,400178,400180-400181 from http://svn.asterisk.org/svn/asterisk/branches/12 git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@400186 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-09-30 18:55:27 +00:00
res = AST_VECTOR_APPEND(table, route);
if (res) {
ROUTE_TABLE_ELEM_CLEANUP(route);
}
return res;
}
static void route_table_dtor(struct route_table *table)
{
Multiple revisions 399887,400138,400178,400180-400181 ........ r399887 | dlee | 2013-09-26 10:41:47 -0500 (Thu, 26 Sep 2013) | 1 line Minor performance bump by not allocate manager variable struct if we don't need it ........ r400138 | dlee | 2013-09-30 10:24:00 -0500 (Mon, 30 Sep 2013) | 23 lines Stasis performance improvements This patch addresses several performance problems that were found in the initial performance testing of Asterisk 12. The Stasis dispatch object was allocated as an AO2 object, even though it has a very confined lifecycle. This was replaced with a straight ast_malloc(). The Stasis message router was spending an inordinate amount of time searching hash tables. In this case, most of our routers had 6 or fewer routes in them to begin with. This was replaced with an array that's searched linearly for the route. We more heavily rely on AO2 objects in Asterisk 12, and the memset() in ao2_ref() actually became noticeable on the profile. This was #ifdef'ed to only run when AO2_DEBUG was enabled. After being misled by an erroneous comment in taskprocessor.c during profiling, the wrong comment was removed. Review: https://reviewboard.asterisk.org/r/2873/ ........ r400178 | dlee | 2013-09-30 13:26:27 -0500 (Mon, 30 Sep 2013) | 24 lines Taskprocessor optimization; switch Stasis to use taskprocessors This patch optimizes taskprocessor to use a semaphore for signaling, which the OS can do a better job at managing contention and waiting that we can with a mutex and condition. The taskprocessor execution was also slightly optimized to reduce the number of locks taken. The only observable difference in the taskprocessor implementation is that when the final reference to the taskprocessor goes away, it will execute all tasks to completion instead of discarding the unexecuted tasks. For systems where unnamed semaphores are not supported, a really simple semaphore implementation is provided. (Which gives identical performance as the original taskprocessor implementation). The way we ended up implementing Stasis caused the threadpool to be a burden instead of a boost to performance. This was switched to just use taskprocessors directly for subscriptions. Review: https://reviewboard.asterisk.org/r/2881/ ........ r400180 | dlee | 2013-09-30 13:39:34 -0500 (Mon, 30 Sep 2013) | 28 lines Optimize how Stasis forwards are dispatched This patch optimizes how forwards are dispatched in Stasis. Originally, forwards were dispatched as subscriptions that are invoked on the publishing thread. This did not account for the vast number of forwards we would end up having in the system, and the amount of work it would take to walk though the forward subscriptions. This patch modifies Stasis so that rather than walking the tree of forwards on every dispatch, when forwards and subscriptions are changed, the subscriber list for every topic in the tree is changed. This has a couple of benefits. First, this reduces the workload of dispatching messages. It also reduces contention when dispatching to different topics that happen to forward to the same aggregation topic (as happens with all of the channel, bridge and endpoint topics). Since forwards are no longer subscriptions, the bulk of this patch is simply changing stasis_subscription objects to stasis_forward objects (which, admittedly, I should have done in the first place.) Since this required me to yet again put in a growing array, I finally abstracted that out into a set of ast_vector macros in asterisk/vector.h. Review: https://reviewboard.asterisk.org/r/2883/ ........ r400181 | dlee | 2013-09-30 13:48:57 -0500 (Mon, 30 Sep 2013) | 28 lines Remove dispatch object allocation from Stasis publishing While looking for areas for performance improvement, I realized that an unused feature in Stasis was negatively impacting performance. When a message is sent to a subscriber, a dispatch object is allocated for the dispatch, containing the topic the message was published to, the subscriber the message is being sent to, and the message itself. The topic is actually unused by any subscriber in Asterisk today. And the subscriber is associated with the taskprocessor the message is being dispatched to. First, this patch removes the unused topic parameter from Stasis subscription callbacks. Second, this patch introduces the concept of taskprocessor local data, data that may be set on a taskprocessor and provided along with the data pointer when a task is pushed using the ast_taskprocessor_push_local() call. This allows the task to have both data specific to that taskprocessor, in addition to data specific to that invocation. With those two changes, the dispatch object can be removed completely, and the message is simply refcounted and sent directly to the taskprocessor. Review: https://reviewboard.asterisk.org/r/2884/ ........ Merged revisions 399887,400138,400178,400180-400181 from http://svn.asterisk.org/svn/asterisk/branches/12 git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@400186 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-09-30 18:55:27 +00:00
size_t idx;
struct stasis_message_route *route;
Multiple revisions 399887,400138,400178,400180-400181 ........ r399887 | dlee | 2013-09-26 10:41:47 -0500 (Thu, 26 Sep 2013) | 1 line Minor performance bump by not allocate manager variable struct if we don't need it ........ r400138 | dlee | 2013-09-30 10:24:00 -0500 (Mon, 30 Sep 2013) | 23 lines Stasis performance improvements This patch addresses several performance problems that were found in the initial performance testing of Asterisk 12. The Stasis dispatch object was allocated as an AO2 object, even though it has a very confined lifecycle. This was replaced with a straight ast_malloc(). The Stasis message router was spending an inordinate amount of time searching hash tables. In this case, most of our routers had 6 or fewer routes in them to begin with. This was replaced with an array that's searched linearly for the route. We more heavily rely on AO2 objects in Asterisk 12, and the memset() in ao2_ref() actually became noticeable on the profile. This was #ifdef'ed to only run when AO2_DEBUG was enabled. After being misled by an erroneous comment in taskprocessor.c during profiling, the wrong comment was removed. Review: https://reviewboard.asterisk.org/r/2873/ ........ r400178 | dlee | 2013-09-30 13:26:27 -0500 (Mon, 30 Sep 2013) | 24 lines Taskprocessor optimization; switch Stasis to use taskprocessors This patch optimizes taskprocessor to use a semaphore for signaling, which the OS can do a better job at managing contention and waiting that we can with a mutex and condition. The taskprocessor execution was also slightly optimized to reduce the number of locks taken. The only observable difference in the taskprocessor implementation is that when the final reference to the taskprocessor goes away, it will execute all tasks to completion instead of discarding the unexecuted tasks. For systems where unnamed semaphores are not supported, a really simple semaphore implementation is provided. (Which gives identical performance as the original taskprocessor implementation). The way we ended up implementing Stasis caused the threadpool to be a burden instead of a boost to performance. This was switched to just use taskprocessors directly for subscriptions. Review: https://reviewboard.asterisk.org/r/2881/ ........ r400180 | dlee | 2013-09-30 13:39:34 -0500 (Mon, 30 Sep 2013) | 28 lines Optimize how Stasis forwards are dispatched This patch optimizes how forwards are dispatched in Stasis. Originally, forwards were dispatched as subscriptions that are invoked on the publishing thread. This did not account for the vast number of forwards we would end up having in the system, and the amount of work it would take to walk though the forward subscriptions. This patch modifies Stasis so that rather than walking the tree of forwards on every dispatch, when forwards and subscriptions are changed, the subscriber list for every topic in the tree is changed. This has a couple of benefits. First, this reduces the workload of dispatching messages. It also reduces contention when dispatching to different topics that happen to forward to the same aggregation topic (as happens with all of the channel, bridge and endpoint topics). Since forwards are no longer subscriptions, the bulk of this patch is simply changing stasis_subscription objects to stasis_forward objects (which, admittedly, I should have done in the first place.) Since this required me to yet again put in a growing array, I finally abstracted that out into a set of ast_vector macros in asterisk/vector.h. Review: https://reviewboard.asterisk.org/r/2883/ ........ r400181 | dlee | 2013-09-30 13:48:57 -0500 (Mon, 30 Sep 2013) | 28 lines Remove dispatch object allocation from Stasis publishing While looking for areas for performance improvement, I realized that an unused feature in Stasis was negatively impacting performance. When a message is sent to a subscriber, a dispatch object is allocated for the dispatch, containing the topic the message was published to, the subscriber the message is being sent to, and the message itself. The topic is actually unused by any subscriber in Asterisk today. And the subscriber is associated with the taskprocessor the message is being dispatched to. First, this patch removes the unused topic parameter from Stasis subscription callbacks. Second, this patch introduces the concept of taskprocessor local data, data that may be set on a taskprocessor and provided along with the data pointer when a task is pushed using the ast_taskprocessor_push_local() call. This allows the task to have both data specific to that taskprocessor, in addition to data specific to that invocation. With those two changes, the dispatch object can be removed completely, and the message is simply refcounted and sent directly to the taskprocessor. Review: https://reviewboard.asterisk.org/r/2884/ ........ Merged revisions 399887,400138,400178,400180-400181 from http://svn.asterisk.org/svn/asterisk/branches/12 git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@400186 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-09-30 18:55:27 +00:00
for (idx = 0; idx < AST_VECTOR_SIZE(table); ++idx) {
route = AST_VECTOR_GET_ADDR(table, idx);
ROUTE_TABLE_ELEM_CLEANUP(*route);
Multiple revisions 399887,400138,400178,400180-400181 ........ r399887 | dlee | 2013-09-26 10:41:47 -0500 (Thu, 26 Sep 2013) | 1 line Minor performance bump by not allocate manager variable struct if we don't need it ........ r400138 | dlee | 2013-09-30 10:24:00 -0500 (Mon, 30 Sep 2013) | 23 lines Stasis performance improvements This patch addresses several performance problems that were found in the initial performance testing of Asterisk 12. The Stasis dispatch object was allocated as an AO2 object, even though it has a very confined lifecycle. This was replaced with a straight ast_malloc(). The Stasis message router was spending an inordinate amount of time searching hash tables. In this case, most of our routers had 6 or fewer routes in them to begin with. This was replaced with an array that's searched linearly for the route. We more heavily rely on AO2 objects in Asterisk 12, and the memset() in ao2_ref() actually became noticeable on the profile. This was #ifdef'ed to only run when AO2_DEBUG was enabled. After being misled by an erroneous comment in taskprocessor.c during profiling, the wrong comment was removed. Review: https://reviewboard.asterisk.org/r/2873/ ........ r400178 | dlee | 2013-09-30 13:26:27 -0500 (Mon, 30 Sep 2013) | 24 lines Taskprocessor optimization; switch Stasis to use taskprocessors This patch optimizes taskprocessor to use a semaphore for signaling, which the OS can do a better job at managing contention and waiting that we can with a mutex and condition. The taskprocessor execution was also slightly optimized to reduce the number of locks taken. The only observable difference in the taskprocessor implementation is that when the final reference to the taskprocessor goes away, it will execute all tasks to completion instead of discarding the unexecuted tasks. For systems where unnamed semaphores are not supported, a really simple semaphore implementation is provided. (Which gives identical performance as the original taskprocessor implementation). The way we ended up implementing Stasis caused the threadpool to be a burden instead of a boost to performance. This was switched to just use taskprocessors directly for subscriptions. Review: https://reviewboard.asterisk.org/r/2881/ ........ r400180 | dlee | 2013-09-30 13:39:34 -0500 (Mon, 30 Sep 2013) | 28 lines Optimize how Stasis forwards are dispatched This patch optimizes how forwards are dispatched in Stasis. Originally, forwards were dispatched as subscriptions that are invoked on the publishing thread. This did not account for the vast number of forwards we would end up having in the system, and the amount of work it would take to walk though the forward subscriptions. This patch modifies Stasis so that rather than walking the tree of forwards on every dispatch, when forwards and subscriptions are changed, the subscriber list for every topic in the tree is changed. This has a couple of benefits. First, this reduces the workload of dispatching messages. It also reduces contention when dispatching to different topics that happen to forward to the same aggregation topic (as happens with all of the channel, bridge and endpoint topics). Since forwards are no longer subscriptions, the bulk of this patch is simply changing stasis_subscription objects to stasis_forward objects (which, admittedly, I should have done in the first place.) Since this required me to yet again put in a growing array, I finally abstracted that out into a set of ast_vector macros in asterisk/vector.h. Review: https://reviewboard.asterisk.org/r/2883/ ........ r400181 | dlee | 2013-09-30 13:48:57 -0500 (Mon, 30 Sep 2013) | 28 lines Remove dispatch object allocation from Stasis publishing While looking for areas for performance improvement, I realized that an unused feature in Stasis was negatively impacting performance. When a message is sent to a subscriber, a dispatch object is allocated for the dispatch, containing the topic the message was published to, the subscriber the message is being sent to, and the message itself. The topic is actually unused by any subscriber in Asterisk today. And the subscriber is associated with the taskprocessor the message is being dispatched to. First, this patch removes the unused topic parameter from Stasis subscription callbacks. Second, this patch introduces the concept of taskprocessor local data, data that may be set on a taskprocessor and provided along with the data pointer when a task is pushed using the ast_taskprocessor_push_local() call. This allows the task to have both data specific to that taskprocessor, in addition to data specific to that invocation. With those two changes, the dispatch object can be removed completely, and the message is simply refcounted and sent directly to the taskprocessor. Review: https://reviewboard.asterisk.org/r/2884/ ........ Merged revisions 399887,400138,400178,400180-400181 from http://svn.asterisk.org/svn/asterisk/branches/12 git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@400186 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-09-30 18:55:27 +00:00
}
AST_VECTOR_FREE(table);
}
/*! \internal */
struct stasis_message_router {
/*! Subscription to the upstream topic */
struct stasis_subscription *subscription;
/*! Subscribed routes */
struct route_table routes;
Multiple revisions 399887,400138,400178,400180-400181 ........ r399887 | dlee | 2013-09-26 10:41:47 -0500 (Thu, 26 Sep 2013) | 1 line Minor performance bump by not allocate manager variable struct if we don't need it ........ r400138 | dlee | 2013-09-30 10:24:00 -0500 (Mon, 30 Sep 2013) | 23 lines Stasis performance improvements This patch addresses several performance problems that were found in the initial performance testing of Asterisk 12. The Stasis dispatch object was allocated as an AO2 object, even though it has a very confined lifecycle. This was replaced with a straight ast_malloc(). The Stasis message router was spending an inordinate amount of time searching hash tables. In this case, most of our routers had 6 or fewer routes in them to begin with. This was replaced with an array that's searched linearly for the route. We more heavily rely on AO2 objects in Asterisk 12, and the memset() in ao2_ref() actually became noticeable on the profile. This was #ifdef'ed to only run when AO2_DEBUG was enabled. After being misled by an erroneous comment in taskprocessor.c during profiling, the wrong comment was removed. Review: https://reviewboard.asterisk.org/r/2873/ ........ r400178 | dlee | 2013-09-30 13:26:27 -0500 (Mon, 30 Sep 2013) | 24 lines Taskprocessor optimization; switch Stasis to use taskprocessors This patch optimizes taskprocessor to use a semaphore for signaling, which the OS can do a better job at managing contention and waiting that we can with a mutex and condition. The taskprocessor execution was also slightly optimized to reduce the number of locks taken. The only observable difference in the taskprocessor implementation is that when the final reference to the taskprocessor goes away, it will execute all tasks to completion instead of discarding the unexecuted tasks. For systems where unnamed semaphores are not supported, a really simple semaphore implementation is provided. (Which gives identical performance as the original taskprocessor implementation). The way we ended up implementing Stasis caused the threadpool to be a burden instead of a boost to performance. This was switched to just use taskprocessors directly for subscriptions. Review: https://reviewboard.asterisk.org/r/2881/ ........ r400180 | dlee | 2013-09-30 13:39:34 -0500 (Mon, 30 Sep 2013) | 28 lines Optimize how Stasis forwards are dispatched This patch optimizes how forwards are dispatched in Stasis. Originally, forwards were dispatched as subscriptions that are invoked on the publishing thread. This did not account for the vast number of forwards we would end up having in the system, and the amount of work it would take to walk though the forward subscriptions. This patch modifies Stasis so that rather than walking the tree of forwards on every dispatch, when forwards and subscriptions are changed, the subscriber list for every topic in the tree is changed. This has a couple of benefits. First, this reduces the workload of dispatching messages. It also reduces contention when dispatching to different topics that happen to forward to the same aggregation topic (as happens with all of the channel, bridge and endpoint topics). Since forwards are no longer subscriptions, the bulk of this patch is simply changing stasis_subscription objects to stasis_forward objects (which, admittedly, I should have done in the first place.) Since this required me to yet again put in a growing array, I finally abstracted that out into a set of ast_vector macros in asterisk/vector.h. Review: https://reviewboard.asterisk.org/r/2883/ ........ r400181 | dlee | 2013-09-30 13:48:57 -0500 (Mon, 30 Sep 2013) | 28 lines Remove dispatch object allocation from Stasis publishing While looking for areas for performance improvement, I realized that an unused feature in Stasis was negatively impacting performance. When a message is sent to a subscriber, a dispatch object is allocated for the dispatch, containing the topic the message was published to, the subscriber the message is being sent to, and the message itself. The topic is actually unused by any subscriber in Asterisk today. And the subscriber is associated with the taskprocessor the message is being dispatched to. First, this patch removes the unused topic parameter from Stasis subscription callbacks. Second, this patch introduces the concept of taskprocessor local data, data that may be set on a taskprocessor and provided along with the data pointer when a task is pushed using the ast_taskprocessor_push_local() call. This allows the task to have both data specific to that taskprocessor, in addition to data specific to that invocation. With those two changes, the dispatch object can be removed completely, and the message is simply refcounted and sent directly to the taskprocessor. Review: https://reviewboard.asterisk.org/r/2884/ ........ Merged revisions 399887,400138,400178,400180-400181 from http://svn.asterisk.org/svn/asterisk/branches/12 git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@400186 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-09-30 18:55:27 +00:00
/*! Subscribed routes for \ref stasis_cache_update messages */
struct route_table cache_routes;
/*! Route of last resort */
Multiple revisions 399887,400138,400178,400180-400181 ........ r399887 | dlee | 2013-09-26 10:41:47 -0500 (Thu, 26 Sep 2013) | 1 line Minor performance bump by not allocate manager variable struct if we don't need it ........ r400138 | dlee | 2013-09-30 10:24:00 -0500 (Mon, 30 Sep 2013) | 23 lines Stasis performance improvements This patch addresses several performance problems that were found in the initial performance testing of Asterisk 12. The Stasis dispatch object was allocated as an AO2 object, even though it has a very confined lifecycle. This was replaced with a straight ast_malloc(). The Stasis message router was spending an inordinate amount of time searching hash tables. In this case, most of our routers had 6 or fewer routes in them to begin with. This was replaced with an array that's searched linearly for the route. We more heavily rely on AO2 objects in Asterisk 12, and the memset() in ao2_ref() actually became noticeable on the profile. This was #ifdef'ed to only run when AO2_DEBUG was enabled. After being misled by an erroneous comment in taskprocessor.c during profiling, the wrong comment was removed. Review: https://reviewboard.asterisk.org/r/2873/ ........ r400178 | dlee | 2013-09-30 13:26:27 -0500 (Mon, 30 Sep 2013) | 24 lines Taskprocessor optimization; switch Stasis to use taskprocessors This patch optimizes taskprocessor to use a semaphore for signaling, which the OS can do a better job at managing contention and waiting that we can with a mutex and condition. The taskprocessor execution was also slightly optimized to reduce the number of locks taken. The only observable difference in the taskprocessor implementation is that when the final reference to the taskprocessor goes away, it will execute all tasks to completion instead of discarding the unexecuted tasks. For systems where unnamed semaphores are not supported, a really simple semaphore implementation is provided. (Which gives identical performance as the original taskprocessor implementation). The way we ended up implementing Stasis caused the threadpool to be a burden instead of a boost to performance. This was switched to just use taskprocessors directly for subscriptions. Review: https://reviewboard.asterisk.org/r/2881/ ........ r400180 | dlee | 2013-09-30 13:39:34 -0500 (Mon, 30 Sep 2013) | 28 lines Optimize how Stasis forwards are dispatched This patch optimizes how forwards are dispatched in Stasis. Originally, forwards were dispatched as subscriptions that are invoked on the publishing thread. This did not account for the vast number of forwards we would end up having in the system, and the amount of work it would take to walk though the forward subscriptions. This patch modifies Stasis so that rather than walking the tree of forwards on every dispatch, when forwards and subscriptions are changed, the subscriber list for every topic in the tree is changed. This has a couple of benefits. First, this reduces the workload of dispatching messages. It also reduces contention when dispatching to different topics that happen to forward to the same aggregation topic (as happens with all of the channel, bridge and endpoint topics). Since forwards are no longer subscriptions, the bulk of this patch is simply changing stasis_subscription objects to stasis_forward objects (which, admittedly, I should have done in the first place.) Since this required me to yet again put in a growing array, I finally abstracted that out into a set of ast_vector macros in asterisk/vector.h. Review: https://reviewboard.asterisk.org/r/2883/ ........ r400181 | dlee | 2013-09-30 13:48:57 -0500 (Mon, 30 Sep 2013) | 28 lines Remove dispatch object allocation from Stasis publishing While looking for areas for performance improvement, I realized that an unused feature in Stasis was negatively impacting performance. When a message is sent to a subscriber, a dispatch object is allocated for the dispatch, containing the topic the message was published to, the subscriber the message is being sent to, and the message itself. The topic is actually unused by any subscriber in Asterisk today. And the subscriber is associated with the taskprocessor the message is being dispatched to. First, this patch removes the unused topic parameter from Stasis subscription callbacks. Second, this patch introduces the concept of taskprocessor local data, data that may be set on a taskprocessor and provided along with the data pointer when a task is pushed using the ast_taskprocessor_push_local() call. This allows the task to have both data specific to that taskprocessor, in addition to data specific to that invocation. With those two changes, the dispatch object can be removed completely, and the message is simply refcounted and sent directly to the taskprocessor. Review: https://reviewboard.asterisk.org/r/2884/ ........ Merged revisions 399887,400138,400178,400180-400181 from http://svn.asterisk.org/svn/asterisk/branches/12 git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@400186 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-09-30 18:55:27 +00:00
struct stasis_message_route default_route;
};
static void router_dtor(void *obj)
{
struct stasis_message_router *router = obj;
ast_assert(!stasis_subscription_is_subscribed(router->subscription));
ast_assert(stasis_subscription_is_done(router->subscription));
router->subscription = NULL;
route_table_dtor(&router->routes);
route_table_dtor(&router->cache_routes);
}
Multiple revisions 399887,400138,400178,400180-400181 ........ r399887 | dlee | 2013-09-26 10:41:47 -0500 (Thu, 26 Sep 2013) | 1 line Minor performance bump by not allocate manager variable struct if we don't need it ........ r400138 | dlee | 2013-09-30 10:24:00 -0500 (Mon, 30 Sep 2013) | 23 lines Stasis performance improvements This patch addresses several performance problems that were found in the initial performance testing of Asterisk 12. The Stasis dispatch object was allocated as an AO2 object, even though it has a very confined lifecycle. This was replaced with a straight ast_malloc(). The Stasis message router was spending an inordinate amount of time searching hash tables. In this case, most of our routers had 6 or fewer routes in them to begin with. This was replaced with an array that's searched linearly for the route. We more heavily rely on AO2 objects in Asterisk 12, and the memset() in ao2_ref() actually became noticeable on the profile. This was #ifdef'ed to only run when AO2_DEBUG was enabled. After being misled by an erroneous comment in taskprocessor.c during profiling, the wrong comment was removed. Review: https://reviewboard.asterisk.org/r/2873/ ........ r400178 | dlee | 2013-09-30 13:26:27 -0500 (Mon, 30 Sep 2013) | 24 lines Taskprocessor optimization; switch Stasis to use taskprocessors This patch optimizes taskprocessor to use a semaphore for signaling, which the OS can do a better job at managing contention and waiting that we can with a mutex and condition. The taskprocessor execution was also slightly optimized to reduce the number of locks taken. The only observable difference in the taskprocessor implementation is that when the final reference to the taskprocessor goes away, it will execute all tasks to completion instead of discarding the unexecuted tasks. For systems where unnamed semaphores are not supported, a really simple semaphore implementation is provided. (Which gives identical performance as the original taskprocessor implementation). The way we ended up implementing Stasis caused the threadpool to be a burden instead of a boost to performance. This was switched to just use taskprocessors directly for subscriptions. Review: https://reviewboard.asterisk.org/r/2881/ ........ r400180 | dlee | 2013-09-30 13:39:34 -0500 (Mon, 30 Sep 2013) | 28 lines Optimize how Stasis forwards are dispatched This patch optimizes how forwards are dispatched in Stasis. Originally, forwards were dispatched as subscriptions that are invoked on the publishing thread. This did not account for the vast number of forwards we would end up having in the system, and the amount of work it would take to walk though the forward subscriptions. This patch modifies Stasis so that rather than walking the tree of forwards on every dispatch, when forwards and subscriptions are changed, the subscriber list for every topic in the tree is changed. This has a couple of benefits. First, this reduces the workload of dispatching messages. It also reduces contention when dispatching to different topics that happen to forward to the same aggregation topic (as happens with all of the channel, bridge and endpoint topics). Since forwards are no longer subscriptions, the bulk of this patch is simply changing stasis_subscription objects to stasis_forward objects (which, admittedly, I should have done in the first place.) Since this required me to yet again put in a growing array, I finally abstracted that out into a set of ast_vector macros in asterisk/vector.h. Review: https://reviewboard.asterisk.org/r/2883/ ........ r400181 | dlee | 2013-09-30 13:48:57 -0500 (Mon, 30 Sep 2013) | 28 lines Remove dispatch object allocation from Stasis publishing While looking for areas for performance improvement, I realized that an unused feature in Stasis was negatively impacting performance. When a message is sent to a subscriber, a dispatch object is allocated for the dispatch, containing the topic the message was published to, the subscriber the message is being sent to, and the message itself. The topic is actually unused by any subscriber in Asterisk today. And the subscriber is associated with the taskprocessor the message is being dispatched to. First, this patch removes the unused topic parameter from Stasis subscription callbacks. Second, this patch introduces the concept of taskprocessor local data, data that may be set on a taskprocessor and provided along with the data pointer when a task is pushed using the ast_taskprocessor_push_local() call. This allows the task to have both data specific to that taskprocessor, in addition to data specific to that invocation. With those two changes, the dispatch object can be removed completely, and the message is simply refcounted and sent directly to the taskprocessor. Review: https://reviewboard.asterisk.org/r/2884/ ........ Merged revisions 399887,400138,400178,400180-400181 from http://svn.asterisk.org/svn/asterisk/branches/12 git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@400186 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-09-30 18:55:27 +00:00
static int find_route(
struct stasis_message_router *router,
Multiple revisions 399887,400138,400178,400180-400181 ........ r399887 | dlee | 2013-09-26 10:41:47 -0500 (Thu, 26 Sep 2013) | 1 line Minor performance bump by not allocate manager variable struct if we don't need it ........ r400138 | dlee | 2013-09-30 10:24:00 -0500 (Mon, 30 Sep 2013) | 23 lines Stasis performance improvements This patch addresses several performance problems that were found in the initial performance testing of Asterisk 12. The Stasis dispatch object was allocated as an AO2 object, even though it has a very confined lifecycle. This was replaced with a straight ast_malloc(). The Stasis message router was spending an inordinate amount of time searching hash tables. In this case, most of our routers had 6 or fewer routes in them to begin with. This was replaced with an array that's searched linearly for the route. We more heavily rely on AO2 objects in Asterisk 12, and the memset() in ao2_ref() actually became noticeable on the profile. This was #ifdef'ed to only run when AO2_DEBUG was enabled. After being misled by an erroneous comment in taskprocessor.c during profiling, the wrong comment was removed. Review: https://reviewboard.asterisk.org/r/2873/ ........ r400178 | dlee | 2013-09-30 13:26:27 -0500 (Mon, 30 Sep 2013) | 24 lines Taskprocessor optimization; switch Stasis to use taskprocessors This patch optimizes taskprocessor to use a semaphore for signaling, which the OS can do a better job at managing contention and waiting that we can with a mutex and condition. The taskprocessor execution was also slightly optimized to reduce the number of locks taken. The only observable difference in the taskprocessor implementation is that when the final reference to the taskprocessor goes away, it will execute all tasks to completion instead of discarding the unexecuted tasks. For systems where unnamed semaphores are not supported, a really simple semaphore implementation is provided. (Which gives identical performance as the original taskprocessor implementation). The way we ended up implementing Stasis caused the threadpool to be a burden instead of a boost to performance. This was switched to just use taskprocessors directly for subscriptions. Review: https://reviewboard.asterisk.org/r/2881/ ........ r400180 | dlee | 2013-09-30 13:39:34 -0500 (Mon, 30 Sep 2013) | 28 lines Optimize how Stasis forwards are dispatched This patch optimizes how forwards are dispatched in Stasis. Originally, forwards were dispatched as subscriptions that are invoked on the publishing thread. This did not account for the vast number of forwards we would end up having in the system, and the amount of work it would take to walk though the forward subscriptions. This patch modifies Stasis so that rather than walking the tree of forwards on every dispatch, when forwards and subscriptions are changed, the subscriber list for every topic in the tree is changed. This has a couple of benefits. First, this reduces the workload of dispatching messages. It also reduces contention when dispatching to different topics that happen to forward to the same aggregation topic (as happens with all of the channel, bridge and endpoint topics). Since forwards are no longer subscriptions, the bulk of this patch is simply changing stasis_subscription objects to stasis_forward objects (which, admittedly, I should have done in the first place.) Since this required me to yet again put in a growing array, I finally abstracted that out into a set of ast_vector macros in asterisk/vector.h. Review: https://reviewboard.asterisk.org/r/2883/ ........ r400181 | dlee | 2013-09-30 13:48:57 -0500 (Mon, 30 Sep 2013) | 28 lines Remove dispatch object allocation from Stasis publishing While looking for areas for performance improvement, I realized that an unused feature in Stasis was negatively impacting performance. When a message is sent to a subscriber, a dispatch object is allocated for the dispatch, containing the topic the message was published to, the subscriber the message is being sent to, and the message itself. The topic is actually unused by any subscriber in Asterisk today. And the subscriber is associated with the taskprocessor the message is being dispatched to. First, this patch removes the unused topic parameter from Stasis subscription callbacks. Second, this patch introduces the concept of taskprocessor local data, data that may be set on a taskprocessor and provided along with the data pointer when a task is pushed using the ast_taskprocessor_push_local() call. This allows the task to have both data specific to that taskprocessor, in addition to data specific to that invocation. With those two changes, the dispatch object can be removed completely, and the message is simply refcounted and sent directly to the taskprocessor. Review: https://reviewboard.asterisk.org/r/2884/ ........ Merged revisions 399887,400138,400178,400180-400181 from http://svn.asterisk.org/svn/asterisk/branches/12 git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@400186 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-09-30 18:55:27 +00:00
struct stasis_message *message,
struct stasis_message_route *route_out)
{
Multiple revisions 399887,400138,400178,400180-400181 ........ r399887 | dlee | 2013-09-26 10:41:47 -0500 (Thu, 26 Sep 2013) | 1 line Minor performance bump by not allocate manager variable struct if we don't need it ........ r400138 | dlee | 2013-09-30 10:24:00 -0500 (Mon, 30 Sep 2013) | 23 lines Stasis performance improvements This patch addresses several performance problems that were found in the initial performance testing of Asterisk 12. The Stasis dispatch object was allocated as an AO2 object, even though it has a very confined lifecycle. This was replaced with a straight ast_malloc(). The Stasis message router was spending an inordinate amount of time searching hash tables. In this case, most of our routers had 6 or fewer routes in them to begin with. This was replaced with an array that's searched linearly for the route. We more heavily rely on AO2 objects in Asterisk 12, and the memset() in ao2_ref() actually became noticeable on the profile. This was #ifdef'ed to only run when AO2_DEBUG was enabled. After being misled by an erroneous comment in taskprocessor.c during profiling, the wrong comment was removed. Review: https://reviewboard.asterisk.org/r/2873/ ........ r400178 | dlee | 2013-09-30 13:26:27 -0500 (Mon, 30 Sep 2013) | 24 lines Taskprocessor optimization; switch Stasis to use taskprocessors This patch optimizes taskprocessor to use a semaphore for signaling, which the OS can do a better job at managing contention and waiting that we can with a mutex and condition. The taskprocessor execution was also slightly optimized to reduce the number of locks taken. The only observable difference in the taskprocessor implementation is that when the final reference to the taskprocessor goes away, it will execute all tasks to completion instead of discarding the unexecuted tasks. For systems where unnamed semaphores are not supported, a really simple semaphore implementation is provided. (Which gives identical performance as the original taskprocessor implementation). The way we ended up implementing Stasis caused the threadpool to be a burden instead of a boost to performance. This was switched to just use taskprocessors directly for subscriptions. Review: https://reviewboard.asterisk.org/r/2881/ ........ r400180 | dlee | 2013-09-30 13:39:34 -0500 (Mon, 30 Sep 2013) | 28 lines Optimize how Stasis forwards are dispatched This patch optimizes how forwards are dispatched in Stasis. Originally, forwards were dispatched as subscriptions that are invoked on the publishing thread. This did not account for the vast number of forwards we would end up having in the system, and the amount of work it would take to walk though the forward subscriptions. This patch modifies Stasis so that rather than walking the tree of forwards on every dispatch, when forwards and subscriptions are changed, the subscriber list for every topic in the tree is changed. This has a couple of benefits. First, this reduces the workload of dispatching messages. It also reduces contention when dispatching to different topics that happen to forward to the same aggregation topic (as happens with all of the channel, bridge and endpoint topics). Since forwards are no longer subscriptions, the bulk of this patch is simply changing stasis_subscription objects to stasis_forward objects (which, admittedly, I should have done in the first place.) Since this required me to yet again put in a growing array, I finally abstracted that out into a set of ast_vector macros in asterisk/vector.h. Review: https://reviewboard.asterisk.org/r/2883/ ........ r400181 | dlee | 2013-09-30 13:48:57 -0500 (Mon, 30 Sep 2013) | 28 lines Remove dispatch object allocation from Stasis publishing While looking for areas for performance improvement, I realized that an unused feature in Stasis was negatively impacting performance. When a message is sent to a subscriber, a dispatch object is allocated for the dispatch, containing the topic the message was published to, the subscriber the message is being sent to, and the message itself. The topic is actually unused by any subscriber in Asterisk today. And the subscriber is associated with the taskprocessor the message is being dispatched to. First, this patch removes the unused topic parameter from Stasis subscription callbacks. Second, this patch introduces the concept of taskprocessor local data, data that may be set on a taskprocessor and provided along with the data pointer when a task is pushed using the ast_taskprocessor_push_local() call. This allows the task to have both data specific to that taskprocessor, in addition to data specific to that invocation. With those two changes, the dispatch object can be removed completely, and the message is simply refcounted and sent directly to the taskprocessor. Review: https://reviewboard.asterisk.org/r/2884/ ........ Merged revisions 399887,400138,400178,400180-400181 from http://svn.asterisk.org/svn/asterisk/branches/12 git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@400186 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-09-30 18:55:27 +00:00
struct stasis_message_route *route = NULL;
struct stasis_message_type *type = stasis_message_type(message);
SCOPED_AO2LOCK(lock, router);
Multiple revisions 399887,400138,400178,400180-400181 ........ r399887 | dlee | 2013-09-26 10:41:47 -0500 (Thu, 26 Sep 2013) | 1 line Minor performance bump by not allocate manager variable struct if we don't need it ........ r400138 | dlee | 2013-09-30 10:24:00 -0500 (Mon, 30 Sep 2013) | 23 lines Stasis performance improvements This patch addresses several performance problems that were found in the initial performance testing of Asterisk 12. The Stasis dispatch object was allocated as an AO2 object, even though it has a very confined lifecycle. This was replaced with a straight ast_malloc(). The Stasis message router was spending an inordinate amount of time searching hash tables. In this case, most of our routers had 6 or fewer routes in them to begin with. This was replaced with an array that's searched linearly for the route. We more heavily rely on AO2 objects in Asterisk 12, and the memset() in ao2_ref() actually became noticeable on the profile. This was #ifdef'ed to only run when AO2_DEBUG was enabled. After being misled by an erroneous comment in taskprocessor.c during profiling, the wrong comment was removed. Review: https://reviewboard.asterisk.org/r/2873/ ........ r400178 | dlee | 2013-09-30 13:26:27 -0500 (Mon, 30 Sep 2013) | 24 lines Taskprocessor optimization; switch Stasis to use taskprocessors This patch optimizes taskprocessor to use a semaphore for signaling, which the OS can do a better job at managing contention and waiting that we can with a mutex and condition. The taskprocessor execution was also slightly optimized to reduce the number of locks taken. The only observable difference in the taskprocessor implementation is that when the final reference to the taskprocessor goes away, it will execute all tasks to completion instead of discarding the unexecuted tasks. For systems where unnamed semaphores are not supported, a really simple semaphore implementation is provided. (Which gives identical performance as the original taskprocessor implementation). The way we ended up implementing Stasis caused the threadpool to be a burden instead of a boost to performance. This was switched to just use taskprocessors directly for subscriptions. Review: https://reviewboard.asterisk.org/r/2881/ ........ r400180 | dlee | 2013-09-30 13:39:34 -0500 (Mon, 30 Sep 2013) | 28 lines Optimize how Stasis forwards are dispatched This patch optimizes how forwards are dispatched in Stasis. Originally, forwards were dispatched as subscriptions that are invoked on the publishing thread. This did not account for the vast number of forwards we would end up having in the system, and the amount of work it would take to walk though the forward subscriptions. This patch modifies Stasis so that rather than walking the tree of forwards on every dispatch, when forwards and subscriptions are changed, the subscriber list for every topic in the tree is changed. This has a couple of benefits. First, this reduces the workload of dispatching messages. It also reduces contention when dispatching to different topics that happen to forward to the same aggregation topic (as happens with all of the channel, bridge and endpoint topics). Since forwards are no longer subscriptions, the bulk of this patch is simply changing stasis_subscription objects to stasis_forward objects (which, admittedly, I should have done in the first place.) Since this required me to yet again put in a growing array, I finally abstracted that out into a set of ast_vector macros in asterisk/vector.h. Review: https://reviewboard.asterisk.org/r/2883/ ........ r400181 | dlee | 2013-09-30 13:48:57 -0500 (Mon, 30 Sep 2013) | 28 lines Remove dispatch object allocation from Stasis publishing While looking for areas for performance improvement, I realized that an unused feature in Stasis was negatively impacting performance. When a message is sent to a subscriber, a dispatch object is allocated for the dispatch, containing the topic the message was published to, the subscriber the message is being sent to, and the message itself. The topic is actually unused by any subscriber in Asterisk today. And the subscriber is associated with the taskprocessor the message is being dispatched to. First, this patch removes the unused topic parameter from Stasis subscription callbacks. Second, this patch introduces the concept of taskprocessor local data, data that may be set on a taskprocessor and provided along with the data pointer when a task is pushed using the ast_taskprocessor_push_local() call. This allows the task to have both data specific to that taskprocessor, in addition to data specific to that invocation. With those two changes, the dispatch object can be removed completely, and the message is simply refcounted and sent directly to the taskprocessor. Review: https://reviewboard.asterisk.org/r/2884/ ........ Merged revisions 399887,400138,400178,400180-400181 from http://svn.asterisk.org/svn/asterisk/branches/12 git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@400186 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-09-30 18:55:27 +00:00
ast_assert(route_out != NULL);
if (type == stasis_cache_update_type()) {
/* Find a cache route */
struct stasis_cache_update *update =
stasis_message_data(message);
route = route_table_find(&router->cache_routes, update->type);
}
if (route == NULL) {
/* Find a regular route */
route = route_table_find(&router->routes, type);
}
Multiple revisions 399887,400138,400178,400180-400181 ........ r399887 | dlee | 2013-09-26 10:41:47 -0500 (Thu, 26 Sep 2013) | 1 line Minor performance bump by not allocate manager variable struct if we don't need it ........ r400138 | dlee | 2013-09-30 10:24:00 -0500 (Mon, 30 Sep 2013) | 23 lines Stasis performance improvements This patch addresses several performance problems that were found in the initial performance testing of Asterisk 12. The Stasis dispatch object was allocated as an AO2 object, even though it has a very confined lifecycle. This was replaced with a straight ast_malloc(). The Stasis message router was spending an inordinate amount of time searching hash tables. In this case, most of our routers had 6 or fewer routes in them to begin with. This was replaced with an array that's searched linearly for the route. We more heavily rely on AO2 objects in Asterisk 12, and the memset() in ao2_ref() actually became noticeable on the profile. This was #ifdef'ed to only run when AO2_DEBUG was enabled. After being misled by an erroneous comment in taskprocessor.c during profiling, the wrong comment was removed. Review: https://reviewboard.asterisk.org/r/2873/ ........ r400178 | dlee | 2013-09-30 13:26:27 -0500 (Mon, 30 Sep 2013) | 24 lines Taskprocessor optimization; switch Stasis to use taskprocessors This patch optimizes taskprocessor to use a semaphore for signaling, which the OS can do a better job at managing contention and waiting that we can with a mutex and condition. The taskprocessor execution was also slightly optimized to reduce the number of locks taken. The only observable difference in the taskprocessor implementation is that when the final reference to the taskprocessor goes away, it will execute all tasks to completion instead of discarding the unexecuted tasks. For systems where unnamed semaphores are not supported, a really simple semaphore implementation is provided. (Which gives identical performance as the original taskprocessor implementation). The way we ended up implementing Stasis caused the threadpool to be a burden instead of a boost to performance. This was switched to just use taskprocessors directly for subscriptions. Review: https://reviewboard.asterisk.org/r/2881/ ........ r400180 | dlee | 2013-09-30 13:39:34 -0500 (Mon, 30 Sep 2013) | 28 lines Optimize how Stasis forwards are dispatched This patch optimizes how forwards are dispatched in Stasis. Originally, forwards were dispatched as subscriptions that are invoked on the publishing thread. This did not account for the vast number of forwards we would end up having in the system, and the amount of work it would take to walk though the forward subscriptions. This patch modifies Stasis so that rather than walking the tree of forwards on every dispatch, when forwards and subscriptions are changed, the subscriber list for every topic in the tree is changed. This has a couple of benefits. First, this reduces the workload of dispatching messages. It also reduces contention when dispatching to different topics that happen to forward to the same aggregation topic (as happens with all of the channel, bridge and endpoint topics). Since forwards are no longer subscriptions, the bulk of this patch is simply changing stasis_subscription objects to stasis_forward objects (which, admittedly, I should have done in the first place.) Since this required me to yet again put in a growing array, I finally abstracted that out into a set of ast_vector macros in asterisk/vector.h. Review: https://reviewboard.asterisk.org/r/2883/ ........ r400181 | dlee | 2013-09-30 13:48:57 -0500 (Mon, 30 Sep 2013) | 28 lines Remove dispatch object allocation from Stasis publishing While looking for areas for performance improvement, I realized that an unused feature in Stasis was negatively impacting performance. When a message is sent to a subscriber, a dispatch object is allocated for the dispatch, containing the topic the message was published to, the subscriber the message is being sent to, and the message itself. The topic is actually unused by any subscriber in Asterisk today. And the subscriber is associated with the taskprocessor the message is being dispatched to. First, this patch removes the unused topic parameter from Stasis subscription callbacks. Second, this patch introduces the concept of taskprocessor local data, data that may be set on a taskprocessor and provided along with the data pointer when a task is pushed using the ast_taskprocessor_push_local() call. This allows the task to have both data specific to that taskprocessor, in addition to data specific to that invocation. With those two changes, the dispatch object can be removed completely, and the message is simply refcounted and sent directly to the taskprocessor. Review: https://reviewboard.asterisk.org/r/2884/ ........ Merged revisions 399887,400138,400178,400180-400181 from http://svn.asterisk.org/svn/asterisk/branches/12 git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@400186 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-09-30 18:55:27 +00:00
if (route == NULL && router->default_route.callback) {
/* Maybe the default route, then? */
Multiple revisions 399887,400138,400178,400180-400181 ........ r399887 | dlee | 2013-09-26 10:41:47 -0500 (Thu, 26 Sep 2013) | 1 line Minor performance bump by not allocate manager variable struct if we don't need it ........ r400138 | dlee | 2013-09-30 10:24:00 -0500 (Mon, 30 Sep 2013) | 23 lines Stasis performance improvements This patch addresses several performance problems that were found in the initial performance testing of Asterisk 12. The Stasis dispatch object was allocated as an AO2 object, even though it has a very confined lifecycle. This was replaced with a straight ast_malloc(). The Stasis message router was spending an inordinate amount of time searching hash tables. In this case, most of our routers had 6 or fewer routes in them to begin with. This was replaced with an array that's searched linearly for the route. We more heavily rely on AO2 objects in Asterisk 12, and the memset() in ao2_ref() actually became noticeable on the profile. This was #ifdef'ed to only run when AO2_DEBUG was enabled. After being misled by an erroneous comment in taskprocessor.c during profiling, the wrong comment was removed. Review: https://reviewboard.asterisk.org/r/2873/ ........ r400178 | dlee | 2013-09-30 13:26:27 -0500 (Mon, 30 Sep 2013) | 24 lines Taskprocessor optimization; switch Stasis to use taskprocessors This patch optimizes taskprocessor to use a semaphore for signaling, which the OS can do a better job at managing contention and waiting that we can with a mutex and condition. The taskprocessor execution was also slightly optimized to reduce the number of locks taken. The only observable difference in the taskprocessor implementation is that when the final reference to the taskprocessor goes away, it will execute all tasks to completion instead of discarding the unexecuted tasks. For systems where unnamed semaphores are not supported, a really simple semaphore implementation is provided. (Which gives identical performance as the original taskprocessor implementation). The way we ended up implementing Stasis caused the threadpool to be a burden instead of a boost to performance. This was switched to just use taskprocessors directly for subscriptions. Review: https://reviewboard.asterisk.org/r/2881/ ........ r400180 | dlee | 2013-09-30 13:39:34 -0500 (Mon, 30 Sep 2013) | 28 lines Optimize how Stasis forwards are dispatched This patch optimizes how forwards are dispatched in Stasis. Originally, forwards were dispatched as subscriptions that are invoked on the publishing thread. This did not account for the vast number of forwards we would end up having in the system, and the amount of work it would take to walk though the forward subscriptions. This patch modifies Stasis so that rather than walking the tree of forwards on every dispatch, when forwards and subscriptions are changed, the subscriber list for every topic in the tree is changed. This has a couple of benefits. First, this reduces the workload of dispatching messages. It also reduces contention when dispatching to different topics that happen to forward to the same aggregation topic (as happens with all of the channel, bridge and endpoint topics). Since forwards are no longer subscriptions, the bulk of this patch is simply changing stasis_subscription objects to stasis_forward objects (which, admittedly, I should have done in the first place.) Since this required me to yet again put in a growing array, I finally abstracted that out into a set of ast_vector macros in asterisk/vector.h. Review: https://reviewboard.asterisk.org/r/2883/ ........ r400181 | dlee | 2013-09-30 13:48:57 -0500 (Mon, 30 Sep 2013) | 28 lines Remove dispatch object allocation from Stasis publishing While looking for areas for performance improvement, I realized that an unused feature in Stasis was negatively impacting performance. When a message is sent to a subscriber, a dispatch object is allocated for the dispatch, containing the topic the message was published to, the subscriber the message is being sent to, and the message itself. The topic is actually unused by any subscriber in Asterisk today. And the subscriber is associated with the taskprocessor the message is being dispatched to. First, this patch removes the unused topic parameter from Stasis subscription callbacks. Second, this patch introduces the concept of taskprocessor local data, data that may be set on a taskprocessor and provided along with the data pointer when a task is pushed using the ast_taskprocessor_push_local() call. This allows the task to have both data specific to that taskprocessor, in addition to data specific to that invocation. With those two changes, the dispatch object can be removed completely, and the message is simply refcounted and sent directly to the taskprocessor. Review: https://reviewboard.asterisk.org/r/2884/ ........ Merged revisions 399887,400138,400178,400180-400181 from http://svn.asterisk.org/svn/asterisk/branches/12 git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@400186 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-09-30 18:55:27 +00:00
route = &router->default_route;
}
Multiple revisions 399887,400138,400178,400180-400181 ........ r399887 | dlee | 2013-09-26 10:41:47 -0500 (Thu, 26 Sep 2013) | 1 line Minor performance bump by not allocate manager variable struct if we don't need it ........ r400138 | dlee | 2013-09-30 10:24:00 -0500 (Mon, 30 Sep 2013) | 23 lines Stasis performance improvements This patch addresses several performance problems that were found in the initial performance testing of Asterisk 12. The Stasis dispatch object was allocated as an AO2 object, even though it has a very confined lifecycle. This was replaced with a straight ast_malloc(). The Stasis message router was spending an inordinate amount of time searching hash tables. In this case, most of our routers had 6 or fewer routes in them to begin with. This was replaced with an array that's searched linearly for the route. We more heavily rely on AO2 objects in Asterisk 12, and the memset() in ao2_ref() actually became noticeable on the profile. This was #ifdef'ed to only run when AO2_DEBUG was enabled. After being misled by an erroneous comment in taskprocessor.c during profiling, the wrong comment was removed. Review: https://reviewboard.asterisk.org/r/2873/ ........ r400178 | dlee | 2013-09-30 13:26:27 -0500 (Mon, 30 Sep 2013) | 24 lines Taskprocessor optimization; switch Stasis to use taskprocessors This patch optimizes taskprocessor to use a semaphore for signaling, which the OS can do a better job at managing contention and waiting that we can with a mutex and condition. The taskprocessor execution was also slightly optimized to reduce the number of locks taken. The only observable difference in the taskprocessor implementation is that when the final reference to the taskprocessor goes away, it will execute all tasks to completion instead of discarding the unexecuted tasks. For systems where unnamed semaphores are not supported, a really simple semaphore implementation is provided. (Which gives identical performance as the original taskprocessor implementation). The way we ended up implementing Stasis caused the threadpool to be a burden instead of a boost to performance. This was switched to just use taskprocessors directly for subscriptions. Review: https://reviewboard.asterisk.org/r/2881/ ........ r400180 | dlee | 2013-09-30 13:39:34 -0500 (Mon, 30 Sep 2013) | 28 lines Optimize how Stasis forwards are dispatched This patch optimizes how forwards are dispatched in Stasis. Originally, forwards were dispatched as subscriptions that are invoked on the publishing thread. This did not account for the vast number of forwards we would end up having in the system, and the amount of work it would take to walk though the forward subscriptions. This patch modifies Stasis so that rather than walking the tree of forwards on every dispatch, when forwards and subscriptions are changed, the subscriber list for every topic in the tree is changed. This has a couple of benefits. First, this reduces the workload of dispatching messages. It also reduces contention when dispatching to different topics that happen to forward to the same aggregation topic (as happens with all of the channel, bridge and endpoint topics). Since forwards are no longer subscriptions, the bulk of this patch is simply changing stasis_subscription objects to stasis_forward objects (which, admittedly, I should have done in the first place.) Since this required me to yet again put in a growing array, I finally abstracted that out into a set of ast_vector macros in asterisk/vector.h. Review: https://reviewboard.asterisk.org/r/2883/ ........ r400181 | dlee | 2013-09-30 13:48:57 -0500 (Mon, 30 Sep 2013) | 28 lines Remove dispatch object allocation from Stasis publishing While looking for areas for performance improvement, I realized that an unused feature in Stasis was negatively impacting performance. When a message is sent to a subscriber, a dispatch object is allocated for the dispatch, containing the topic the message was published to, the subscriber the message is being sent to, and the message itself. The topic is actually unused by any subscriber in Asterisk today. And the subscriber is associated with the taskprocessor the message is being dispatched to. First, this patch removes the unused topic parameter from Stasis subscription callbacks. Second, this patch introduces the concept of taskprocessor local data, data that may be set on a taskprocessor and provided along with the data pointer when a task is pushed using the ast_taskprocessor_push_local() call. This allows the task to have both data specific to that taskprocessor, in addition to data specific to that invocation. With those two changes, the dispatch object can be removed completely, and the message is simply refcounted and sent directly to the taskprocessor. Review: https://reviewboard.asterisk.org/r/2884/ ........ Merged revisions 399887,400138,400178,400180-400181 from http://svn.asterisk.org/svn/asterisk/branches/12 git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@400186 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-09-30 18:55:27 +00:00
if (!route) {
return -1;
}
Multiple revisions 399887,400138,400178,400180-400181 ........ r399887 | dlee | 2013-09-26 10:41:47 -0500 (Thu, 26 Sep 2013) | 1 line Minor performance bump by not allocate manager variable struct if we don't need it ........ r400138 | dlee | 2013-09-30 10:24:00 -0500 (Mon, 30 Sep 2013) | 23 lines Stasis performance improvements This patch addresses several performance problems that were found in the initial performance testing of Asterisk 12. The Stasis dispatch object was allocated as an AO2 object, even though it has a very confined lifecycle. This was replaced with a straight ast_malloc(). The Stasis message router was spending an inordinate amount of time searching hash tables. In this case, most of our routers had 6 or fewer routes in them to begin with. This was replaced with an array that's searched linearly for the route. We more heavily rely on AO2 objects in Asterisk 12, and the memset() in ao2_ref() actually became noticeable on the profile. This was #ifdef'ed to only run when AO2_DEBUG was enabled. After being misled by an erroneous comment in taskprocessor.c during profiling, the wrong comment was removed. Review: https://reviewboard.asterisk.org/r/2873/ ........ r400178 | dlee | 2013-09-30 13:26:27 -0500 (Mon, 30 Sep 2013) | 24 lines Taskprocessor optimization; switch Stasis to use taskprocessors This patch optimizes taskprocessor to use a semaphore for signaling, which the OS can do a better job at managing contention and waiting that we can with a mutex and condition. The taskprocessor execution was also slightly optimized to reduce the number of locks taken. The only observable difference in the taskprocessor implementation is that when the final reference to the taskprocessor goes away, it will execute all tasks to completion instead of discarding the unexecuted tasks. For systems where unnamed semaphores are not supported, a really simple semaphore implementation is provided. (Which gives identical performance as the original taskprocessor implementation). The way we ended up implementing Stasis caused the threadpool to be a burden instead of a boost to performance. This was switched to just use taskprocessors directly for subscriptions. Review: https://reviewboard.asterisk.org/r/2881/ ........ r400180 | dlee | 2013-09-30 13:39:34 -0500 (Mon, 30 Sep 2013) | 28 lines Optimize how Stasis forwards are dispatched This patch optimizes how forwards are dispatched in Stasis. Originally, forwards were dispatched as subscriptions that are invoked on the publishing thread. This did not account for the vast number of forwards we would end up having in the system, and the amount of work it would take to walk though the forward subscriptions. This patch modifies Stasis so that rather than walking the tree of forwards on every dispatch, when forwards and subscriptions are changed, the subscriber list for every topic in the tree is changed. This has a couple of benefits. First, this reduces the workload of dispatching messages. It also reduces contention when dispatching to different topics that happen to forward to the same aggregation topic (as happens with all of the channel, bridge and endpoint topics). Since forwards are no longer subscriptions, the bulk of this patch is simply changing stasis_subscription objects to stasis_forward objects (which, admittedly, I should have done in the first place.) Since this required me to yet again put in a growing array, I finally abstracted that out into a set of ast_vector macros in asterisk/vector.h. Review: https://reviewboard.asterisk.org/r/2883/ ........ r400181 | dlee | 2013-09-30 13:48:57 -0500 (Mon, 30 Sep 2013) | 28 lines Remove dispatch object allocation from Stasis publishing While looking for areas for performance improvement, I realized that an unused feature in Stasis was negatively impacting performance. When a message is sent to a subscriber, a dispatch object is allocated for the dispatch, containing the topic the message was published to, the subscriber the message is being sent to, and the message itself. The topic is actually unused by any subscriber in Asterisk today. And the subscriber is associated with the taskprocessor the message is being dispatched to. First, this patch removes the unused topic parameter from Stasis subscription callbacks. Second, this patch introduces the concept of taskprocessor local data, data that may be set on a taskprocessor and provided along with the data pointer when a task is pushed using the ast_taskprocessor_push_local() call. This allows the task to have both data specific to that taskprocessor, in addition to data specific to that invocation. With those two changes, the dispatch object can be removed completely, and the message is simply refcounted and sent directly to the taskprocessor. Review: https://reviewboard.asterisk.org/r/2884/ ........ Merged revisions 399887,400138,400178,400180-400181 from http://svn.asterisk.org/svn/asterisk/branches/12 git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@400186 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-09-30 18:55:27 +00:00
*route_out = *route;
return 0;
}
static void router_dispatch(void *data,
struct stasis_subscription *sub,
struct stasis_message *message)
{
struct stasis_message_router *router = data;
Multiple revisions 399887,400138,400178,400180-400181 ........ r399887 | dlee | 2013-09-26 10:41:47 -0500 (Thu, 26 Sep 2013) | 1 line Minor performance bump by not allocate manager variable struct if we don't need it ........ r400138 | dlee | 2013-09-30 10:24:00 -0500 (Mon, 30 Sep 2013) | 23 lines Stasis performance improvements This patch addresses several performance problems that were found in the initial performance testing of Asterisk 12. The Stasis dispatch object was allocated as an AO2 object, even though it has a very confined lifecycle. This was replaced with a straight ast_malloc(). The Stasis message router was spending an inordinate amount of time searching hash tables. In this case, most of our routers had 6 or fewer routes in them to begin with. This was replaced with an array that's searched linearly for the route. We more heavily rely on AO2 objects in Asterisk 12, and the memset() in ao2_ref() actually became noticeable on the profile. This was #ifdef'ed to only run when AO2_DEBUG was enabled. After being misled by an erroneous comment in taskprocessor.c during profiling, the wrong comment was removed. Review: https://reviewboard.asterisk.org/r/2873/ ........ r400178 | dlee | 2013-09-30 13:26:27 -0500 (Mon, 30 Sep 2013) | 24 lines Taskprocessor optimization; switch Stasis to use taskprocessors This patch optimizes taskprocessor to use a semaphore for signaling, which the OS can do a better job at managing contention and waiting that we can with a mutex and condition. The taskprocessor execution was also slightly optimized to reduce the number of locks taken. The only observable difference in the taskprocessor implementation is that when the final reference to the taskprocessor goes away, it will execute all tasks to completion instead of discarding the unexecuted tasks. For systems where unnamed semaphores are not supported, a really simple semaphore implementation is provided. (Which gives identical performance as the original taskprocessor implementation). The way we ended up implementing Stasis caused the threadpool to be a burden instead of a boost to performance. This was switched to just use taskprocessors directly for subscriptions. Review: https://reviewboard.asterisk.org/r/2881/ ........ r400180 | dlee | 2013-09-30 13:39:34 -0500 (Mon, 30 Sep 2013) | 28 lines Optimize how Stasis forwards are dispatched This patch optimizes how forwards are dispatched in Stasis. Originally, forwards were dispatched as subscriptions that are invoked on the publishing thread. This did not account for the vast number of forwards we would end up having in the system, and the amount of work it would take to walk though the forward subscriptions. This patch modifies Stasis so that rather than walking the tree of forwards on every dispatch, when forwards and subscriptions are changed, the subscriber list for every topic in the tree is changed. This has a couple of benefits. First, this reduces the workload of dispatching messages. It also reduces contention when dispatching to different topics that happen to forward to the same aggregation topic (as happens with all of the channel, bridge and endpoint topics). Since forwards are no longer subscriptions, the bulk of this patch is simply changing stasis_subscription objects to stasis_forward objects (which, admittedly, I should have done in the first place.) Since this required me to yet again put in a growing array, I finally abstracted that out into a set of ast_vector macros in asterisk/vector.h. Review: https://reviewboard.asterisk.org/r/2883/ ........ r400181 | dlee | 2013-09-30 13:48:57 -0500 (Mon, 30 Sep 2013) | 28 lines Remove dispatch object allocation from Stasis publishing While looking for areas for performance improvement, I realized that an unused feature in Stasis was negatively impacting performance. When a message is sent to a subscriber, a dispatch object is allocated for the dispatch, containing the topic the message was published to, the subscriber the message is being sent to, and the message itself. The topic is actually unused by any subscriber in Asterisk today. And the subscriber is associated with the taskprocessor the message is being dispatched to. First, this patch removes the unused topic parameter from Stasis subscription callbacks. Second, this patch introduces the concept of taskprocessor local data, data that may be set on a taskprocessor and provided along with the data pointer when a task is pushed using the ast_taskprocessor_push_local() call. This allows the task to have both data specific to that taskprocessor, in addition to data specific to that invocation. With those two changes, the dispatch object can be removed completely, and the message is simply refcounted and sent directly to the taskprocessor. Review: https://reviewboard.asterisk.org/r/2884/ ........ Merged revisions 399887,400138,400178,400180-400181 from http://svn.asterisk.org/svn/asterisk/branches/12 git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@400186 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-09-30 18:55:27 +00:00
struct stasis_message_route route;
Multiple revisions 399887,400138,400178,400180-400181 ........ r399887 | dlee | 2013-09-26 10:41:47 -0500 (Thu, 26 Sep 2013) | 1 line Minor performance bump by not allocate manager variable struct if we don't need it ........ r400138 | dlee | 2013-09-30 10:24:00 -0500 (Mon, 30 Sep 2013) | 23 lines Stasis performance improvements This patch addresses several performance problems that were found in the initial performance testing of Asterisk 12. The Stasis dispatch object was allocated as an AO2 object, even though it has a very confined lifecycle. This was replaced with a straight ast_malloc(). The Stasis message router was spending an inordinate amount of time searching hash tables. In this case, most of our routers had 6 or fewer routes in them to begin with. This was replaced with an array that's searched linearly for the route. We more heavily rely on AO2 objects in Asterisk 12, and the memset() in ao2_ref() actually became noticeable on the profile. This was #ifdef'ed to only run when AO2_DEBUG was enabled. After being misled by an erroneous comment in taskprocessor.c during profiling, the wrong comment was removed. Review: https://reviewboard.asterisk.org/r/2873/ ........ r400178 | dlee | 2013-09-30 13:26:27 -0500 (Mon, 30 Sep 2013) | 24 lines Taskprocessor optimization; switch Stasis to use taskprocessors This patch optimizes taskprocessor to use a semaphore for signaling, which the OS can do a better job at managing contention and waiting that we can with a mutex and condition. The taskprocessor execution was also slightly optimized to reduce the number of locks taken. The only observable difference in the taskprocessor implementation is that when the final reference to the taskprocessor goes away, it will execute all tasks to completion instead of discarding the unexecuted tasks. For systems where unnamed semaphores are not supported, a really simple semaphore implementation is provided. (Which gives identical performance as the original taskprocessor implementation). The way we ended up implementing Stasis caused the threadpool to be a burden instead of a boost to performance. This was switched to just use taskprocessors directly for subscriptions. Review: https://reviewboard.asterisk.org/r/2881/ ........ r400180 | dlee | 2013-09-30 13:39:34 -0500 (Mon, 30 Sep 2013) | 28 lines Optimize how Stasis forwards are dispatched This patch optimizes how forwards are dispatched in Stasis. Originally, forwards were dispatched as subscriptions that are invoked on the publishing thread. This did not account for the vast number of forwards we would end up having in the system, and the amount of work it would take to walk though the forward subscriptions. This patch modifies Stasis so that rather than walking the tree of forwards on every dispatch, when forwards and subscriptions are changed, the subscriber list for every topic in the tree is changed. This has a couple of benefits. First, this reduces the workload of dispatching messages. It also reduces contention when dispatching to different topics that happen to forward to the same aggregation topic (as happens with all of the channel, bridge and endpoint topics). Since forwards are no longer subscriptions, the bulk of this patch is simply changing stasis_subscription objects to stasis_forward objects (which, admittedly, I should have done in the first place.) Since this required me to yet again put in a growing array, I finally abstracted that out into a set of ast_vector macros in asterisk/vector.h. Review: https://reviewboard.asterisk.org/r/2883/ ........ r400181 | dlee | 2013-09-30 13:48:57 -0500 (Mon, 30 Sep 2013) | 28 lines Remove dispatch object allocation from Stasis publishing While looking for areas for performance improvement, I realized that an unused feature in Stasis was negatively impacting performance. When a message is sent to a subscriber, a dispatch object is allocated for the dispatch, containing the topic the message was published to, the subscriber the message is being sent to, and the message itself. The topic is actually unused by any subscriber in Asterisk today. And the subscriber is associated with the taskprocessor the message is being dispatched to. First, this patch removes the unused topic parameter from Stasis subscription callbacks. Second, this patch introduces the concept of taskprocessor local data, data that may be set on a taskprocessor and provided along with the data pointer when a task is pushed using the ast_taskprocessor_push_local() call. This allows the task to have both data specific to that taskprocessor, in addition to data specific to that invocation. With those two changes, the dispatch object can be removed completely, and the message is simply refcounted and sent directly to the taskprocessor. Review: https://reviewboard.asterisk.org/r/2884/ ........ Merged revisions 399887,400138,400178,400180-400181 from http://svn.asterisk.org/svn/asterisk/branches/12 git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@400186 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-09-30 18:55:27 +00:00
if (find_route(router, message, &route) == 0) {
route.callback(route.data, sub, message);
}
if (stasis_subscription_final_message(sub, message)) {
ao2_cleanup(router);
}
}
struct stasis_message_router *stasis_message_router_create(
struct stasis_topic *topic)
{
int res;
RAII_VAR(struct stasis_message_router *, router, NULL, ao2_cleanup);
router = ao2_t_alloc(sizeof(*router), router_dtor, stasis_topic_name(topic));
if (!router) {
return NULL;
}
res = 0;
res |= AST_VECTOR_INIT(&router->routes, 0);
res |= AST_VECTOR_INIT(&router->cache_routes, 0);
if (res) {
return NULL;
}
router->subscription = stasis_subscribe(topic, router_dispatch, router);
if (!router->subscription) {
return NULL;
}
ao2_ref(router, +1);
return router;
}
void stasis_message_router_unsubscribe(struct stasis_message_router *router)
{
if (!router) {
return;
}
stasis_unsubscribe(router->subscription);
}
void stasis_message_router_unsubscribe_and_join(
struct stasis_message_router *router)
{
if (!router) {
return;
}
stasis_unsubscribe_and_join(router->subscription);
}
int stasis_message_router_is_done(struct stasis_message_router *router)
{
if (!router) {
/* Null router is about as done as you can get */
return 1;
}
return stasis_subscription_is_done(router->subscription);
}
void stasis_message_router_publish_sync(struct stasis_message_router *router,
struct stasis_message *message)
{
ast_assert(router != NULL);
ao2_bump(router);
stasis_publish_sync(router->subscription, message);
ao2_cleanup(router);
}
int stasis_message_router_add(struct stasis_message_router *router,
struct stasis_message_type *message_type,
stasis_subscription_cb callback, void *data)
{
int res;
ast_assert(router != NULL);
if (!message_type) {
/* Cannot route to NULL type. */
return -1;
}
ao2_lock(router);
res = route_table_add(&router->routes, message_type, callback, data);
ao2_unlock(router);
return res;
}
int stasis_message_router_add_cache_update(struct stasis_message_router *router,
struct stasis_message_type *message_type,
stasis_subscription_cb callback, void *data)
{
int res;
ast_assert(router != NULL);
if (!message_type) {
/* Cannot cache a route to NULL type. */
return -1;
}
ao2_lock(router);
res = route_table_add(&router->cache_routes, message_type, callback, data);
ao2_unlock(router);
return res;
}
void stasis_message_router_remove(struct stasis_message_router *router,
struct stasis_message_type *message_type)
{
ast_assert(router != NULL);
if (!message_type) {
/* Cannot remove a NULL type. */
return;
}
ao2_lock(router);
route_table_remove(&router->routes, message_type);
ao2_unlock(router);
}
void stasis_message_router_remove_cache_update(
struct stasis_message_router *router,
struct stasis_message_type *message_type)
{
ast_assert(router != NULL);
if (!message_type) {
/* Cannot remove a NULL type. */
return;
}
ao2_lock(router);
route_table_remove(&router->cache_routes, message_type);
ao2_unlock(router);
}
int stasis_message_router_set_default(struct stasis_message_router *router,
stasis_subscription_cb callback,
void *data)
{
ast_assert(router != NULL);
ast_assert(callback != NULL);
ao2_lock(router);
Multiple revisions 399887,400138,400178,400180-400181 ........ r399887 | dlee | 2013-09-26 10:41:47 -0500 (Thu, 26 Sep 2013) | 1 line Minor performance bump by not allocate manager variable struct if we don't need it ........ r400138 | dlee | 2013-09-30 10:24:00 -0500 (Mon, 30 Sep 2013) | 23 lines Stasis performance improvements This patch addresses several performance problems that were found in the initial performance testing of Asterisk 12. The Stasis dispatch object was allocated as an AO2 object, even though it has a very confined lifecycle. This was replaced with a straight ast_malloc(). The Stasis message router was spending an inordinate amount of time searching hash tables. In this case, most of our routers had 6 or fewer routes in them to begin with. This was replaced with an array that's searched linearly for the route. We more heavily rely on AO2 objects in Asterisk 12, and the memset() in ao2_ref() actually became noticeable on the profile. This was #ifdef'ed to only run when AO2_DEBUG was enabled. After being misled by an erroneous comment in taskprocessor.c during profiling, the wrong comment was removed. Review: https://reviewboard.asterisk.org/r/2873/ ........ r400178 | dlee | 2013-09-30 13:26:27 -0500 (Mon, 30 Sep 2013) | 24 lines Taskprocessor optimization; switch Stasis to use taskprocessors This patch optimizes taskprocessor to use a semaphore for signaling, which the OS can do a better job at managing contention and waiting that we can with a mutex and condition. The taskprocessor execution was also slightly optimized to reduce the number of locks taken. The only observable difference in the taskprocessor implementation is that when the final reference to the taskprocessor goes away, it will execute all tasks to completion instead of discarding the unexecuted tasks. For systems where unnamed semaphores are not supported, a really simple semaphore implementation is provided. (Which gives identical performance as the original taskprocessor implementation). The way we ended up implementing Stasis caused the threadpool to be a burden instead of a boost to performance. This was switched to just use taskprocessors directly for subscriptions. Review: https://reviewboard.asterisk.org/r/2881/ ........ r400180 | dlee | 2013-09-30 13:39:34 -0500 (Mon, 30 Sep 2013) | 28 lines Optimize how Stasis forwards are dispatched This patch optimizes how forwards are dispatched in Stasis. Originally, forwards were dispatched as subscriptions that are invoked on the publishing thread. This did not account for the vast number of forwards we would end up having in the system, and the amount of work it would take to walk though the forward subscriptions. This patch modifies Stasis so that rather than walking the tree of forwards on every dispatch, when forwards and subscriptions are changed, the subscriber list for every topic in the tree is changed. This has a couple of benefits. First, this reduces the workload of dispatching messages. It also reduces contention when dispatching to different topics that happen to forward to the same aggregation topic (as happens with all of the channel, bridge and endpoint topics). Since forwards are no longer subscriptions, the bulk of this patch is simply changing stasis_subscription objects to stasis_forward objects (which, admittedly, I should have done in the first place.) Since this required me to yet again put in a growing array, I finally abstracted that out into a set of ast_vector macros in asterisk/vector.h. Review: https://reviewboard.asterisk.org/r/2883/ ........ r400181 | dlee | 2013-09-30 13:48:57 -0500 (Mon, 30 Sep 2013) | 28 lines Remove dispatch object allocation from Stasis publishing While looking for areas for performance improvement, I realized that an unused feature in Stasis was negatively impacting performance. When a message is sent to a subscriber, a dispatch object is allocated for the dispatch, containing the topic the message was published to, the subscriber the message is being sent to, and the message itself. The topic is actually unused by any subscriber in Asterisk today. And the subscriber is associated with the taskprocessor the message is being dispatched to. First, this patch removes the unused topic parameter from Stasis subscription callbacks. Second, this patch introduces the concept of taskprocessor local data, data that may be set on a taskprocessor and provided along with the data pointer when a task is pushed using the ast_taskprocessor_push_local() call. This allows the task to have both data specific to that taskprocessor, in addition to data specific to that invocation. With those two changes, the dispatch object can be removed completely, and the message is simply refcounted and sent directly to the taskprocessor. Review: https://reviewboard.asterisk.org/r/2884/ ........ Merged revisions 399887,400138,400178,400180-400181 from http://svn.asterisk.org/svn/asterisk/branches/12 git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@400186 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-09-30 18:55:27 +00:00
router->default_route.callback = callback;
router->default_route.data = data;
ao2_unlock(router);
Multiple revisions 399887,400138,400178,400180-400181 ........ r399887 | dlee | 2013-09-26 10:41:47 -0500 (Thu, 26 Sep 2013) | 1 line Minor performance bump by not allocate manager variable struct if we don't need it ........ r400138 | dlee | 2013-09-30 10:24:00 -0500 (Mon, 30 Sep 2013) | 23 lines Stasis performance improvements This patch addresses several performance problems that were found in the initial performance testing of Asterisk 12. The Stasis dispatch object was allocated as an AO2 object, even though it has a very confined lifecycle. This was replaced with a straight ast_malloc(). The Stasis message router was spending an inordinate amount of time searching hash tables. In this case, most of our routers had 6 or fewer routes in them to begin with. This was replaced with an array that's searched linearly for the route. We more heavily rely on AO2 objects in Asterisk 12, and the memset() in ao2_ref() actually became noticeable on the profile. This was #ifdef'ed to only run when AO2_DEBUG was enabled. After being misled by an erroneous comment in taskprocessor.c during profiling, the wrong comment was removed. Review: https://reviewboard.asterisk.org/r/2873/ ........ r400178 | dlee | 2013-09-30 13:26:27 -0500 (Mon, 30 Sep 2013) | 24 lines Taskprocessor optimization; switch Stasis to use taskprocessors This patch optimizes taskprocessor to use a semaphore for signaling, which the OS can do a better job at managing contention and waiting that we can with a mutex and condition. The taskprocessor execution was also slightly optimized to reduce the number of locks taken. The only observable difference in the taskprocessor implementation is that when the final reference to the taskprocessor goes away, it will execute all tasks to completion instead of discarding the unexecuted tasks. For systems where unnamed semaphores are not supported, a really simple semaphore implementation is provided. (Which gives identical performance as the original taskprocessor implementation). The way we ended up implementing Stasis caused the threadpool to be a burden instead of a boost to performance. This was switched to just use taskprocessors directly for subscriptions. Review: https://reviewboard.asterisk.org/r/2881/ ........ r400180 | dlee | 2013-09-30 13:39:34 -0500 (Mon, 30 Sep 2013) | 28 lines Optimize how Stasis forwards are dispatched This patch optimizes how forwards are dispatched in Stasis. Originally, forwards were dispatched as subscriptions that are invoked on the publishing thread. This did not account for the vast number of forwards we would end up having in the system, and the amount of work it would take to walk though the forward subscriptions. This patch modifies Stasis so that rather than walking the tree of forwards on every dispatch, when forwards and subscriptions are changed, the subscriber list for every topic in the tree is changed. This has a couple of benefits. First, this reduces the workload of dispatching messages. It also reduces contention when dispatching to different topics that happen to forward to the same aggregation topic (as happens with all of the channel, bridge and endpoint topics). Since forwards are no longer subscriptions, the bulk of this patch is simply changing stasis_subscription objects to stasis_forward objects (which, admittedly, I should have done in the first place.) Since this required me to yet again put in a growing array, I finally abstracted that out into a set of ast_vector macros in asterisk/vector.h. Review: https://reviewboard.asterisk.org/r/2883/ ........ r400181 | dlee | 2013-09-30 13:48:57 -0500 (Mon, 30 Sep 2013) | 28 lines Remove dispatch object allocation from Stasis publishing While looking for areas for performance improvement, I realized that an unused feature in Stasis was negatively impacting performance. When a message is sent to a subscriber, a dispatch object is allocated for the dispatch, containing the topic the message was published to, the subscriber the message is being sent to, and the message itself. The topic is actually unused by any subscriber in Asterisk today. And the subscriber is associated with the taskprocessor the message is being dispatched to. First, this patch removes the unused topic parameter from Stasis subscription callbacks. Second, this patch introduces the concept of taskprocessor local data, data that may be set on a taskprocessor and provided along with the data pointer when a task is pushed using the ast_taskprocessor_push_local() call. This allows the task to have both data specific to that taskprocessor, in addition to data specific to that invocation. With those two changes, the dispatch object can be removed completely, and the message is simply refcounted and sent directly to the taskprocessor. Review: https://reviewboard.asterisk.org/r/2884/ ........ Merged revisions 399887,400138,400178,400180-400181 from http://svn.asterisk.org/svn/asterisk/branches/12 git-svn-id: https://origsvn.digium.com/svn/asterisk/trunk@400186 65c4cc65-6c06-0410-ace0-fbb531ad65f3
2013-09-30 18:55:27 +00:00
/* While this implementation can never fail, it used to be able to */
return 0;
}