[ViewVC] Diff of: cvs/AnyEvent/lib/AnyEvent.pm

Comparing AnyEvent/lib/AnyEvent.pm (file contents):
Revision 1.90 by root, Fri Apr 25 14:24:29 2008 UTC vs.
Revision 1.99 by root, Sun Apr 27 17:09:33 2008 UTC

…		…
66		66
67	Of course, if you want lots of policy (this can arguably be somewhat	67	Of course, if you want lots of policy (this can arguably be somewhat
68	useful) and you want to force your users to use the one and only event	68	useful) and you want to force your users to use the one and only event
69	model, you should I<not> use this module.	69	model, you should I<not> use this module.
70		70
		71	#TODO#
		72
		73	Net::IRC3
		74	AnyEvent::HTTPD
		75	AnyEvent::DNS
		76	IO::AnyEvent
		77	Net::FPing
		78	Net::XMPP2
		79	Coro
		80
		81	AnyEvent::IRC
		82	AnyEvent::HTTPD
		83	AnyEvent::DNS
		84	AnyEvent::Handle
		85	AnyEvent::Socket
		86	AnyEvent::FPing
		87	AnyEvent::XMPP
		88	AnyEvent::SNMP
		89	Coro
71		90
72	=head1 DESCRIPTION	91	=head1 DESCRIPTION
73		92
74	L<AnyEvent> provides an identical interface to multiple event loops. This	93	L<AnyEvent> provides an identical interface to multiple event loops. This
75	allows module authors to utilise an event loop without forcing module	94	allows module authors to utilise an event loop without forcing module
…		…
894	});	913	});
895		914
896	$quit->wait;	915	$quit->wait;
897		916
898		917
899	=head1 BENCHMARK	918	=head1 BENCHMARKS
900		919
901	To give you an idea of the performance and overheads that AnyEvent adds	920	To give you an idea of the performance and overheads that AnyEvent adds
902	over the event loops themselves (and to give you an impression of the	921	over the event loops themselves and to give you an impression of the speed
903	speed of various event loops), here is a benchmark of various supported	922	of various event loops I prepared some benchmarks.
904	event models natively and with anyevent. The benchmark creates a lot of	923
905	timers (with a zero timeout) and I/O watchers (watching STDOUT, a pty, to	924	=head2 BENCHMARKING ANYEVENT OVERHEAD
		925
		926	Here is a benchmark of various supported event models used natively and
		927	through anyevent. The benchmark creates a lot of timers (with a zero
		928	timeout) and I/O watchers (watching STDOUT, a pty, to become writable,
906	become writable, which it is), lets them fire exactly once and destroys	929	which it is), lets them fire exactly once and destroys them again.
907	them again.
908		930
909	Rewriting the benchmark to use many different sockets instead of using	931	Source code for this benchmark is found as F<eg/bench> in the AnyEvent
910	the same filehandle for all I/O watchers results in a much longer runtime	932	distribution.
911	(socket creation is expensive), but qualitatively the same figures, so it
912	was not used.
913		933
914	=head2 Explanation of the columns	934	=head3 Explanation of the columns
915		935
916	I<watcher> is the number of event watchers created/destroyed. Since	936	I<watcher> is the number of event watchers created/destroyed. Since
917	different event models feature vastly different performances, each event	937	different event models feature vastly different performances, each event
918	loop was given a number of watchers so that overall runtime is acceptable	938	loop was given a number of watchers so that overall runtime is acceptable
919	and similar between tested event loop (and keep them from crashing): Glib	939	and similar between tested event loop (and keep them from crashing): Glib
…		…
935	signal the end of this phase.	955	signal the end of this phase.
936		956
937	I<destroy> is the time, in microseconds, that it takes to destroy a single	957	I<destroy> is the time, in microseconds, that it takes to destroy a single
938	watcher.	958	watcher.
939		959
940	=head2 Results	960	=head3 Results
941		961
942	name watchers bytes create invoke destroy comment	962	name watchers bytes create invoke destroy comment
943	EV/EV 400000 244 0.56 0.46 0.31 EV native interface	963	EV/EV 400000 244 0.56 0.46 0.31 EV native interface
944	EV/Any 100000 244 2.50 0.46 0.29 EV + AnyEvent watchers	964	EV/Any 100000 244 2.50 0.46 0.29 EV + AnyEvent watchers
945	CoroEV/Any 100000 244 2.49 0.44 0.29 coroutines + Coro::Signal	965	CoroEV/Any 100000 244 2.49 0.44 0.29 coroutines + Coro::Signal
946	Perl/Any 100000 513 4.92 0.87 1.12 pure perl implementation	966	Perl/Any 100000 513 4.92 0.87 1.12 pure perl implementation
947	Event/Event 16000 516 31.88 31.30 0.85 Event native interface	967	Event/Event 16000 516 31.88 31.30 0.85 Event native interface
948	Event/Any 16000 936 39.17 33.63 1.43 Event + AnyEvent watchers	968	Event/Any 16000 590 35.75 31.42 1.08 Event + AnyEvent watchers
949	Glib/Any 16000 1357 98.22 12.41 54.00 quadratic behaviour	969	Glib/Any 16000 1357 98.22 12.41 54.00 quadratic behaviour
950	Tk/Any 2000 1860 26.97 67.98 14.00 SEGV with >> 2000 watchers	970	Tk/Any 2000 1860 26.97 67.98 14.00 SEGV with >> 2000 watchers
951	POE/Event 2000 6644 108.64 736.02 14.73 via POE::Loop::Event	971	POE/Event 2000 6644 108.64 736.02 14.73 via POE::Loop::Event
952	POE/Select 2000 6343 94.13 809.12 565.96 via POE::Loop::Select	972	POE/Select 2000 6343 94.13 809.12 565.96 via POE::Loop::Select
953		973
954	=head2 Discussion	974	=head3 Discussion
955		975
956	The benchmark does I<not> measure scalability of the event loop very	976	The benchmark does I<not> measure scalability of the event loop very
957	well. For example, a select-based event loop (such as the pure perl one)	977	well. For example, a select-based event loop (such as the pure perl one)
958	can never compete with an event loop that uses epoll when the number of	978	can never compete with an event loop that uses epoll when the number of
959	file descriptors grows high. In this benchmark, all events become ready at	979	file descriptors grows high. In this benchmark, all events become ready at
960	the same time, so select/poll-based implementations get an unnatural speed	980	the same time, so select/poll-based implementations get an unnatural speed
961	boost.	981	boost.
		982
		983	Also, note that the number of watchers usually has a nonlinear effect on
		984	overall speed, that is, creating twice as many watchers doesn't take twice
		985	the time - usually it takes longer. This puts event loops tested with a
		986	higher number of watchers at a disadvantage.
		987
		988	To put the range of results into perspective, consider that on the
		989	benchmark machine, handling an event takes roughly 1600 CPU cycles with
		990	EV, 3100 CPU cycles with AnyEvent's pure perl loop and almost 3000000 CPU
		991	cycles with POE.
962		992
963	C<EV> is the sole leader regarding speed and memory use, which are both	993	C<EV> is the sole leader regarding speed and memory use, which are both
964	maximal/minimal, respectively. Even when going through AnyEvent, it uses	994	maximal/minimal, respectively. Even when going through AnyEvent, it uses
965	far less memory than any other event loop and is still faster than Event	995	far less memory than any other event loop and is still faster than Event
966	natively.	996	natively.
…		…
1001	implementation. The design of the POE adaptor class in AnyEvent can not	1031	implementation. The design of the POE adaptor class in AnyEvent can not
1002	really account for this, as session creation overhead is small compared	1032	really account for this, as session creation overhead is small compared
1003	to execution of the state machine, which is coded pretty optimally within	1033	to execution of the state machine, which is coded pretty optimally within
1004	L<AnyEvent::Impl::POE>. POE simply seems to be abysmally slow.	1034	L<AnyEvent::Impl::POE>. POE simply seems to be abysmally slow.
1005		1035
1006	=head2 Summary	1036	=head3 Summary
1007		1037
1008	=over 4	1038	=over 4
1009		1039
1010	=item * Using EV through AnyEvent is faster than any other event loop	1040	=item * Using EV through AnyEvent is faster than any other event loop
1011	(even when used without AnyEvent), but most event loops have acceptable	1041	(even when used without AnyEvent), but most event loops have acceptable
…		…
1015	the actual event loop, only with extremely fast event loops such as EV	1045	the actual event loop, only with extremely fast event loops such as EV
1016	adds AnyEvent significant overhead.	1046	adds AnyEvent significant overhead.
1017		1047
1018	=item * You should avoid POE like the plague if you want performance or	1048	=item * You should avoid POE like the plague if you want performance or
1019	reasonable memory usage.	1049	reasonable memory usage.
		1050
		1051	=back
		1052
		1053	=head2 BENCHMARKING THE LARGE SERVER CASE
		1054
		1055	This benchmark atcually benchmarks the event loop itself. It works by
		1056	creating a number of "servers": each server consists of a socketpair, a
		1057	timeout watcher that gets reset on activity (but never fires), and an I/O
		1058	watcher waiting for input on one side of the socket. Each time the socket
		1059	watcher reads a byte it will write that byte to a random other "server".
		1060
		1061	The effect is that there will be a lot of I/O watchers, only part of which
		1062	are active at any one point (so there is a constant number of active
		1063	fds for each loop iterstaion, but which fds these are is random). The
		1064	timeout is reset each time something is read because that reflects how
		1065	most timeouts work (and puts extra pressure on the event loops).
		1066
		1067	In this benchmark, we use 10000 socketpairs (20000 sockets), of which 100
		1068	(1%) are active. This mirrors the activity of large servers with many
		1069	connections, most of which are idle at any one point in time.
		1070
		1071	Source code for this benchmark is found as F<eg/bench2> in the AnyEvent
		1072	distribution.
		1073
		1074	=head3 Explanation of the columns
		1075
		1076	I<sockets> is the number of sockets, and twice the number of "servers" (as
		1077	each server has a read and write socket end).
		1078
		1079	I<create> is the time it takes to create a socketpair (which is
		1080	nontrivial) and two watchers: an I/O watcher and a timeout watcher.
		1081
		1082	I<request>, the most important value, is the time it takes to handle a
		1083	single "request", that is, reading the token from the pipe and forwarding
		1084	it to another server. This includes deleting the old timeout and creating
		1085	a new one that moves the timeout into the future.
		1086
		1087	=head3 Results
		1088
		1089	name sockets create request
		1090	EV 20000 69.01 11.16
		1091	Perl 20000 73.32 35.87
		1092	Event 20000 212.62 257.32
		1093	Glib 20000 651.16 1896.30
		1094	POE 20000 349.67 12317.24 uses POE::Loop::Event
		1095
		1096	=head3 Discussion
		1097
		1098	This benchmark I<does> measure scalability and overall performance of the
		1099	particular event loop.
		1100
		1101	EV is again fastest. Since it is using epoll on my system, the setup time
		1102	is relatively high, though.
		1103
		1104	Perl surprisingly comes second. It is much faster than the C-based event
		1105	loops Event and Glib.
		1106
		1107	Event suffers from high setup time as well (look at its code and you will
		1108	understand why). Callback invocation also has a high overhead compared to
		1109	the C<< $_->() for .. >>-style loop that the Perl event loop uses. Event
		1110	uses select or poll in basically all documented configurations.
		1111
		1112	Glib is hit hard by its quadratic behaviour w.r.t. many watchers. It
		1113	clearly fails to perform with many filehandles or in busy servers.
		1114
		1115	POE is still completely out of the picture, taking over 1000 times as long
		1116	as EV, and over 100 times as long as the Perl implementation, even though
		1117	it uses a C-based event loop in this case.
		1118
		1119	=head3 Summary
		1120
		1121	=over 4
		1122
		1123	=item * The pure perl implementation performs extremely well, considering
		1124	that it uses select.
		1125
		1126	=item * Avoid Glib or POE in large projects where performance matters.
		1127
		1128	=back
		1129
		1130	=head2 BENCHMARKING SMALL SERVERS
		1131
		1132	While event loops should scale (and select-based ones do not...) even to
		1133	large servers, most programs we (or I :) actually write have only a few
		1134	I/O watchers.
		1135
		1136	In this benchmark, I use the same benchmark program as in the large server
		1137	case, but it uses only eight "servers", of which three are active at any
		1138	one time. This should reflect performance for a small server relatively
		1139	well.
		1140
		1141	The columns are identical to the previous table.
		1142
		1143	=head3 Results
		1144
		1145	name sockets create request
		1146	EV 16 20.00 6.54
		1147	Perl 16 25.75 12.62
		1148	Event 16 81.27 35.86
		1149	Glib 16 32.63 15.48
		1150	POE 16 261.87 276.28 uses POE::Loop::Event
		1151
		1152	=head3 Discussion
		1153
		1154	The benchmark tries to test the performance of a typical small
		1155	server. While knowing how various event loops perform is interesting, keep
		1156	in mind that their overhead in this case is usually not as important, due
		1157	to the small absolute number of watchers (that is, you need efficiency and
		1158	speed most when you have lots of watchers, not when you only have a few of
		1159	them).
		1160
		1161	EV is again fastest.
		1162
		1163	The C-based event loops Event and Glib come in second this time, as the
		1164	overhead of running an iteration is much smaller in C than in Perl (little
		1165	code to execute in the inner loop, and perl's function calling overhead is
		1166	high, and updating all the data structures is costly).
		1167
		1168	The pure perl event loop is much slower, but still competitive.
		1169
		1170	POE also performs much better in this case, but is is still far behind the
		1171	others.
		1172
		1173	=head3 Summary
		1174
		1175	=over 4
		1176
		1177	=item * C-based event loops perform very well with small number of
		1178	watchers, as the management overhead dominates.
1020		1179
1021	=back	1180	=back
1022		1181
1023		1182
1024	=head1 FORK	1183	=head1 FORK

Diff Legend

-–
+Removed lines
-+
+Added lines
-<
+Changed lines
->
+Changed lines

Comparing AnyEvent/lib/AnyEvent.pm (file contents): Revision 1.90 by root, Fri Apr 25 14:24:29 2008 UTC vs. Revision 1.99 by root, Sun Apr 27 17:09:33 2008 UTC

Diff Legend

Comparing AnyEvent/lib/AnyEvent.pm (file contents):
Revision 1.90 by root, Fri Apr 25 14:24:29 2008 UTC vs.
Revision 1.99 by root, Sun Apr 27 17:09:33 2008 UTC