… | |
… | |
3 | AnyEvent::Fork::Pool - simple process pool manager on top of AnyEvent::Fork |
3 | AnyEvent::Fork::Pool - simple process pool manager on top of AnyEvent::Fork |
4 | |
4 | |
5 | =head1 SYNOPSIS |
5 | =head1 SYNOPSIS |
6 | |
6 | |
7 | use AnyEvent; |
7 | use AnyEvent; |
|
|
8 | use AnyEvent::Fork; |
8 | use AnyEvent::Fork::Pool; |
9 | use AnyEvent::Fork::Pool; |
9 | # use AnyEvent::Fork is not needed |
|
|
10 | |
10 | |
11 | # all parameters with default values |
11 | # all possible parameters shown, with default values |
12 | my $pool = AnyEvent::Fork |
12 | my $pool = AnyEvent::Fork |
13 | ->new |
13 | ->new |
14 | ->require ("MyWorker") |
14 | ->require ("MyWorker") |
15 | ->AnyEvent::Fork::Pool::run ( |
15 | ->AnyEvent::Fork::Pool::run ( |
16 | "MyWorker::run", # the worker function |
16 | "MyWorker::run", # the worker function |
17 | |
17 | |
18 | # pool management |
18 | # pool management |
19 | max => 4, # absolute maximum # of processes |
19 | max => 4, # absolute maximum # of processes |
20 | idle => 2, # minimum # of idle processes |
20 | idle => 0, # minimum # of idle processes |
21 | load => 2, # queue at most this number of jobs per process |
21 | load => 2, # queue at most this number of jobs per process |
22 | start => 0.1, # wait this many seconds before starting a new process |
22 | start => 0.1, # wait this many seconds before starting a new process |
23 | stop => 1, # wait this many seconds before stopping an idle process |
23 | stop => 10, # wait this many seconds before stopping an idle process |
24 | on_destroy => (my $finish = AE::cv), # called when object is destroyed |
24 | on_destroy => (my $finish = AE::cv), # called when object is destroyed |
25 | |
25 | |
26 | # parameters passed to AnyEvent::Fork::RPC |
26 | # parameters passed to AnyEvent::Fork::RPC |
27 | async => 0, |
27 | async => 0, |
28 | on_error => sub { die "FATAL: $_[0]\n" }, |
28 | on_error => sub { die "FATAL: $_[0]\n" }, |
… | |
… | |
41 | |
41 | |
42 | $finish->recv; |
42 | $finish->recv; |
43 | |
43 | |
44 | =head1 DESCRIPTION |
44 | =head1 DESCRIPTION |
45 | |
45 | |
46 | This module uses processes created via L<AnyEvent::Fork> and the RPC |
46 | This module uses processes created via L<AnyEvent::Fork> (or |
47 | protocol implement in L<AnyEvent::Fork::RPC> to create a load-balanced |
47 | L<AnyEvent::Fork::Remote>) and the RPC protocol implement in |
48 | pool of processes that handles jobs. |
48 | L<AnyEvent::Fork::RPC> to create a load-balanced pool of processes that |
|
|
49 | handles jobs. |
49 | |
50 | |
50 | Understanding of L<AnyEvent::Fork> is helpful but not critical to be able |
51 | Understanding L<AnyEvent::Fork> is helpful but not required to use this |
51 | to use this module, but a thorough understanding of L<AnyEvent::Fork::RPC> |
52 | module, but a thorough understanding of L<AnyEvent::Fork::RPC> is, as |
52 | is, as it defines the actual API that needs to be implemented in the |
53 | it defines the actual API that needs to be implemented in the worker |
53 | children. |
54 | processes. |
54 | |
|
|
55 | =head1 EXAMPLES |
|
|
56 | |
55 | |
57 | =head1 PARENT USAGE |
56 | =head1 PARENT USAGE |
|
|
57 | |
|
|
58 | To create a pool, you first have to create a L<AnyEvent::Fork> object - |
|
|
59 | this object becomes your template process. Whenever a new worker process |
|
|
60 | is needed, it is forked from this template process. Then you need to |
|
|
61 | "hand off" this template process to the C<AnyEvent::Fork::Pool> module by |
|
|
62 | calling its run method on it: |
|
|
63 | |
|
|
64 | my $template = AnyEvent::Fork |
|
|
65 | ->new |
|
|
66 | ->require ("SomeModule", "MyWorkerModule"); |
|
|
67 | |
|
|
68 | my $pool = $template->AnyEvent::Fork::Pool::run ("MyWorkerModule::myfunction"); |
|
|
69 | |
|
|
70 | The pool "object" is not a regular Perl object, but a code reference that |
|
|
71 | you can call and that works roughly like calling the worker function |
|
|
72 | directly, except that it returns nothing but instead you need to specify a |
|
|
73 | callback to be invoked once results are in: |
|
|
74 | |
|
|
75 | $pool->(1, 2, 3, sub { warn "myfunction(1,2,3) returned @_" }); |
58 | |
76 | |
59 | =over 4 |
77 | =over 4 |
60 | |
78 | |
61 | =cut |
79 | =cut |
62 | |
80 | |
… | |
… | |
68 | |
86 | |
69 | use Guard (); |
87 | use Guard (); |
70 | use Array::Heap (); |
88 | use Array::Heap (); |
71 | |
89 | |
72 | use AnyEvent; |
90 | use AnyEvent; |
73 | use AnyEvent::Fork; # we don't actually depend on it, this is for convenience |
|
|
74 | use AnyEvent::Fork::RPC; |
91 | use AnyEvent::Fork::RPC; |
75 | |
92 | |
76 | # these are used for the first and last argument of events |
93 | # these are used for the first and last argument of events |
77 | # in the hope of not colliding. yes, I don't like it either, |
94 | # in the hope of not colliding. yes, I don't like it either, |
78 | # but didn't come up with an obviously better alternative. |
95 | # but didn't come up with an obviously better alternative. |
79 | my $magic0 = ':t6Z@HK1N%Dx@_7?=~-7NQgWDdAs6a,jFN=wLO0*jD*1%P'; |
96 | my $magic0 = ':t6Z@HK1N%Dx@_7?=~-7NQgWDdAs6a,jFN=wLO0*jD*1%P'; |
80 | my $magic1 = '<~53rexz.U`!]X[A235^"fyEoiTF\T~oH1l/N6+Djep9b~bI9`\1x%B~vWO1q*'; |
97 | my $magic1 = '<~53rexz.U`!]X[A235^"fyEoiTF\T~oH1l/N6+Djep9b~bI9`\1x%B~vWO1q*'; |
81 | |
98 | |
82 | our $VERSION = 0.1; |
99 | our $VERSION = 1.2; |
83 | |
100 | |
84 | =item my $pool = AnyEvent::Fork::Pool::run $fork, $function, [key => value...] |
101 | =item my $pool = AnyEvent::Fork::Pool::run $fork, $function, [key => value...] |
85 | |
102 | |
86 | The traditional way to call it. But it is way cooler to call it in the |
103 | The traditional way to call the pool creation function. But it is way |
87 | following way: |
104 | cooler to call it in the following way: |
88 | |
105 | |
89 | =item my $pool = $fork->AnyEvent::Fork::Pool::run ($function, [key => value...]) |
106 | =item my $pool = $fork->AnyEvent::Fork::Pool::run ($function, [key => value...]) |
90 | |
107 | |
91 | Creates a new pool object with the specified C<$function> as function |
108 | Creates a new pool object with the specified C<$function> as function |
92 | (name) to call for each request. The pool uses the C<$fork> object as the |
109 | (name) to call for each request. The pool uses the C<$fork> object as the |
… | |
… | |
105 | |
122 | |
106 | The pool consists of a certain number of worker processes. These options |
123 | The pool consists of a certain number of worker processes. These options |
107 | decide how many of these processes exist and when they are started and |
124 | decide how many of these processes exist and when they are started and |
108 | stopped. |
125 | stopped. |
109 | |
126 | |
|
|
127 | The worker pool is dynamically resized, according to (perceived :) |
|
|
128 | load. The minimum size is given by the C<idle> parameter and the maximum |
|
|
129 | size is given by the C<max> parameter. A new worker is started every |
|
|
130 | C<start> seconds at most, and an idle worker is stopped at most every |
|
|
131 | C<stop> second. |
|
|
132 | |
|
|
133 | You can specify the amount of jobs sent to a worker concurrently using the |
|
|
134 | C<load> parameter. |
|
|
135 | |
110 | =over 4 |
136 | =over 4 |
111 | |
137 | |
112 | =item idle => $count (default: 0) |
138 | =item idle => $count (default: 0) |
113 | |
139 | |
114 | The minimum amount of idle processes in the pool - when there are fewer |
140 | The minimum amount of idle processes in the pool - when there are fewer |
115 | than this many idle workers, C<AnyEvent::Fork::Pool> will try to start new |
141 | than this many idle workers, C<AnyEvent::Fork::Pool> will try to start new |
116 | ones, subject to C<max> and C<start>. |
142 | ones, subject to the limits set by C<max> and C<start>. |
117 | |
143 | |
118 | This is also the initial/minimum amount of workers in the pool. The |
144 | This is also the initial amount of workers in the pool. The default of |
119 | default of zero means that the pool starts empty and can shrink back to |
145 | zero means that the pool starts empty and can shrink back to zero workers |
120 | zero workers over time. |
146 | over time. |
121 | |
147 | |
122 | =item max => $count (default: 4) |
148 | =item max => $count (default: 4) |
123 | |
149 | |
124 | The maximum number of processes in the pool, in addition to the template |
150 | The maximum number of processes in the pool, in addition to the template |
125 | process. C<AnyEvent::Fork::Pool> will never create more than this number |
151 | process. C<AnyEvent::Fork::Pool> will never have more than this number of |
126 | of worker processes, although there can be more temporarily when a worker |
152 | worker processes, although there can be more temporarily when a worker is |
127 | is shut down and hasn't exited yet. |
153 | shut down and hasn't exited yet. |
128 | |
154 | |
129 | =item load => $count (default: 2) |
155 | =item load => $count (default: 2) |
130 | |
156 | |
131 | The maximum number of concurrent jobs sent to a single worker |
157 | The maximum number of concurrent jobs sent to a single worker process. |
132 | process. Worker processes that handle this number of jobs already are |
|
|
133 | called "busy". |
|
|
134 | |
158 | |
135 | Jobs that cannot be sent to a worker immediately (because all workers are |
159 | Jobs that cannot be sent to a worker immediately (because all workers are |
136 | busy) will be queued until a worker is available. |
160 | busy) will be queued until a worker is available. |
137 | |
161 | |
|
|
162 | Setting this low improves latency. For example, at C<1>, every job that |
|
|
163 | is sent to a worker is sent to a completely idle worker that doesn't run |
|
|
164 | any other jobs. The downside is that throughput is reduced - a worker that |
|
|
165 | finishes a job needs to wait for a new job from the parent. |
|
|
166 | |
|
|
167 | The default of C<2> is usually a good compromise. |
|
|
168 | |
138 | =item start => $seconds (default: 0.1) |
169 | =item start => $seconds (default: 0.1) |
139 | |
170 | |
140 | When a job is queued and all workers are busy, a timer is started. If the |
171 | When there are fewer than C<idle> workers (or all workers are completely |
141 | timer elapses and there are still jobs that cannot be queued to a worker, |
172 | busy), then a timer is started. If the timer elapses and there are still |
142 | a new worker is started. |
173 | jobs that cannot be queued to a worker, a new worker is started. |
143 | |
174 | |
144 | This configurs the time that all workers must be busy before a new worker |
175 | This sets the minimum time that all workers must be busy before a new |
145 | is started. Or, put differently, the minimum delay betwene starting new |
176 | worker is started. Or, put differently, the minimum delay between starting |
146 | workers. |
177 | new workers. |
147 | |
178 | |
148 | The delay is zero by default, which means new workers will be started |
179 | The delay is small by default, which means new workers will be started |
149 | without delay. |
180 | relatively quickly. A delay of C<0> is possible, and ensures that the pool |
|
|
181 | will grow as quickly as possible under load. |
150 | |
182 | |
|
|
183 | Non-zero values are useful to avoid "exploding" a pool because a lot of |
|
|
184 | jobs are queued in an instant. |
|
|
185 | |
|
|
186 | Higher values are often useful to improve efficiency at the cost of |
|
|
187 | latency - when fewer processes can do the job over time, starting more and |
|
|
188 | more is not necessarily going to help. |
|
|
189 | |
151 | =item stop => $seconds (default: 1) |
190 | =item stop => $seconds (default: 10) |
152 | |
191 | |
153 | When a worker has no jobs to execute it becomes idle. An idle worker that |
192 | When a worker has no jobs to execute it becomes idle. An idle worker that |
154 | hasn't executed a job within this amount of time will be stopped, unless |
193 | hasn't executed a job within this amount of time will be stopped, unless |
155 | the other parameters say otherwise. |
194 | the other parameters say otherwise. |
156 | |
195 | |
|
|
196 | Setting this to a very high value means that workers stay around longer, |
|
|
197 | even when they have nothing to do, which can be good as they don't have to |
|
|
198 | be started on the netx load spike again. |
|
|
199 | |
|
|
200 | Setting this to a lower value can be useful to avoid memory or simply |
|
|
201 | process table wastage. |
|
|
202 | |
|
|
203 | Usually, setting this to a time longer than the time between load spikes |
|
|
204 | is best - if you expect a lot of requests every minute and little work |
|
|
205 | in between, setting this to longer than a minute avoids having to stop |
|
|
206 | and start workers. On the other hand, you have to ask yourself if letting |
|
|
207 | workers run idle is a good use of your resources. Try to find a good |
|
|
208 | balance between resource usage of your workers and the time to start new |
|
|
209 | workers - the processes created by L<AnyEvent::Fork> itself is fats at |
|
|
210 | creating workers while not using much memory for them, so most of the |
|
|
211 | overhead is likely from your own code. |
|
|
212 | |
157 | =item on_destroy => $callback->() (default: none) |
213 | =item on_destroy => $callback->() (default: none) |
158 | |
214 | |
159 | When a pool object goes out of scope, it will still handle all outstanding |
215 | When a pool object goes out of scope, the outstanding requests are still |
160 | jobs. After that, it will destroy all workers (and also the template |
216 | handled till completion. Only after handling all jobs will the workers |
161 | process if it isn't referenced otherwise). |
217 | be destroyed (and also the template process if it isn't referenced |
|
|
218 | otherwise). |
|
|
219 | |
|
|
220 | To find out when a pool I<really> has finished its work, you can set this |
|
|
221 | callback, which will be called when the pool has been destroyed. |
162 | |
222 | |
163 | =back |
223 | =back |
164 | |
224 | |
165 | =item Template Process |
225 | =item AnyEvent::Fork::RPC Parameters |
166 | |
226 | |
167 | The worker processes are all forked from a single template |
227 | These parameters are all passed more or less directly to |
168 | process. Ideally, all modules and all cdoe used by the worker, as well as |
228 | L<AnyEvent::Fork::RPC>. They are only briefly mentioned here, for |
169 | any shared data structures should be loaded into the template process, to |
229 | their full documentation please refer to the L<AnyEvent::Fork::RPC> |
170 | take advantage of data sharing via fork. |
230 | documentation. Also, the default values mentioned here are only documented |
171 | |
231 | as a best effort - the L<AnyEvent::Fork::RPC> documentation is binding. |
172 | You can create your own template process by creating a L<AnyEvent::Fork> |
|
|
173 | object yourself and passing it as the C<template> parameter, but |
|
|
174 | C<AnyEvent::Fork::Pool> can create one for you, including some standard |
|
|
175 | options. |
|
|
176 | |
232 | |
177 | =over 4 |
233 | =over 4 |
178 | |
234 | |
179 | =item template => $fork (default: C<< AnyEvent::Fork->new >>) |
|
|
180 | |
|
|
181 | The template process to use, if you want to create your own. |
|
|
182 | |
|
|
183 | =item require => \@modules (default: C<[]>) |
|
|
184 | |
|
|
185 | The modules in this list will be laoded into the template process. |
|
|
186 | |
|
|
187 | =item eval => "# perl code to execute in template" (default: none) |
|
|
188 | |
|
|
189 | This is a perl string that is evaluated after creating the template |
|
|
190 | process and after requiring the modules. It can do whatever it wants to |
|
|
191 | configure the process, but it must not do anything that would keep a later |
|
|
192 | fork from working (so must not create event handlers or (real) threads for |
|
|
193 | example). |
|
|
194 | |
|
|
195 | =back |
|
|
196 | |
|
|
197 | =item AnyEvent::Fork::RPC Parameters |
|
|
198 | |
|
|
199 | These parameters are all passed directly to L<AnyEvent::Fork::RPC>. They |
|
|
200 | are only briefly mentioned here, for their full documentation |
|
|
201 | please refer to the L<AnyEvent::Fork::RPC> documentation. Also, the |
|
|
202 | default values mentioned here are only documented as a best effort - |
|
|
203 | L<AnyEvent::Fork::RPC> documentation is binding. |
|
|
204 | |
|
|
205 | =over 4 |
|
|
206 | |
|
|
207 | =item async => $boolean (default: 0) |
235 | =item async => $boolean (default: 0) |
208 | |
236 | |
209 | Whether to sue the synchronous or asynchronous RPC backend. |
237 | Whether to use the synchronous or asynchronous RPC backend. |
210 | |
238 | |
211 | =item on_error => $callback->($message) (default: die with message) |
239 | =item on_error => $callback->($message) (default: die with message) |
212 | |
240 | |
213 | The callback to call on any (fatal) errors. |
241 | The callback to call on any (fatal) errors. |
214 | |
242 | |
… | |
… | |
235 | |
263 | |
236 | my $max = $arg{max} || 4; |
264 | my $max = $arg{max} || 4; |
237 | my $idle = $arg{idle} || 0, |
265 | my $idle = $arg{idle} || 0, |
238 | my $load = $arg{load} || 2, |
266 | my $load = $arg{load} || 2, |
239 | my $start = $arg{start} || 0.1, |
267 | my $start = $arg{start} || 0.1, |
240 | my $stop = $arg{stop} || 1, |
268 | my $stop = $arg{stop} || 10, |
241 | my $on_event = $arg{on_event} || sub { }, |
269 | my $on_event = $arg{on_event} || sub { }, |
242 | my $on_destroy = $arg{on_destroy}; |
270 | my $on_destroy = $arg{on_destroy}; |
243 | |
271 | |
244 | my @rpc = ( |
272 | my @rpc = ( |
245 | async => $arg{async}, |
273 | async => $arg{async}, |
… | |
… | |
258 | |
286 | |
259 | $template |
287 | $template |
260 | ->require ("AnyEvent::Fork::RPC::" . ($arg{async} ? "Async" : "Sync")) |
288 | ->require ("AnyEvent::Fork::RPC::" . ($arg{async} ? "Async" : "Sync")) |
261 | ->eval (' |
289 | ->eval (' |
262 | my ($magic0, $magic1) = @_; |
290 | my ($magic0, $magic1) = @_; |
263 | sub AnyEvent::Fork::Pool::quit() { |
291 | sub AnyEvent::Fork::Pool::retire() { |
264 | AnyEvent::Fork::RPC::on_event $magic0, "quit", $magic1; |
292 | AnyEvent::Fork::RPC::event $magic0, "quit", $magic1; |
265 | } |
293 | } |
266 | ', $magic0, $magic1) |
294 | ', $magic0, $magic1) |
267 | ->eval ($arg{eval}); |
295 | ; |
268 | |
296 | |
269 | $start_worker = sub { |
297 | $start_worker = sub { |
270 | my $proc = [0, 0, undef]; # load, index, rpc |
298 | my $proc = [0, 0, undef]; # load, index, rpc |
271 | |
299 | |
272 | $proc->[2] = $template |
300 | $proc->[2] = $template |
… | |
… | |
298 | $proc->[0] |
326 | $proc->[0] |
299 | or --$nidle; |
327 | or --$nidle; |
300 | |
328 | |
301 | Array::Heap::splice_heap_idx @pool, $proc->[1] |
329 | Array::Heap::splice_heap_idx @pool, $proc->[1] |
302 | if defined $proc->[1]; |
330 | if defined $proc->[1]; |
|
|
331 | |
|
|
332 | @$proc = 0; # tell others to leave it be |
303 | }; |
333 | }; |
304 | |
334 | |
305 | $want_start = sub { |
335 | $want_start = sub { |
306 | undef $stop_w; |
336 | undef $stop_w; |
307 | |
337 | |
… | |
… | |
326 | }; |
356 | }; |
327 | |
357 | |
328 | $scheduler = sub { |
358 | $scheduler = sub { |
329 | if (@queue) { |
359 | if (@queue) { |
330 | while (@queue) { |
360 | while (@queue) { |
|
|
361 | @pool or $start_worker->(); |
|
|
362 | |
331 | my $proc = $pool[0]; |
363 | my $proc = $pool[0]; |
332 | |
364 | |
333 | if ($proc->[0] < $load) { |
365 | if ($proc->[0] < $load) { |
334 | # found free worker, increase load |
366 | # found free worker, increase load |
335 | unless ($proc->[0]++) { |
367 | unless ($proc->[0]++) { |
… | |
… | |
353 | or $want_stop->(); |
385 | or $want_stop->(); |
354 | |
386 | |
355 | Array::Heap::adjust_heap_idx @pool, $proc->[1] |
387 | Array::Heap::adjust_heap_idx @pool, $proc->[1] |
356 | if defined $proc->[1]; |
388 | if defined $proc->[1]; |
357 | |
389 | |
|
|
390 | &$ocb; |
|
|
391 | |
358 | $scheduler->(); |
392 | $scheduler->(); |
359 | |
|
|
360 | &$ocb; |
|
|
361 | }); |
393 | }); |
362 | } else { |
394 | } else { |
363 | $want_start->() |
395 | $want_start->() |
364 | unless @pool >= $max; |
396 | unless @pool >= $max; |
365 | |
397 | |
366 | last; |
398 | last; |
367 | } |
399 | } |
368 | } |
400 | } |
369 | } elsif ($shutdown) { |
401 | } elsif ($shutdown) { |
370 | @pool = (); |
402 | undef $_->[2] |
|
|
403 | for @pool; |
|
|
404 | |
371 | undef $start_w; |
405 | undef $start_w; |
372 | undef $start_worker; # frees $destroy_guard reference |
406 | undef $start_worker; # frees $destroy_guard reference |
373 | |
407 | |
374 | $stop_worker->($pool[0]) |
408 | $stop_worker->($pool[0]) |
375 | while $nidle; |
409 | while $nidle; |
… | |
… | |
397 | |
431 | |
398 | =item $pool->(..., $cb->(...)) |
432 | =item $pool->(..., $cb->(...)) |
399 | |
433 | |
400 | Call the RPC function of a worker with the given arguments, and when the |
434 | Call the RPC function of a worker with the given arguments, and when the |
401 | worker is done, call the C<$cb> with the results, just like calling the |
435 | worker is done, call the C<$cb> with the results, just like calling the |
402 | L<AnyEvent::Fork::RPC> object directly. |
436 | RPC object durectly - see the L<AnyEvent::Fork::RPC> documentation for |
|
|
437 | details on the RPC API. |
403 | |
438 | |
404 | If there is no free worker, the call will be queued. |
439 | If there is no free worker, the call will be queued until a worker becomes |
|
|
440 | available. |
405 | |
441 | |
406 | Note that there can be considerable time between calling this method and |
442 | Note that there can be considerable time between calling this method and |
407 | the call actually being executed. During this time, the parameters passed |
443 | the call actually being executed. During this time, the parameters passed |
408 | to this function are effectively read-only - modifying them after the call |
444 | to this function are effectively read-only - modifying them after the call |
409 | and before the callback is invoked causes undefined behaviour. |
445 | and before the callback is invoked causes undefined behaviour. |
410 | |
446 | |
411 | =cut |
447 | =cut |
412 | |
448 | |
|
|
449 | =item $cpus = AnyEvent::Fork::Pool::ncpu [$default_cpus] |
|
|
450 | |
|
|
451 | =item ($cpus, $eus) = AnyEvent::Fork::Pool::ncpu [$default_cpus] |
|
|
452 | |
|
|
453 | Tries to detect the number of CPUs (C<$cpus> often called CPU cores |
|
|
454 | nowadays) and execution units (C<$eus>) which include e.g. extra |
|
|
455 | hyperthreaded units). When C<$cpus> cannot be determined reliably, |
|
|
456 | C<$default_cpus> is returned for both values, or C<1> if it is missing. |
|
|
457 | |
|
|
458 | For normal CPU bound uses, it is wise to have as many worker processes |
|
|
459 | as CPUs in the system (C<$cpus>), if nothing else uses the CPU. Using |
|
|
460 | hyperthreading is usually detrimental to performance, but in those rare |
|
|
461 | cases where that really helps it might be beneficial to use more workers |
|
|
462 | (C<$eus>). |
|
|
463 | |
|
|
464 | Currently, F</proc/cpuinfo> is parsed on GNU/Linux systems for both |
|
|
465 | C<$cpus> and C<$eus>, and on {Free,Net,Open}BSD, F<sysctl -n hw.ncpu> is |
|
|
466 | used for C<$cpus>. |
|
|
467 | |
|
|
468 | Example: create a worker pool with as many workers as CPU cores, or C<2>, |
|
|
469 | if the actual number could not be determined. |
|
|
470 | |
|
|
471 | $fork->AnyEvent::Fork::Pool::run ("myworker::function", |
|
|
472 | max => (scalar AnyEvent::Fork::Pool::ncpu 2), |
|
|
473 | ); |
|
|
474 | |
|
|
475 | =cut |
|
|
476 | |
|
|
477 | BEGIN { |
|
|
478 | if ($^O eq "linux") { |
|
|
479 | *ncpu = sub(;$) { |
|
|
480 | my ($cpus, $eus); |
|
|
481 | |
|
|
482 | if (open my $fh, "<", "/proc/cpuinfo") { |
|
|
483 | my %id; |
|
|
484 | |
|
|
485 | while (<$fh>) { |
|
|
486 | if (/^core id\s*:\s*(\d+)/) { |
|
|
487 | ++$eus; |
|
|
488 | undef $id{$1}; |
|
|
489 | } |
|
|
490 | } |
|
|
491 | |
|
|
492 | $cpus = scalar keys %id; |
|
|
493 | } else { |
|
|
494 | $cpus = $eus = @_ ? shift : 1; |
|
|
495 | } |
|
|
496 | wantarray ? ($cpus, $eus) : $cpus |
|
|
497 | }; |
|
|
498 | } elsif ($^O eq "freebsd" || $^O eq "netbsd" || $^O eq "openbsd") { |
|
|
499 | *ncpu = sub(;$) { |
|
|
500 | my $cpus = qx<sysctl -n hw.ncpu> * 1 |
|
|
501 | || (@_ ? shift : 1); |
|
|
502 | wantarray ? ($cpus, $cpus) : $cpus |
|
|
503 | }; |
|
|
504 | } else { |
|
|
505 | *ncpu = sub(;$) { |
|
|
506 | my $cpus = @_ ? shift : 1; |
|
|
507 | wantarray ? ($cpus, $cpus) : $cpus |
|
|
508 | }; |
|
|
509 | } |
|
|
510 | } |
|
|
511 | |
413 | =back |
512 | =back |
414 | |
513 | |
|
|
514 | =head1 CHILD USAGE |
|
|
515 | |
|
|
516 | In addition to the L<AnyEvent::Fork::RPC> API, this module implements one |
|
|
517 | more child-side function: |
|
|
518 | |
|
|
519 | =over 4 |
|
|
520 | |
|
|
521 | =item AnyEvent::Fork::Pool::retire () |
|
|
522 | |
|
|
523 | This function sends an event to the parent process to request retirement: |
|
|
524 | the worker is removed from the pool and no new jobs will be sent to it, |
|
|
525 | but it still has to handle the jobs that are already queued. |
|
|
526 | |
|
|
527 | The parentheses are part of the syntax: the function usually isn't defined |
|
|
528 | when you compile your code (because that happens I<before> handing the |
|
|
529 | template process over to C<AnyEvent::Fork::Pool::run>, so you need the |
|
|
530 | empty parentheses to tell Perl that the function is indeed a function. |
|
|
531 | |
|
|
532 | Retiring a worker can be useful to gracefully shut it down when the worker |
|
|
533 | deems this useful. For example, after executing a job, it could check the |
|
|
534 | process size or the number of jobs handled so far, and if either is too |
|
|
535 | high, the worker could request to be retired, to avoid memory leaks to |
|
|
536 | accumulate. |
|
|
537 | |
|
|
538 | Example: retire a worker after it has handled roughly 100 requests. It |
|
|
539 | doesn't matter whether you retire at the beginning or end of your request, |
|
|
540 | as the worker will continue to handle some outstanding requests. Likewise, |
|
|
541 | it's ok to call retire multiple times. |
|
|
542 | |
|
|
543 | my $count = 0; |
|
|
544 | |
|
|
545 | sub my::worker { |
|
|
546 | |
|
|
547 | ++$count == 100 |
|
|
548 | and AnyEvent::Fork::Pool::retire (); |
|
|
549 | |
|
|
550 | ... normal code goes here |
|
|
551 | } |
|
|
552 | |
|
|
553 | =back |
|
|
554 | |
|
|
555 | =head1 POOL PARAMETERS RECIPES |
|
|
556 | |
|
|
557 | This section describes some recipes for pool parameters. These are mostly |
|
|
558 | meant for the synchronous RPC backend, as the asynchronous RPC backend |
|
|
559 | changes the rules considerably, making workers themselves responsible for |
|
|
560 | their scheduling. |
|
|
561 | |
|
|
562 | =over 4 |
|
|
563 | |
|
|
564 | =item low latency - set load = 1 |
|
|
565 | |
|
|
566 | If you need a deterministic low latency, you should set the C<load> |
|
|
567 | parameter to C<1>. This ensures that never more than one job is sent to |
|
|
568 | each worker. This avoids having to wait for a previous job to finish. |
|
|
569 | |
|
|
570 | This makes most sense with the synchronous (default) backend, as the |
|
|
571 | asynchronous backend can handle multiple requests concurrently. |
|
|
572 | |
|
|
573 | =item lowest latency - set load = 1 and idle = max |
|
|
574 | |
|
|
575 | To achieve the lowest latency, you additionally should disable any dynamic |
|
|
576 | resizing of the pool by setting C<idle> to the same value as C<max>. |
|
|
577 | |
|
|
578 | =item high throughput, cpu bound jobs - set load >= 2, max = #cpus |
|
|
579 | |
|
|
580 | To get high throughput with cpu-bound jobs, you should set the maximum |
|
|
581 | pool size to the number of cpus in your system, and C<load> to at least |
|
|
582 | C<2>, to make sure there can be another job waiting for the worker when it |
|
|
583 | has finished one. |
|
|
584 | |
|
|
585 | The value of C<2> for C<load> is the minimum value that I<can> achieve |
|
|
586 | 100% throughput, but if your parent process itself is sometimes busy, you |
|
|
587 | might need higher values. Also there is a limit on the amount of data that |
|
|
588 | can be "in flight" to the worker, so if you send big blobs of data to your |
|
|
589 | worker, C<load> might have much less of an effect. |
|
|
590 | |
|
|
591 | =item high throughput, I/O bound jobs - set load >= 2, max = 1, or very high |
|
|
592 | |
|
|
593 | When your jobs are I/O bound, using more workers usually boils down to |
|
|
594 | higher throughput, depending very much on your actual workload - sometimes |
|
|
595 | having only one worker is best, for example, when you read or write big |
|
|
596 | files at maximum speed, as a second worker will increase seek times. |
|
|
597 | |
|
|
598 | =back |
|
|
599 | |
|
|
600 | =head1 EXCEPTIONS |
|
|
601 | |
|
|
602 | The same "policy" as with L<AnyEvent::Fork::RPC> applies - exceptions |
|
|
603 | will not be caught, and exceptions in both worker and in callbacks causes |
|
|
604 | undesirable or undefined behaviour. |
|
|
605 | |
415 | =head1 SEE ALSO |
606 | =head1 SEE ALSO |
416 | |
607 | |
417 | L<AnyEvent::Fork>, to create the processes in the first place. |
608 | L<AnyEvent::Fork>, to create the processes in the first place. |
|
|
609 | |
|
|
610 | L<AnyEvent::Fork::Remote>, likewise, but helpful for remote processes. |
418 | |
611 | |
419 | L<AnyEvent::Fork::RPC>, which implements the RPC protocol and API. |
612 | L<AnyEvent::Fork::RPC>, which implements the RPC protocol and API. |
420 | |
613 | |
421 | =head1 AUTHOR AND CONTACT INFORMATION |
614 | =head1 AUTHOR AND CONTACT INFORMATION |
422 | |
615 | |