1 |
=head1 NAME |
2 |
|
3 |
BDB - Asynchronous Berkeley DB access |
4 |
|
5 |
=head1 SYNOPSIS |
6 |
|
7 |
use BDB; |
8 |
|
9 |
=head1 DESCRIPTION |
10 |
|
11 |
=head2 EXAMPLE |
12 |
|
13 |
=head1 REQUEST ANATOMY AND LIFETIME |
14 |
|
15 |
Every request method creates a request. which is a C data structure not |
16 |
directly visible to Perl. |
17 |
|
18 |
During their existance, bdb requests travel through the following states, |
19 |
in order: |
20 |
|
21 |
=over 4 |
22 |
|
23 |
=item ready |
24 |
|
25 |
Immediately after a request is created it is put into the ready state, |
26 |
waiting for a thread to execute it. |
27 |
|
28 |
=item execute |
29 |
|
30 |
A thread has accepted the request for processing and is currently |
31 |
executing it (e.g. blocking in read). |
32 |
|
33 |
=item pending |
34 |
|
35 |
The request has been executed and is waiting for result processing. |
36 |
|
37 |
While request submission and execution is fully asynchronous, result |
38 |
processing is not and relies on the perl interpreter calling C<poll_cb> |
39 |
(or another function with the same effect). |
40 |
|
41 |
=item result |
42 |
|
43 |
The request results are processed synchronously by C<poll_cb>. |
44 |
|
45 |
The C<poll_cb> function will process all outstanding aio requests by |
46 |
calling their callbacks, freeing memory associated with them and managing |
47 |
any groups they are contained in. |
48 |
|
49 |
=item done |
50 |
|
51 |
Request has reached the end of its lifetime and holds no resources anymore |
52 |
(except possibly for the Perl object, but its connection to the actual |
53 |
aio request is severed and calling its methods will either do nothing or |
54 |
result in a runtime error). |
55 |
|
56 |
=back |
57 |
|
58 |
=cut |
59 |
|
60 |
package BDB; |
61 |
|
62 |
no warnings; |
63 |
use strict 'vars'; |
64 |
|
65 |
use base 'Exporter'; |
66 |
|
67 |
BEGIN { |
68 |
our $VERSION = '0.1'; |
69 |
|
70 |
our @BDB_REQ = qw( |
71 |
db_env_create db_env_open db_env_close |
72 |
db_create db_open db_close db_compact db_sync db_put |
73 |
); |
74 |
our @EXPORT = (@BDB_REQ, qw(dbreq_pri dbreq_nice)); |
75 |
our @EXPORT_OK = qw(poll_fileno poll_cb poll_wait flush |
76 |
min_parallel max_parallel max_idle |
77 |
nreqs nready npending nthreads |
78 |
max_poll_time max_poll_reqs); |
79 |
|
80 |
require XSLoader; |
81 |
XSLoader::load ("BDB", $VERSION); |
82 |
} |
83 |
|
84 |
=head2 SUPPORT FUNCTIONS |
85 |
|
86 |
=head3 EVENT PROCESSING AND EVENT LOOP INTEGRATION |
87 |
|
88 |
=over 4 |
89 |
|
90 |
=item $fileno = BDB::poll_fileno |
91 |
|
92 |
Return the I<request result pipe file descriptor>. This filehandle must be |
93 |
polled for reading by some mechanism outside this module (e.g. Event or |
94 |
select, see below or the SYNOPSIS). If the pipe becomes readable you have |
95 |
to call C<poll_cb> to check the results. |
96 |
|
97 |
See C<poll_cb> for an example. |
98 |
|
99 |
=item BDB::poll_cb |
100 |
|
101 |
Process some outstanding events on the result pipe. You have to call this |
102 |
regularly. Returns the number of events processed. Returns immediately |
103 |
when no events are outstanding. The amount of events processed depends on |
104 |
the settings of C<BDB::max_poll_req> and C<BDB::max_poll_time>. |
105 |
|
106 |
If not all requests were processed for whatever reason, the filehandle |
107 |
will still be ready when C<poll_cb> returns. |
108 |
|
109 |
Example: Install an Event watcher that automatically calls |
110 |
BDB::poll_cb with high priority: |
111 |
|
112 |
Event->io (fd => BDB::poll_fileno, |
113 |
poll => 'r', async => 1, |
114 |
cb => \&BDB::poll_cb); |
115 |
|
116 |
=item BDB::max_poll_reqs $nreqs |
117 |
|
118 |
=item BDB::max_poll_time $seconds |
119 |
|
120 |
These set the maximum number of requests (default C<0>, meaning infinity) |
121 |
that are being processed by C<BDB::poll_cb> in one call, respectively |
122 |
the maximum amount of time (default C<0>, meaning infinity) spent in |
123 |
C<BDB::poll_cb> to process requests (more correctly the mininum amount |
124 |
of time C<poll_cb> is allowed to use). |
125 |
|
126 |
Setting C<max_poll_time> to a non-zero value creates an overhead of one |
127 |
syscall per request processed, which is not normally a problem unless your |
128 |
callbacks are really really fast or your OS is really really slow (I am |
129 |
not mentioning Solaris here). Using C<max_poll_reqs> incurs no overhead. |
130 |
|
131 |
Setting these is useful if you want to ensure some level of |
132 |
interactiveness when perl is not fast enough to process all requests in |
133 |
time. |
134 |
|
135 |
For interactive programs, values such as C<0.01> to C<0.1> should be fine. |
136 |
|
137 |
Example: Install an Event watcher that automatically calls |
138 |
BDB::poll_cb with low priority, to ensure that other parts of the |
139 |
program get the CPU sometimes even under high AIO load. |
140 |
|
141 |
# try not to spend much more than 0.1s in poll_cb |
142 |
BDB::max_poll_time 0.1; |
143 |
|
144 |
# use a low priority so other tasks have priority |
145 |
Event->io (fd => BDB::poll_fileno, |
146 |
poll => 'r', nice => 1, |
147 |
cb => &BDB::poll_cb); |
148 |
|
149 |
=item BDB::poll_wait |
150 |
|
151 |
If there are any outstanding requests and none of them in the result |
152 |
phase, wait till the result filehandle becomes ready for reading (simply |
153 |
does a C<select> on the filehandle. This is useful if you want to |
154 |
synchronously wait for some requests to finish). |
155 |
|
156 |
See C<nreqs> for an example. |
157 |
|
158 |
=item BDB::poll |
159 |
|
160 |
Waits until some requests have been handled. |
161 |
|
162 |
Returns the number of requests processed, but is otherwise strictly |
163 |
equivalent to: |
164 |
|
165 |
BDB::poll_wait, BDB::poll_cb |
166 |
|
167 |
=item BDB::flush |
168 |
|
169 |
Wait till all outstanding AIO requests have been handled. |
170 |
|
171 |
Strictly equivalent to: |
172 |
|
173 |
BDB::poll_wait, BDB::poll_cb |
174 |
while BDB::nreqs; |
175 |
|
176 |
=head3 CONTROLLING THE NUMBER OF THREADS |
177 |
|
178 |
=item BDB::min_parallel $nthreads |
179 |
|
180 |
Set the minimum number of AIO threads to C<$nthreads>. The current |
181 |
default is C<8>, which means eight asynchronous operations can execute |
182 |
concurrently at any one time (the number of outstanding requests, |
183 |
however, is unlimited). |
184 |
|
185 |
BDB starts threads only on demand, when an AIO request is queued and |
186 |
no free thread exists. Please note that queueing up a hundred requests can |
187 |
create demand for a hundred threads, even if it turns out that everything |
188 |
is in the cache and could have been processed faster by a single thread. |
189 |
|
190 |
It is recommended to keep the number of threads relatively low, as some |
191 |
Linux kernel versions will scale negatively with the number of threads |
192 |
(higher parallelity => MUCH higher latency). With current Linux 2.6 |
193 |
versions, 4-32 threads should be fine. |
194 |
|
195 |
Under most circumstances you don't need to call this function, as the |
196 |
module selects a default that is suitable for low to moderate load. |
197 |
|
198 |
=item BDB::max_parallel $nthreads |
199 |
|
200 |
Sets the maximum number of AIO threads to C<$nthreads>. If more than the |
201 |
specified number of threads are currently running, this function kills |
202 |
them. This function blocks until the limit is reached. |
203 |
|
204 |
While C<$nthreads> are zero, aio requests get queued but not executed |
205 |
until the number of threads has been increased again. |
206 |
|
207 |
This module automatically runs C<max_parallel 0> at program end, to ensure |
208 |
that all threads are killed and that there are no outstanding requests. |
209 |
|
210 |
Under normal circumstances you don't need to call this function. |
211 |
|
212 |
=item BDB::max_idle $nthreads |
213 |
|
214 |
Limit the number of threads (default: 4) that are allowed to idle (i.e., |
215 |
threads that did not get a request to process within 10 seconds). That |
216 |
means if a thread becomes idle while C<$nthreads> other threads are also |
217 |
idle, it will free its resources and exit. |
218 |
|
219 |
This is useful when you allow a large number of threads (e.g. 100 or 1000) |
220 |
to allow for extremely high load situations, but want to free resources |
221 |
under normal circumstances (1000 threads can easily consume 30MB of RAM). |
222 |
|
223 |
The default is probably ok in most situations, especially if thread |
224 |
creation is fast. If thread creation is very slow on your system you might |
225 |
want to use larger values. |
226 |
|
227 |
=item $oldmaxreqs = BDB::max_outstanding $maxreqs |
228 |
|
229 |
This is a very bad function to use in interactive programs because it |
230 |
blocks, and a bad way to reduce concurrency because it is inexact: Better |
231 |
use an C<aio_group> together with a feed callback. |
232 |
|
233 |
Sets the maximum number of outstanding requests to C<$nreqs>. If you |
234 |
to queue up more than this number of requests, the next call to the |
235 |
C<poll_cb> (and C<poll_some> and other functions calling C<poll_cb>) |
236 |
function will block until the limit is no longer exceeded. |
237 |
|
238 |
The default value is very large, so there is no practical limit on the |
239 |
number of outstanding requests. |
240 |
|
241 |
You can still queue as many requests as you want. Therefore, |
242 |
C<max_oustsanding> is mainly useful in simple scripts (with low values) or |
243 |
as a stop gap to shield against fatal memory overflow (with large values). |
244 |
|
245 |
=item BDB::set_sync_prepare $cb |
246 |
|
247 |
Sets a callback that is called whenever a request is created without an |
248 |
explicit callback. It has to return two code references. The first is used |
249 |
as the request callback, and the second is called to wait until the first |
250 |
callback has been called. The default implementation works like this: |
251 |
|
252 |
sub { |
253 |
my $status; |
254 |
( |
255 |
sub { $status = $! }, |
256 |
sub { BDB::poll while !defined $status; $! = $status }, |
257 |
) |
258 |
} |
259 |
|
260 |
=back |
261 |
|
262 |
=head3 STATISTICAL INFORMATION |
263 |
|
264 |
=over 4 |
265 |
|
266 |
=item BDB::nreqs |
267 |
|
268 |
Returns the number of requests currently in the ready, execute or pending |
269 |
states (i.e. for which their callback has not been invoked yet). |
270 |
|
271 |
Example: wait till there are no outstanding requests anymore: |
272 |
|
273 |
BDB::poll_wait, BDB::poll_cb |
274 |
while BDB::nreqs; |
275 |
|
276 |
=item BDB::nready |
277 |
|
278 |
Returns the number of requests currently in the ready state (not yet |
279 |
executed). |
280 |
|
281 |
=item BDB::npending |
282 |
|
283 |
Returns the number of requests currently in the pending state (executed, |
284 |
but not yet processed by poll_cb). |
285 |
|
286 |
=back |
287 |
|
288 |
=cut |
289 |
|
290 |
set_sync_prepare { |
291 |
my $status; |
292 |
( |
293 |
sub { |
294 |
$status = $!; |
295 |
}, |
296 |
sub { |
297 |
BDB::poll while !defined $status; |
298 |
$! = $status; |
299 |
}, |
300 |
) |
301 |
}; |
302 |
|
303 |
min_parallel 8; |
304 |
|
305 |
END { flush } |
306 |
|
307 |
1; |
308 |
|
309 |
=head2 FORK BEHAVIOUR |
310 |
|
311 |
This module should do "the right thing" when the process using it forks: |
312 |
|
313 |
Before the fork, IO::AIO enters a quiescent state where no requests |
314 |
can be added in other threads and no results will be processed. After |
315 |
the fork the parent simply leaves the quiescent state and continues |
316 |
request/result processing, while the child frees the request/result queue |
317 |
(so that the requests started before the fork will only be handled in the |
318 |
parent). Threads will be started on demand until the limit set in the |
319 |
parent process has been reached again. |
320 |
|
321 |
In short: the parent will, after a short pause, continue as if fork had |
322 |
not been called, while the child will act as if IO::AIO has not been used |
323 |
yet. |
324 |
|
325 |
=head2 MEMORY USAGE |
326 |
|
327 |
Per-request usage: |
328 |
|
329 |
Each aio request uses - depending on your architecture - around 100-200 |
330 |
bytes of memory. In addition, stat requests need a stat buffer (possibly |
331 |
a few hundred bytes), readdir requires a result buffer and so on. Perl |
332 |
scalars and other data passed into aio requests will also be locked and |
333 |
will consume memory till the request has entered the done state. |
334 |
|
335 |
This is now awfully much, so queuing lots of requests is not usually a |
336 |
problem. |
337 |
|
338 |
Per-thread usage: |
339 |
|
340 |
In the execution phase, some aio requests require more memory for |
341 |
temporary buffers, and each thread requires a stack and other data |
342 |
structures (usually around 16k-128k, depending on the OS). |
343 |
|
344 |
=head1 KNOWN BUGS |
345 |
|
346 |
Known bugs will be fixed in the next release. |
347 |
|
348 |
=head1 SEE ALSO |
349 |
|
350 |
L<Coro::AIO>. |
351 |
|
352 |
=head1 AUTHOR |
353 |
|
354 |
Marc Lehmann <schmorp@schmorp.de> |
355 |
http://home.schmorp.de/ |
356 |
|
357 |
=cut |
358 |
|