Using a different threaded test tool that compares the actual output to an expected output, I ran 150k tests with a 100% pass rate. This is with 20 concurrent request threads. To handle the load I added this to my config:
ns_param maxthreads 512 ns_param minthreads 10
I am running PHP 5.0.0, AOLserver 4.0.5 on Linux 2.4.24 (RedHat 7.2).
rob
Dossy Shiobara wrote:
On 2004.07.23, Noah Robin <[EMAIL PROTECTED]> wrote:
So, in the interest of science (and probably masochism) I compiled PHP 4.3.8 (with the aolserver.c patch from panoptic) against aolserver 3.5.10 and 4.0.7 last night, and ran some load tests. Result: on linux 2.4.26, and the only PHP configure option was --with-aolserver=, I was seeing a failure rate of ~6% for a page that only called phpinfo().
Load test: /usr/sbin/ab2 -n100000 -c25 http://localhost:8000/ (ab2 being the version of 'ab' that ships with apache2).
Lets see what I get on my system:
$ uname -a Linux ecwav 2.6.5 #3 Thu May 6 06:57:14 EDT 2004 i686 GNU/Linux
$ cat /proc/cpuinfo | grep "model name" model name : Intel(R) Pentium(R) III CPU family 1266MHz
$ cat test.php <? phpinfo(); ?>
$ ab -c25 -n10000 http://panoptic.com/test.php This is ApacheBench, Version 1.3d <$Revision: 1.67 $> apache-1.3 Copyright (c) 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/ Copyright (c) 1998-2002 The Apache Software Foundation, http://www.apache.org/
Benchmarking panoptic.com (be patient) Completed 1000 requests Completed 2000 requests Completed 3000 requests Completed 4000 requests Completed 5000 requests Completed 6000 requests Completed 7000 requests Completed 8000 requests Completed 9000 requests Finished 10000 requests Server Software: AOLserver/4.0.7 Server Hostname: panoptic.com Server Port: 80
Document Path: /test.php Document Length: 23519 bytes
Concurrency Level: 25 Time taken for tests: 30.294 seconds Complete requests: 10000 Failed requests: 3 (Connect: 0, Length: 3, Exceptions: 0) Broken pipe errors: 0 Total transferred: 236933670 bytes HTML transferred: 235213498 bytes Requests per second: 330.10 [#/sec] (mean) Time per request: 75.73 [ms] (mean) Time per request: 3.03 [ms] (mean, across all concurrent requests) Transfer rate: 7821.14 [Kbytes/sec] received
Connnection Times (ms) min mean[+/-sd] median max Connect: 0 45 370.0 0 9012 Processing: 3 26 139.5 17 6325 Waiting: 0 26 139.5 17 6325 Total: 3 71 398.6 17 9031
Percentage of the requests served within a certain time (ms) 50% 17 66% 18 75% 19 80% 21 90% 29 95% 33 98% 355 99% 3022 100% 9031 (last request)
PHP 4.3.7RC1
This failure rate is obviously not going to cut it for production use, and I was wondering if anyone had seen this before. Nothing in the server.log file during the failures; the access log shows that sometimes the request simply returned zero-length content. Thread stack-size has been set to 1M. Anyone seen this before? =)
Hmm, in my test above, we see 3 failures out of 10,000 -- 0.03% -- is probably reasonable for a production site. Nothing emitted into the server log, nor the php4.log either.
Let me crank up ab to -v99 and see if I can get the failures to happen again ...
Failed requests: 2 (Connect: 0, Length: 2, Exceptions: 0)
OK, so it happened again. Looking at the log ... nothing enlightening.
Lets ask Google. It knows everything ...
Aha. OK -- those "Length" errors can *probably* be ignored -- ApacheBench registers a "Length" failure if the HTTP response length varies from the initial response, which it CAN do for the test.php page that calls phpinfo(), since the output can vary. Look in the "PHP Variables" section towards the bottom of the page at _SERVER["REMOTE_PORT"] -- as it goes from 9999 to 10000, the length of the document goes up by 1 byte causing a length failure. Same thing happens in the other direction, as it goes from 32000 back down to 1000-something.
I agree with you, a true 6% failure rate is unacceptable. On my test environment, I'm seeing effectively a 0% failure rate.
-- Dossy
-- Dossy Shiobara mail: [EMAIL PROTECTED] Panoptic Computer Network web: http://www.panoptic.com/ "He realized the fastest way to change is to laugh at your own folly -- then you can let go and quickly move on." (p. 70)
-- AOLserver - http://www.aolserver.com/
To Remove yourself from this list, simply send an email to <[EMAIL PROTECTED]> with the body of "SIGNOFF AOLSERVER" in the email message. You can leave the Subject: field of your email blank.
-- AOLserver - http://www.aolserver.com/
To Remove yourself from this list, simply send an email to <[EMAIL PROTECTED]> with the body of "SIGNOFF AOLSERVER" in the email message. You can leave the Subject: field of your email blank.
