Apache Benchmark - High # of failed requests
I read up on using the apache benchmark (ab) tool, and I found quite a surprise, I seem to be having some issues with request failing due to length.
root@girls [/etc/apache2]# ab -c 1 -n 10 www.girls-on-bikes.com/
This is ApacheBench, Version 2.0.41-dev <$Revision: 1.141 $> apache-2.0
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 www.girls-on-bikes.com (be patient).....done
Server Software: Apache/2.0.53
Server Hostname: www.girls-on-bikes.com
Server Port: 80
Document Path: /
Document Length: 29425 bytes
Concurrency Level: 1
Time taken for tests: 14.482593 seconds
Complete requests: 10
Failed requests: 9
(Connect: 0, Length: 9, Exceptions: 0)
Write errors: 0
Total transferred: 299966 bytes
HTML transferred: 294456 bytes
Requests per second: 0.69 [#/sec] (mean)
Time per request: 1448.259 [ms] (mean)
Time per request: 1448.259 [ms] (mean, across all concurrent requests)
Transfer rate: 20.16 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 0 0 0.7 0 2
Processing: 717 1446 707.6 1451 2718
Waiting: 709 1433 708.5 1382 2714
Total: 717 1446 707.6 1451 2718
Percentage of the requests served within a certain time (ms)
50% 1451
66% 1641
75% 1837
80% 2418
90% 2718
95% 2718
98% 2718
99% 2718
100% 2718 (longest request)
Any ideas what to check first?
4 Replies
http://www.celebrazio.net/tech/unix/apache_bench.html
Easy way to verify:
$ wget
Repeat that twice, then:
$ diff <retrieved filename=""> <retrieved filename="">.1</retrieved></retrieved>
See if there are differences. If yes, then you can ignore the length-related failures. If no, it's some other cause and need to investigate further.
I did wget twice, and there were large differences in the files. The content is indeed dynamically created, and includes sessionid's in the code, so I see how that would cause the length issues.
I would like to benchmark my apache installation if possible. Do you know of any benchmarkers that could be used with dynamic content?
Thanks,
-Kevin
Thank you for the suggestion on the siege program, this seems to be working.