You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@httpd.apache.org by David Ziegler <dz...@MIT.EDU> on 2001/05/25 05:44:28 UTC
Apache 2.0 ab
I was just benchmarking my server, and I tried benchmarking with the ab
that comes with Apache 2.0 (.16). I get ridiculously slow numbers:
This is ApacheBench, Version 1.3e <$Revision: 1.69 $> apache-2.0
Copyright (c) 1996 Adam Twiss, Zeus Technology Ltd,
http://www.zeustech.net/
Copyright (c) 1998-2001 The Apache Software Foundation,
http://www.apache.org/
Benchmarking localhost (be patient).....done
Server Software: Apache/2.0.16
Server Hostname: localhost
Server Port: 80
Document Path: /
Document Length: 120 bytes
Concurrency Level: 25
Time taken for tests: 0.16919 seconds
Complete requests: 100
Failed requests: 0
Write errors: 0
Total transferred: 38600 bytes
HTML transferred: 12000 bytes
Requests per second: 5.91 [#/sec] (mean)
Time per request: 4229.75 [ms] (mean)
Time per request: 169.19 [ms] (mean, across all concurent
requests)
Transfer rate: 2.28 [Kbytes/sec] received
Connnection Times (ms)
min mean[+/-sd] median max
Connect: 0 0 0.0 0 1
Processing: 11 3148 3627.0 533 10477
Waiting: 0 1572 2312.8 230 5480
Total: 11 3148 3627.0 533 10477
ERROR: The median and mean for the initial connection time are more than
twice the standard
deviation apart. These results are NOT reliable.
Percentage of the requests served within a certain time (ms)
50% 533
66% 5434
75% 5448
80% 5480
90% 10460
95% 10472
98% 10474
99% 10477
100% 10477 (last request)
But when I benchmark with Apache 1.3.19 ab, I get much more realistic
numbers:
This is ApacheBench, Version 1.3c <$Revision: 1.44 $> apache-1.3
Copyright (c) 1996 Adam Twiss, Zeus Technology Ltd,
http://www.zeustech.net/
Copyright (c) 1998-2000 The Apache Group, http://www.apache.org/
Server Software: Apache/2.0.16
Server Hostname: localhost
Server Port: 80
Document Path: /
Document Length: 120 bytes
Concurrency Level: 25
Time taken for tests: 1.365 seconds
Complete requests: 1000
Failed requests: 0
Total transferred: 388702 bytes
HTML transferred: 120840 bytes
Requests per second: 732.60
Transfer rate: 284.76 kb/s received
Connnection Times (ms)
min avg max
Connect: -6 4 18
Processing: 12 27 98
Total: 6 31 116
It's just serving up a static html file, nothing spectacular. Any ideas
what might be causing these huge differences? Is this something to look
at? Thanks!
-David Ziegler
-dziegler@mit.edu
Re: Apache 2.0 ab
Posted by Martin Kraemer <Ma...@Fujitsu-Siemens.com>.
On Fri, May 25, 2001 at 12:25:36AM -0700, Dirk-Willem van Gulik wrote:
>
> Hit the send button too quickly - setting both run's to do the same number
> of requests, say 4x1000 - might fix it; right now your ab-1 test is 10
> times longer tham your ab-2 test.
No, that's not the answer. I have been observing the exact same behavior on
a SVR4 platform: insanely slow connection times for ab-2.0, while
ab-1.3.20 results looked sensible. There MUST be some problem lurking
in the combination of ab with apr (on some platforms, at least. Maybe
Lachman-derived TCP stacks like in SVR4?)
With FreeBSD, the differences are also noticeable (and
I wonder what the "-t 10" switch is supposed to do,
when it actually runs for 160 seconds!??!!?):
---- ab from apache-1.3.21-dev: ----
% ../src/support/ab -t10 -k -c6 http://deejai2.mch.fsc.net:8007/tenkb.raw
Server Hostname: deejai2.mch.fsc.net
Server Port: 8007
Document Path: /tenkb.raw
Document Length: 10240 bytes
Concurrency Level: 6
Time taken for tests: 10.014 seconds
Complete requests: 3546
Failed requests: 0
Broken pipe errors: 0
Keep-Alive requests: 3515
Total transferred: 37488743 bytes
HTML transferred: 36360651 bytes
Requests per second: 354.10 [#/sec] (mean)
Time per request: 16.94 [ms] (mean)
Time per request: 2.82 [ms] (mean, across all concurrent requests)
Transfer rate: 3743.63 [Kbytes/sec] received
Connnection Times (ms)
min mean[+/-sd] median max
Connect: 0 0 1.0 0 22
Processing: 3 16 8.6 15 328
Waiting: 0 14 8.6 13 327
Total: 3 16 8.7 15 328
Percentage of the requests served within a certain time (ms)
50% 15
66% 16
75% 17
80% 17
90% 18
95% 21
98% 28
99% 32
100% 328 (last request)
---- ab from httpd-2.0.19-dev: ----
% httpd-2.0/support/ab -t10 -k -c6 http://deejai2.mch.fsc.net:8007/tenkb.raw
Server Software: Apache/2.0.19-dev
Server Hostname: deejai2.mch.fsc.net
Server Port: 8007
Document Path: /tenkb.raw
Document Length: 10240 bytes
Concurrency Level: 6
Time taken for tests: 159.76375 seconds
Complete requests: 50000
Failed requests: 0
Write errors: 0
Keep-Alive requests: 49509
Total transferred: 527918871 bytes
HTML transferred: 512039370 bytes
Requests per second: 314.31 [#/sec] (mean)
Time per request: 0.019 [ms] (mean)
Time per request: 0.003 [ms] (mean, across all concurent requests)
Transfer rate: 3240.86 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 0 0 0.2 0 9
Processing: 14 965 568.2 954 2667
Waiting: 0 8 6.5 7 261
Total: 14 965 568.2 954 2667
Percentage of the requests served within a certain time (ms)
50% 954
66% 1254
75% 1428
80% 1526
90% 1728
95% 1895
98% 2025
99% 2121
100% 2667 (longest request)
Martin
--
<Ma...@Fujitsu-Siemens.com> | Fujitsu Siemens
<ma...@apache.org> | 81730 Munich, Germany
Re: Apache 2.0 ab
Posted by Dirk-Willem van Gulik <di...@covalent.net>.
Hit the send button too quickly - setting both run's to do the same number
of requests, say 4x1000 - might fix it; right now your ab-1 test is 10
times longer tham your ab-2 test.
Dw.
Re: Apache 2.0 ab
Posted by Dirk-Willem van Gulik <di...@covalent.net>.
On Thu, 24 May 2001, David Ziegler wrote:
> I was just benchmarking my server, and I tried benchmarking with the ab
> that comes with Apache 2.0 (.16). I get ridiculously slow numbers:
Note the 'warning:'
...
> Total: 11 3148 3627.0 533 10477
> ERROR: The median and mean for the initial connection time are more than
> twice the standarrd deviation apart. These results are NOT reliable.
If you look at the table further down:
> Percentage of the requests served within a certain time (ms)
> 50% 533
> 66% 5434
> 75% 5448
> 80% 5480
> 90% 10460
> 95% 10472
> 98% 10474
> 99% 10477
> 100% 10477 (last request)
You get an idea of what is happening; somehow about half the requests are
handled 10! times as quick as the other half.
Dw