-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1
Martin,
On 5/18/2009 10:47 AM, Martin Gainty wrote:
> the apache httpd [crowd] may cry foul because you are testing with a
> prefork config instead of worker assuming you can scare up another
> processor
I'm happy to re-run the tests using a worker MPM as well. I'm already
testing lots of different TC connectors, so I may as well test the
worker MPM.
> is there a way to run the same httpd test suite with apache worker?
Sure. I've never configured an httpd for worker MPM, but this is
intended to be an out-of-the-box test anyway, so I don't need to know
anything about tweaking that type of worker :)
> data transfer rate refers to the achieved average net bit rate that
> is delivered to the application layer, exclusive of all protocol
> overhead, data packets retransmissions, etc. i would be interested to
> know how you detect protocol overhead and packet retransmits (to
> exclude for your data_transfer_rate calculations)
I don't. I'm allowing ApacheBench to count whatever it wants to count.
This is the full output of an 'ab' run in case you haven't seen it
before (example for my 1MiB test):
- --------- CUT ---------
Server Software: Apache-Coyote/1.1
Server Hostname: localhost
Server Port: 8001
Document Path: /tomcat/1MiB.bin
Document Length: 1048576 bytes
Concurrency Level: 1
Time taken for tests: 609.926 seconds
Complete requests: 100000
Failed requests: 0
Write errors: 0
Total transferred: 104882200000 bytes
HTML transferred: 104857600000 bytes
Requests per second: 163.95 [#/sec] (mean)
Time per request: 6.099 [ms] (mean)
Time per request: 6.099 [ms] (mean, across all concurrent requests)
Transfer rate: 167928.56 [Kbytes/sec] received
Connection Times (ms)
min mean[+/-sd] median max
Connect: 0 0 0.4 0 112
Processing: 5 6 2.6 6 525
Waiting: 0 0 0.1 0 31
Total: 6 6 2.6 6 525
Percentage of the requests served within a certain time (ms)
50% 6
66% 6
75% 6
80% 6
90% 6
95% 6
98% 7
99% 7
100% 525 (longest request)
- --------- CUT ---------
ApacheBench indicates that 104882200000 bytes were transferred but that
104857600000 (24600000 fewer) "HTML" bytes were transferred. Those extra
bytes are most likely the HTTP headers. There are no counts for TCP
handshake, ACK packets, or any other low-level protocol statistics, but
they would be the same for any software since the kernel is taking care
of all of that detail in any case. I'm not comparing OSs, here: I'm
comparing HTTP server software :)
- -chris
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (MingW32)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org/
iEYEARECAAYFAkoRgEUACgkQ9CaO5/Lv0PC4FgCgq0XG6nG/JYn8nqqCuLfLn1e0
S7sAn0daa0Hah643MolLVUYxc8CixPjJ
=Oexn
-----END PGP SIGNATURE-----
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscr...@tomcat.apache.org
For additional commands, e-mail: users-h...@tomcat.apache.org
