I decided to run some quick tests (very short and small datasets)using Web
Bench to see if there was any differences in the new shm session caching
code, here's what I found. This test was really here to see if there was a
difference between the caching code, but the results are somewhat
surprising!
The server is a dual processor SGI Origin with 6 clients of various speeds
as clients. The original server build is:
Apache/1.3.12 (Unix) mod_ssl/2.6.6 OpenSSL/0.9.5a
The new server build is:
Apache/1.3.14 (Unix) mod_ssl/2.7.1 OpenSSL/0.9.6
These were all SSL tests.
A is Apache/1.3.14/mod_ssl/2.7.1 with the shmcb (new sharing code)
B is Apache/1.3.14/mod_ssl/2.7.1 with the shmht (old session code)
C is Apache/1.3.12/mod_ssl/2.6.6 with shm (shmht)
Test 1 Test 2 Test 3 Avg
A 141.00 146.25 147.00 144.75
B 148.50 139.50 150.00 146.00
C 86.25 90.00 90.00 88.75
The results are pretty close between the shmcb/shmht tests, close enough
that I'd call them the same. But the difference between the old/new build
is huge and way more than I expected! (But hey, I'll take a 65% speed boost
for free anyday!)
Has anyone else seen these types of results? I would be curious to know if
I'm just crazy or made a mistake in my benchmarking somewhere. It doesn't
seem to be a difference between Apache 1.3.12/14, as static tests show
results that are very close, so I assume that it was some change between
mod_ssl 2.6.6/2.7.1 or openssl 0.9.5a/0.9.6.
I also figure that the tests I ran don't show the differences between the
session sharing code very well, but in the real world I assume the shmcb
code does make a real difference.
-Dave
______________________________________________________________________
Apache Interface to OpenSSL (mod_ssl) www.modssl.org
User Support Mailing List [EMAIL PROTECTED]
Automated List Manager [EMAIL PROTECTED]
