Alexander Lazic wrote:
>
> Hi,
>
> I have heard on the lighttpd-list that you plan some new benchmarks ;-)
>
> I have read the Mail-Thread from May 2006 and would suggest to not
> exclue the nginx because it's also a server which people use ;-)
I think that a fair comparison between Cherokee and many other mature
web servers could be done only when Cherokee reaches
at least version 0.9.x.
Meanwhile the results of benchmarks should only point to areas where
things should be improved;
of course benchmarks should be performed on a regular basis
(let's say every 3-6 months) to show how much things improve
during the time.
So, it would be a good thing to discuss about
what kind of tests are useful / meaningful
for the current stage of development of Cherokee.
My current list is:
- areas to test:
- static content:
- file range: 0.1KB, 0.8, 2KB, 4KB, 8KB, 16KB ... 16GB
- data set with: 1, 2, 4, 8, 16 ... 16 millions of files
- directory listings:
- files per directory: 1, 10, 100, 1000, 2000, 4000, 8000
- number of directories: 1, 2, 5, 10, 20, 50, 100, 200, ...
- fcgi
- scgi
- cgi:
- program size: 8 KB ... 8 MB
- number of active programs
- error-handling:
- HTTP 200, 3xx, 4xx, 5xx responses.
- scalability:
- number of active connections: 1, 2, 5, 10, 20, 50, 100 ... 50000
- number of threads per CPU: 1 ... 10;
- number of CPUs: 1, 2, 4, 8, 16
- performances:
- requests x sec. + MB/sec. with 1 ... 50000 connections, ecc.
- latency (milliseconds) when accepting a new connection
under load;
- etc.
--
Nick Name: A.D.F.
E-Mail: <adefacc () tin ! it>
--
_______________________________________________
Cherokee mailing list
[email protected]
http://www.0x50.org/cgi-bin/mailman/listinfo/cherokee
