httpd-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From Dan Kegel <>
Subject real-world benchmarks?
Date Sun, 22 Aug 1999 03:56:42 GMT
Dear Apache-ites,
I'm interested in your opinion about SPECWeb99 
[ ].  Its key features seem to be:
1. It measures how many simultaneous active connections the
   server can handle.
2. Each client maintains a 400kilobit/second load.
3. The load is 70% static, 30% dynamic of various sorts.
4. The static load file set appears to be big enough
   to not fit in most servers' RAM, and grows linearly with 
   number of clients.
5. MSS is set to 1500 bytes (no jumbo packets).

It seems rather closer to real life than previous benchmarks
(Specweb96, ZD WebBench) because the static load file set is 
much larger, and it tests performance of dynamic page 
generation, which earlier tests did not require.

While it does not model the effects of latency in
the real Internet, it is probably a much more fair
benchmark than we've had before.

Do you folks agree?
Anyone tried it with Apache yet?
- Dan

(The above is my opinion alone, and not that of my employer)

View raw message