Extremely preliminary WWW benchmark numbers with ISS test3

David S. Miller (davem@jenolan.rutgers.edu)
Wed, 5 Mar 1997 17:35:57 -0500


These are extremely preliminary, please keep this in mind. I've
essentially been setting up a test network to try and get WebStone-2.0
benchmark runs going, to give people some sort of idea of what Linux
is capable of doing on large web servers with the ISS changes.

The current set of results I am about to show were on an SS10 with one
142MHZ HyperSparc processor. Using the on-board Lance ethernet and
all of the server stuff on it's internal 500mb slow scsi disk. This
machine is running stock SparcLinux-2.0.29 + ISS test3 which went out
today. The benchmark client programs were run on 3 SS20's running
SunOS-4.1.3_U1, each of these have 2 50Mhz SuperSparc processors in
them. The web server used was the latest Apache release, compiled
from stock sources.

Essentially, I still believe that there are two problems with these
runs:

1) The 10baseT is saturated completely, it is limiting how
much further I can push the server.

2) I need to add more clients to get more meaningful results
under extremely high load, I'm going to try and get a run
using 20 machines or so, with 16 clients going on each
machine to see what that does. Preferably I'd like to run
all of this over 100baseT, but I lack the facilities
personally to do that at this time.

3) I also need to run a more coherent set of runs with the
recommended client distributions and time lengths from the
WebSTONE home page.

Nevertheless, have a look, this is Linux "Internet Shit Stirrer"
coming at ya:

WEBSTONE 2.0 results:
Total number of clients: 48
Test time: 5 minutes
Server connection rate: 121.00 connections/sec
Server error rate: 0.0000 err/sec
Server thruput: 1.44 Mbit/sec
Little's Load Factor: 47.93
Average response time: 0.3961 sec
Error Level: 0.0000 %
Average client thruput: 31 Kbit/sec
Sum of client response times: 14378.491732 sec
Total number of pages read: 36301

36301 connection(s) to server, 0 errors

Average Std Dev Minimum Maximum

Connect time (sec) 0.114010 0.712250 0.000793 42.033459
Response time (sec) 0.396090 0.825057 0.032083 42.336637
Response size (bytes) 1556 0 1556 1556
Body size (bytes) 759 0 759 759

27552459 body bytes moved + 28931897 header bytes moved = 56484356 total

So, someone who knows something about these things, tell me do those
numbers suck raw eggs or what?

---------------------------------------------////
Yow! 11.26 MB/s remote host TCP bandwidth & ////
199 usec remote TCP latency over 100Mb/s ////
ethernet. Beat that! ////
-----------------------------------------////__________ o
David S. Miller, davem@caip.rutgers.edu /_____________/ / // /_/ ><