Re: ssl and polygraph

On Wed, 29 Nov 2000, Paonia Ezrine wrote:

> lets say the squid box is not the bottle neck what kind of hit/sec can a
> single pentium 800 (running linux or FreeBSD) running polygrpah generate?

With PolyMix-3, you can get 400-700 req/sec (230-400 hit/sec), far
more than an average single Squid can handle. By altering
(simplifying) the workload, one can saturate 100Mbit link with a
single Polygraph pair.
 
> How much work would it be?

It would probably take about 7-14 days to develop a useable version,
assuming we can use the existing SSL libraries. The libraries'
encryption/decryption speed may be a problem here, but it can be
solved by adding more client server pairs, of course. Please contact
me off-list of you are interested.

> do you know of any other tools that would do this and fast and
> free? tried pagepoker and wtest both of too slow?

I read that the latest release of httperf had SSL support. Httperf
has somewhat different objectives from Polygraph, but may be worth a
try: 

	http://www.epm.ornl.gov/~sigmetrics/cgi-bin/viewmsg.cgi?id=8
	http://www.hpl.hp.com/personal/David_Mosberger/httperf.html
	ftp://ftp.hpl.hp.com/pub/httperf/

Good luck,

Alex.