Oh, yeah...threads.  No, in neither case did threads ever approach 900. 
  In fact, at the time threads had a hard limit in Linux, so I set it to 
512 for my tests, which was plenty for the benchmarks that I did. 
Admittedly, the polygraph benchmark was running on local machines, so 
each request could complete in ~3 seconds for misses and under a second 
for hits.  So the number of concurrent connections doesn't have to be 
very high.

Joe Cooper wrote:
> I've benchmarked it on both Linux and FreeBSD.  I have very little 
> knowledge of FreeBSD, however, so it may not have been properly tuned (I 
> followed Squid and Polygraph tuning instructions, which I think should 
> be suitable for Oops as well).
> 
> I found them to perform very similarly.  Not enough performance 
> difference to consider significant.
> 
> And as I've said before, Oops is great at low loads client numbers, but 
> when scaling up to much higher request rates (anything over about 70 on 
> my test hardware) it falls over.  Same hardware sustains 90-110 running 
> Squid depending on Squid version (2.2STABLE5 being the fastest, with 2.4 
> and 2.5 being somewhat close...2.3 is the slowest).
> 
> The hardware is an old 850 model of our Tsunami line...450MHz K6-2, 256 
> or 512 MB of RAM (have tested with both along the way) and two 7200 RPM 
> IDE disks.
> 
> Edward Millington wrote:
> 
>> All of this do sound good.
>>
>> Joe! When you ran oops, was it on linux?
>>
>> What was the machine config?
>>
>> did you see the thread hit 1000?
>>
>> Thank you very much.

-- 
Joe Cooper <[EMAIL PROTECTED]>
http://www.swelltech.com
Web Caching Appliances and Support

=====================================================================
If you would like to unsubscribe from this list send message to
[EMAIL PROTECTED] with "unsubscribe oops-eng" in message body.
Archive is accessible on http://lists.paco.net/oops-eng/

Дати відповідь електронним листом