admin@xxxxxx wrote:
Dnia Wt Czerwca 10 2008, 04:19, Amos Jeffries napisał(a):
Hello,
Is there any reason to migrate to Squid3 (or head snapshot 3.1)?
Did anybody do compare functions/performance between 2.6 and 3.x
version?
I have high load (~400req/s) server with COSS filesystems and tproxy on
squid 2.6.18 (now tproxy doesn't work correctly, only transparent
proxy).
Under 600req/sec there is no noticible difference in 2.6+/3.x. Tproxy 4
support may be a reason for you.
What's mean "Under 600req/sec"?
Squid-3 has tested well according to some basic speed tests up to
650req/sec at 500/1000/1500 concurrent connections.
600req/sec on what hardware? Mayby I have slower server than You and there
will be noticable difference?
I have a i386/ 2.1GHz single-CPU machine doing well. I think Adrians
test were done one a late 2.x GHz machine. Slow stuff by most standards.
Exact details of Adrians tests and hardware should be listed in the mail
archives under benchmarking.
Tproxy 4.1 is important reason for me (mayby one and only).
But the COSS support in 3 is buggy with nobody actively testing it.
you are able to change FS or wanting to assist testing and debugging
before you migrate, that may be a blocker problem.
It's strange that nobody testing it, it's very good fs.
Another question.
How stable is HEAD 3.1 with working Tproxy?
I've only had responses from 3 people outside those of us coding it up.
Two were very happy, one had issues that turned out to be kernel
back-port problems. There has been no confirmed bugs in the squid code
itself. Not bad.
I'm considering moving my production interception machine over to using
it in a few weeks.
When will be released stable 3.1?
I was hoping for initial beta release early this month, but things have
slowed a little. The current 3-HEAD code is fairly stable, only 14 known
bugs. The next features to go in are large enough to cause some upset.
So now is a good time to grab alpha copies if you need them for any
non-testing use.
Amos
--
Please use Squid 2.7.STABLE1 or 3.0.STABLE6
