<html>
<head>
<meta content="text/html; charset=utf-8" http-equiv="Content-Type">
</head>
<body text="#000000" bgcolor="#FFFFFF">
<br>
-----BEGIN PGP SIGNED MESSAGE----- <br>
Hash: SHA256 <br>
<br>
The rule is simple.<br>
<br>
If threads on processor(s) are in the queue to the disk - the
bottleneck is disk.<br>
If the disks or network interfaces (IO threads) waits execution on
processor(s) - CPU(s) bottleneck.<br>
<br>
PS. And, man, 1600 users is not a high load. Really high load
starting with 15-30k users. ;)<br>
<br>
24.02.16 2:04, Yuri Voinov пишет:<br>
<span style="white-space: pre;">><br>
> Agreed.<br>
><br>
> High-load big enough caches must utilize _an adequate_
hardware<br>
> configuration with enough capacity to meet you expectations.<br>
><br>
> And, of course, cache software configuration must fit this
hardware, to<br>
> maximize approaches.<br>
><br>
> 24.02.16 1:55, Amos Jeffries пишет:<br>
> > [ pPS please dont hijack other peoples threads ... this
has nothing to<br>
> > do with YouTube ]<br>
><br>
> > On 24/02/2016 8:11 a.m., Heiler Bemerguy wrote:<br>
> >><br>
> >> Thanks Alex.<br>
> >><br>
> >> We have a simple cache_dir config like this, with no
"workers" defined:<br>
> >> cache_dir rock /cache2 80000 min-size=0
max-size=32767<br>
> >> cache_dir aufs /cache 320000 96 256 min-size=32768<br>
> >><br>
> >> And we are suffering from a 100% CPU use by a single
squid thread. We<br>
> >> have lots of ram, cores and disk space..<br>
><br>
> > Squid is essentially single-threaded (not completely, so
dual-core has<br>
> > benefit, but close). Without SMP enabled you will not
benefit from those<br>
> > "lots of cores".<br>
><br>
><br>
> >> but also too many users:<br>
> >> Number of clients accessing cache: 1634<br>
> >> Number of HTTP requests received: 3276691<br>
> >> Average HTTP requests per minute since start:
12807.1<br>
> >> Select loop called: 60353401 times, 22.017 ms avg<br>
> >><br>
><br>
> > What GHz rating is each CPU core? 200-250 RPS is
roughly in the range I<br>
> > would expect from a 1.xGHz core going full speed / 100%
usage.<br>
><br>
> > Are you using RAID on the disk storage? IME, RAID can
more than halve<br>
> > the speed of the proxy. Although the CPU thrashing
effect is mostly<br>
> > hidden away out of sight in the disk controller
processor(s).<br>
><br>
><br>
> >> Getting rid of this big aufs and spreading to many
rock stores will<br>
> >> improve things here? I've already shrunk the acls
and<br>
> patterns/regexes etc<br>
> >><br>
><br>
> > YMMV but I doubt it. AUFS has 64 disk I/O threads taking
advantage of<br>
> > those other cores. Without SMP rock is restricted to
fewer threads for<br>
> > its I/O and most work is being done by the main worker
core anyway<br>
> > without the disk IO portion.<br>
><br>
><br>
> > With CPU maxing out as the bottleneck I would be looking
at config<br>
> > perforemance (you say you have done that already), then
Squid SMP<br>
> > workers as the next workaround. With disk efficiencies
later if/when<br>
> > they become relevant.<br>
><br>
> > Amos<br>
><br>
> > _______________________________________________<br>
> > squid-users mailing list<br>
> > <a class="moz-txt-link-abbreviated" href="mailto:squid-users@lists.squid-cache.org">squid-users@lists.squid-cache.org</a><br>
> > <a class="moz-txt-link-freetext" href="http://lists.squid-cache.org/listinfo/squid-users">http://lists.squid-cache.org/listinfo/squid-users</a><br>
><br>
></span><br>
<br>
-----BEGIN PGP SIGNATURE-----
<br>
Version: GnuPG v2
<br>
<br>
iQEcBAEBCAAGBQJWzMJRAAoJENNXIZxhPexGYYkH/jzBXz01wAl3/kvGD75Ya6Nr
<br>
lmdMX6nVx+3UHPIwC24RJJcJrxaRfHAuXA1NSArvu5iM90O9WDkYcJYLO6FCOeuk
<br>
qha3eMl+KTcFpk7GunaZNF1G6O/kIS5VymFs0lc9tuz6W1GNUmuvfcVIXCc7XDC5
<br>
SM6SY/u7gnnRvxhUyYlrLhb7TdS4uQ5HShqBhaMtkAM5LFVLnPAK9rTxEYr2A7Qq
<br>
hQVkn1lBm/sTImd1fhts3Qkr7F3GDUsOpfmMGeGUIvKCPm9sMTc01trK1Rcb3cre
<br>
dhuluPvbFCguuxvDRCMvoU6DpEkvZDjNL+kN+gLI5SfrqlcZjb4xRosKt5d18Y0=
<br>
=i5/S
<br>
-----END PGP SIGNATURE-----
<br>
<br>
</body>
</html>