RE: [squid-users] What is decent/good squid performance and architecture

From: Chris Robertson <crobertson@dont-contact.us>
Date: Thu, 7 Jul 2005 13:48:45 -0800

> -----Original Message-----
> From: Jos Houtman [mailto:jos@hyves.nl]
> Sent: Wednesday, July 06, 2005 8:09 AM
> To: Robert Borkowski
> Cc: squid-users@squid-cache.org
> Subject: Re: [squid-users] What is decent/good squid performance and
> architecture
>
>
>> That's better, though with such a large growth rate you will need to
>> anticipate network bottlenecks far ahead, and be ready to switch to
>> gigabit on the squids or grow the number of squids, whichever one is
>> cheaper. Set up MRTG or something equivalent to keep an eye on this.
>> Squid has a built in SNMP server that can produce some useful graphs
>> though MRTG.
>
> i was just trying to produce graphs using squidclient. but i will try
> snmp tomorrow :).
>
>>> i think that loadbalancing is based on source ip, instead of url.
>>> so carp wouldnt be an option.
>>> Is that the same CARP I was looking at?
>>> http://squid-docs.sourceforge.net/latest/html/x2398.html
>>
> obviously not, i googled from carp load balacing and it came up with a
> loadbalancer solutions for BSD.

Using ICP queries will likely work fine in your current situation, but if you can put CARP in use (I'm not sure if LVS supports it), it might give better results (less overlap between caches). A bit of poking for LVS and CARP turns up http://www.austintek.com/LVS/LVS-HOWTO/HOWTO/LVS-HOWTO.L7_switch.html#id2949641, which suggests using multicast ICP on separate NICs. *shrug*

>
>>>
>>>>> If you have a load balancer with packet inspection capabilities you
>>>>> can also direct traffic that way. On F5 BigIPs the facility is
>>>>> called iRules. I'm pretty sure NetScaler can do that too.
>>>>>
>>>> That is the kinda solution iam looking for, but then without the
>>>> cost we are pretty new company without the money to buy expensive
>>>> solutions. so we prefer open source solutions.
>>>>
>>>> another point:
>>>> what is your experience with ext2/3 reiserfs?
>>>> our ext3 partitions tend to get corrupted, when used for squid
>>>> caches or simular purposes.
>>>> i tend to change things to reiserfs entirely, but its just a guess.
>>>> does anyone have the same experience?
>>>
>>>
>>> Read the flames on the LKML about ReiserFS and decide if it's stable
>>> for production use ;-)
>>>
>>> I've got six squids handling a similar traffic load to what you
>>> describe (though on a smaller working set) on ext3 with no corruption
>>> issues.
>>> No corruption issues on any other server using ext3 either. Looks
>>> like you have a serious issue to fix there.
>>
> LKLM? i havent been around for long, so please forgive my lack of
> vocabulair :P
> hmm, its strange it only happens on partitions with large directory's
> with alot of small files in it.
> strange, worth a closer look in the future

LKML = Linux Kernel Mailing List. Some say Reiser is unreliable, and ext3 is the only sane choice. Others blame their troubles on ext3 and claim that Reiser is the best.

I only have experience with ext3, so I can't take sides. OTOH, none of my experience has been bad.

Chris
Received on Thu Jul 07 2005 - 15:48:50 MDT

This archive was generated by hypermail pre-2.1.9 : Mon Aug 01 2005 - 12:00:02 MDT