Re: [squid-users] Sizing a pretty Large Squid Install

Hi Mark,

I've written a tuning article about Squid on Linux in the past (as soon
as I have the time, I'll update it to reflect that Squid 2.4 is now
about as fast as the version documented in the article):

http://www.swelltech.com/pengies/joe/squidtuneup/t1.html

Tuning any reasonably modern hardware with enough RAM using the steps in
the article will get you up to about 100 reqs sec, pretty easily (you'll
want more than 512MB is you're using 32GB of disk space, however--768MB
is probably a safe minimum).

Clustering becomes more cost-effective than building a bigger box
somewhere between 150 and 220 reqs/sec. client load (and by that time
you're into a very big box--at least 1GB RAM, and probably 4 10,000 RPM
disks, and a GHz processor).

I've also written an article on building a caching infrastructure with
Squid (it's focused on our products, but if you have tuned your Squid
properly, then the sizing info will apply to you too):

http://www.swelltech.com/support/sizecache/t1.html

It's also getting some age on it...but the advice is still relevant, and
Squid hasn't gotten any more memory efficient since I wrote it.

Good luck!

Mark Pace Balzan wrote:

> Hi all,
>
> Ive looked abit around the faqs but didnt quite find all the answers I need.
>
> My situation:
>
> I am thinking of setting up squid for caching having worked with it in the
> past, but Im unclear about some sizing issue for a large installation.
> We are a broadband (cable modem) based ISP with a some dialups too. My
> immediate need is to service around 50 - 60 http requests per second.
>
> This got me to the big figure of 4,320,000 per day, which I didnt not find
> in the JANET artice by Martin Hamilton on the FAQ. Also with an average 8k
> object size, I gather id need 32GB hard disk and 512MB ram minimum.
>
> We currently run on a 14 Mbps Internet link, this will be up to 30Mbps by
> the end of the year, so the requests per second will be increasing by quite
> an amount !
> Im estimating 64-80GB hard disk and 1GB RAM minimum for this.
>
> My questions:
>
> At what point do I need to consider clustering ? (Note the above is at one
> single physical location.)
> Practically speaking what is the max load seen in the field/production ?
> Does squid break under such heavy load, if running on appropriate
> hardware/memory ?
> Is anyone aware of any hardware, kernel o/s or squid issues with memory
> addressing over 1GB physical RAM
>
> Im open to any o/s that fits the job (linux or solaris) and any sugeestions
> you may have.
>
> Thanks for your time

                                   --
                      Joe Cooper <joe@swelltech.com>
                  Affordable Web Caching Proxy Appliances
                         http://www.swelltech.com
Received on Tue May 08 2001 - 14:47:03 MDT