RE: Busy caches on Linux ???

From: Adam Neat <>
Date: Tue, 30 Jun 1998 10:28:40 +1000

On Tuesday, June 30, 1998 9:01 AM, Dax Kelson []
> On Mon, 29 Jun 1998, Andrew Cormack wrote:
> > We currently run a single site cache on a DEC Alpha box, handling just
> > over 500K TCP requests a day. I'm now considering upgrading it to a
> > multi-cache setup in the interests of fault-tolerance, and wondered
> > whether Intel hardware running Linux would be up to the job ? Although
> > I plan to replace one cache with two, I feel I should plan for at least a
> > doubling of the number of connections, so I'd want each of the Linux
> > caches to be capable of handling 500K requests a day. If you are already
> > running a Linux squid cache of similar (or larger!) size to this, I'd be
> > very grateful for any information on what sort of hardware you are using
> >
> > Thanks in advance,
> > Andrew
> PentiumII 266Mhz, 384MB RAM, 2 x 4.5GB UW SCSI for cache, 500K hits per
> day.
> hardly even feeling it

We run 5 servers here (mainly to distribute load). Our main three in our main
access node share most of the load. The front end proxy runs a 450Mhz Intel
Pentium II box with 512Mb of RAM and around 60Gb of 18.2Gb Disk's. The total
cache runs about 130Gb and around 1.5Gb of RAM.

In total we run about 800,000-1400,000 hits per hour on this box alone
depending on load and so forth.

We looked at Sparc boxes but we could have had almost tripple the disk space
with an Intel box than we could for the same $ on a sparc. And to be honest, I
dont think low - medium level sparc's will have that greater or noticable
difference over an fast, well configured, good quality Intel box.

My $0.02 ! :-)



                  IPAX Systems
              Melbourne, Australia
            National Systems Manager
            Ops Tel: 1800 894 894
            Ops Fax: 03 - 9801-8533
        Web Site
Received on Mon Jun 29 1998 - 17:21:45 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:40:53 MST