RE: [squid-users] 50 requests per second

Aaron,

        First, I must state that I am far from an expert with Cisco products.
However, I believe that it is safe to say that for the most part, Cisco
routers have no moving parts, like Hard drives in a computer system. Due
to this fact alone, they are considerably less likely to fail, in
comparison with a server.

        Sure, there are people that have servers running 24/7 for ten years...
Blah, blah, blah... There is still a likely chance that a hard drive will
crash.

        Having a server with moving parts as a single point of failure is quite
a risky venture. Due to the configuration possibilities of a server it
could take 8 or more hours just to bring it back to the state it was
before the failure. Of course, some or most of those services could be
restored far sooner then 8 hours.

        Again, I am unfamiliar with Cisco routers, but I would imagine that if
you have a back-up of the configuration and a known-good spare unit, the
Cisco router configuration could probably be restored within a
half-hour's time.

        Personally, I would rather rely on nearly anyone's hardware pure router
as a single point of failure instead of a machine with a few spinning
magnetic discs.

Regards,
Robert Adkins
IT Manager/Buyer
IMPEL Industries, Inc.

 -----Original Message-----
From: Aaron Seelye [mailto:AaronS@et-n-m.com]
Sent: Friday, June 21, 2002 2:58 PM
To: mailinglistsquid-users@squid-cache.org; Mark.H.Price@AOC.STATE.NC.US;
Robert Adkins
Cc: squid-users@squid-cache.org
Subject: RE: [squid-users] 50 requests per second?

So your single point of failure is a Cisco 6509? :)

Aaron

> -----Original Message-----
> From: Mark.H.Price@AOC.STATE.NC.US
> [mailto:Mark.H.Price@AOC.STATE.NC.US]
> Sent: Friday, June 21, 2002 10:18 AM
> To: joe@swelltech.com
> Cc: squid-users@squid-cache.org
> Subject: Re: [squid-users] 50 requests per second?
>
>
>
> Yes, while it may not be necessary to use 2 load balanced boxes
> for serving that many, it would be frowned upon in my environment
> to have a single point of failure. Our network is focusing on a Cisco
> 6509 we have that will be doing the load-balancing.
>
> Mark
> ------------------( Forwarded letter 1 follows )---------------------
> Date: Fri, 21 Jun 2002 10:44:32 -0500
> To: Mark.H.Price
> Cc: squid-users@squid-cache.org
> From: joe@swelltech.com
> Sender: squid-users-return-1@gcs.alias
> Subject: Re: [squid-users] 50 requests per second?
>
> Load balancing isn't strictly necessary. We have a single box serving
> 200+ reqs/sec at peak periods (dialup ISP workload, so
> different than a
> LAN, but still should make a huge difference in peak request rate).
>
> It just takes a big box, ReiserFS, and Squid with aufs filesystems.
>
> Mark.H.Price@AOC.STATE.NC.US wrote:
> > I have successfully tested squid with transparent caching, and it is
> > working well for about 100 users. I am going to be
> deploying squid for
> > about 5000 users. I estimate that the proxy will need to
> easily handle
> > 50 requests per second, and perhaps up to 80 or 100 during
> peak usage
> > times.
> >
> > Does anyone have experience with a cache server serving with such
> > a large user base?
> >
> > I am undecided whether I will be using a tree or mesh, but
> load balancing
> > will be necessary. The platform of choice is RedHat Linux.
> >
> > thanks
> >
> > Mark
>
> --
> Joe Cooper <joe@swelltech.com>
> Web caching appliances and support.
> http://www.swelltech.com
>
Received on Mon Jun 24 2002 - 05:47:11 MDT