Re: how many concurrent users?

On Mon, 8 May 2000, Jim Ruderer wrote:

> Hi all,
>
> I've searched the FAQ and the mailing list archive but can't seem to find
> the answer to my question-
>
> - Is there a limit on how many concurrent users can access a Squid cache
> running only in accelerator mode?
>
> Here's a little background info:
>
> - Hardware: Dual PII-400s, 1gb of RAM, 18gb of total storage space. I'm
> not exactly sure how much disk space and RAM will actually be devoted to the
> cache- suggestions welcome!
> - OS is FreeBSD, web server is Apache.
>
> If anyone has any stories about results they've seen running Squid as an
> accelerator, I'm interested in hearing them. Thanks in advance!

If your squid/accelerator is dedicated to one site, then you probably
don't need more disk space than the web server uses.

Measuring concurrent *users* can be tricky, deneding on what
you mean by a user. You will be limited by how many TCP connections
can be open at one time. To be pessimistic, figure each client request
uses two file descriptors in Squid. So if you configure FreeBSD
for 8K file descriptors, then you can probably support 4K client connections
at one time.

If you are reaching that limit, then you may want to carefully consider
persistent connection parameters. You don't want a lot of clients tying
up file descriptors with idle persistent connections. You'd be
better off to lower the timeout to just a few seconds.

Duane W.
Received on Mon May 08 2000 - 15:01:54 MDT