Re: [squid-users] squid performance of 2.4 much worse than 2.2 and 2.3?

Klavs Klavsen wrote:

> - however, I'm sorry to hear, that I need to disable disk-cache to use my
> memory optimally.. all the other reverse-proxy boxes you see have a lot of
> memory -
> and they seemingly use it. Howcome squid doesn't?

Because the Squid cache is not very optimized for accelerator/reverse
proxy use. Reverse proxying / acceleration performance has not been the
major focus of Squid development. For a proxy cache the requirements on
the memory cache is quite different (mostly voilatile) than from an
accelerator (more static) making the current design work quite fine.

Almost all Squid development today are driven by needs, or for the fun
of it. For the majority of the developers other needs than very good
reverse proxy cache performance is higher on the list.

Most of the developers (including me) are open to commercial agreements
if there is someone having a different need but not the human resources
to fulfull it.

Yes, Squid is free software. Free does not mean that new features or
improvements appears magically at no cost, it basically means that
anyone is free to adopt the software to fit his own needs without any
charges other than his own expenses. If no changes is required then
obviously the software cost is extremely low, if changes are required
then the software cost increases one way or another, either indirectly
in time the one having the need needs to spend on changing the software
to fit ones needs, or in a direct cost by hiring someone else to do it.

For example the "rproxy" patch was developed by a need from my company
MARA Systems AB to have more flexible reverse proxy configurations, but
there was no need for higher performance, only flexibility, so no
improvements in performance was made as part of that task.

Similarily, the authentication rewrite in Squid-2.5 was triggered by
another companys need for the NTLM authentication scheme.

and the list continues for many of the features you find in or around
Squid.

Other parts of Squid development is triggered by personal needs of
developers to work (for their own amusement) on what then feel is fun
tasks. Many of these current developments are on performance, but it
will take time before any real results are seen as the spare time that
can be spent on such developments is somewhat limited.

Regards
Henrik Nordström
Squid Hacker
Received on Tue Nov 13 2001 - 01:57:18 MST