Re: [squid-users] Commercial Squid tweak speeds things up significantly!

Il 25/11/2011 23:21, - Mikael - ha scritto:
> Our school dept wants to buy a commercial proxy (Squid based) which
> seems to work a whole lot faster than the standard installation (of
> Squid). The performance difference between the two Squid's seems to be
> in how commercial Squid implementation is handling a missed object.
>
> From what I understand their Squid implementation caches the content
> in a normal way, but once a client gets a cache miss, then their Squid
> allows the client to fetch the record -- apparently Squid doesn't do
> it for the client which seems to make the big performance difference.
> Once the object is fetched by the client, Squid intercepts it and
> stores the object for the other clients. I tested this and that

Hi,
I'm no expert (just a few squid deployments) but that doesn't make any
sense to me.

"caches the content in a normal way", i.e. on a cache miss it fetches
the object from the net and stores it for other clients, while on a
cache hit it serves the file from its local depot (I'm oversimplifying
here).

"but once a client gets a cache miss, then their Squid allows the client
to fetch the record" hmmmm....
"Once the object is fetched by the client, Squid intercepts it and
stores the object for the other clients."

How can squid "intercept" a file _after_ it's been downloaded _directly_
from a client, if that file doesn't pass through squid itself ?

> implementation really, really speeds things up. Now I hate to admit
> this but this commercial product is subscription based and that's the
> money which I would rather spend on students and teachers, school
> supplies for them etc.
>
> Is it possible to configure Squid that way by modifying config file,
> or this is more involved than just editing config file?
>
> Thanks!

Also, how did you test squid and that commercial product ? What does
'speed things up' mean ? Higher throughput ? Smaller latency ?
Did you take DNS resolution times into account ?
Did you submit the same workload to the two setups ? Same wan connection
? Was the wan link being used by others during the test ?
etc.
Also, is squid "powerful" enough for your school or do you really need
that extra "performance" (whatever this means) the commercial product
seems to offer ?

I'm not defending squid, nor do I want to put up a controversy, I just
want to highlight the fact that performance comparisons are a very
complex subject due to the number of variables involved, some of which
are not under direct control of the person doing the test. So wrong
result and apparent performance gains are definitely something to beware of.

My 2 (euro) cents.

-- 
Marcello Romani