GoogleGuy escreveu:
> I have a small home LAN, with 3 people using the router PC as a
> working machine (read: 3 clients) and basically browsing the same sites.
> Plus a couple of laptops. 
    I'm the sysadmin of several networks running squid as transparent 
proxy. My squids were monitored through MRTG and now i'm using 
cacti/rrdtool, so i have several graphs of all of them, including 
REQUEST hit ratio and BYTE hit ratio. For me and seems for you too, the 
REQUEST hit ratio value is useless, we are worried with BYTE hit ratio, 
which really indicated our bandwidth saving.
    What i have noticed, on the last 3-4 years, is that the BYTE hit 
ratio is getting lower each year. And that's somehow expectable. Several 
sites, including those who have stale content, are starting to use site 
generator systems, just like Wiki and other stuff. Not to count that 
sites are really going dynamic and setting the expire values correctly. 
All of these are making the BYTE hit ratio get lower on the last years, 
at least for me. We can't forget all the multemedia content, who usually 
dont get cached because of maximum_object_size. All these stuff 
contributes for the byte hit ratio being NOT too high.
    I have several different networks, including those 'everybody surfs 
the same thing, ACL controlled' and the other side 'everybody surfs 
everything - ISPs'. In both cases, my bandwidth saving varies usually 
from 10 to 15%. That's the 60 minutes counter. The 1 and 5 minutes 
counter varies too much, i dont think they are good for taking those 
measures.
    OK, somebody will say that 10-15% of a LOT of HTTP traffic is a LOT 
of saving. And indeed, it would be. But again, in my cases and certainly 
in yours too, HTTP is not the only internet traffic i have. So, saving 
10-15% in squid doesnt mean i'm saving 10-15% on my internet connection. 
We still have DNS, SMTP/POP3/IMAP4, all those Instant Messaging systems 
and the bandwidth-hungry P2P things.
    With all these numbers and maths, it seems to me that bandwidth 
saving, through squid, wouldnt allow me to have a smaller internet 
connection neither save me from upgrading it on the next month, for example.
    But why using squid anyway ??? That's the point. At least for me, 
squid is VERY useful because of it's ACL control and delay_pool things, 
which allows me traffic-shape HTTP traffic using a VERY complete ACL 
system, which allows me to create rules for virtually anything i neeed 
until now. Logs which can be processed later are other GREAT thing i use 
a lot.
    So, getting back to your situation ..... i'm from Brazil, all those 
squid i told are running in Brazil too. Bandwidth here is not cheap, 
probably as in your country. I would love to have bandwidth cheap as USA 
people has .... but unfortunely i don't and i'm sure you don't as well.
    Anyway, with the actual internet situation (dynamic, multimedia, 
non-http traffic, etc), i still dont think that a 3-6 clients network, 
as it seems to be your case, will give you some interesting BYTE hit 
ratio. It can have some, it will sure did. But i really dont think that 
would make any difference for you or for any other 3-6 clients network.
    And sorry for the huge email ... but i hope you understand my points.
-- Atenciosamente / Sincerily, Leonardo Rodrigues Solutti Tecnologia http://www.solutti.com.br Minha armadilha de SPAM, NÃO mandem email gertrudes@solutti.com.br My SPAMTRAP, do not email it
This archive was generated by hypermail pre-2.1.9 : Wed Aug 01 2007 - 12:00:04 MDT