Re: ghost bandwidth [squid-1.2beta22]

Jaeho Yang wrote:
>
> Dear sir/madam.
>
> I'm using squids for transparent proxying.
> The works are great for us.
>
> But I saw *odd behavior of running squid*.
>
> I measure the I/O traffic bytes of squids via ipfwadm accounting
> features.
>
> [Group1] clients request
> [Group2] squid request for non-cached objects
> [Group3] response from sites (response of [Group2]
> [Group4] response from squid (cached objects + response from request
> for non-cached objects)
>
> So, I can measure the saving rate by [Group4] - [Group3]. [Group4] -
> [Group3] is the
> saved traffic without using real bandwidth.
>
> But Sometimes, [Group3] is greater than [Group4]. It means that squid
> received the traffic
> from other sites, but it ate the traffic and didn't serve them.
>
> If the difference is subtle, I can't notice them. But the difference is
> serveral Megabytes.

In all probability there are differences in
packet-lengths/fragment-sizes between your internet-side and your
client-side, based on network characteristics, congestion,
lack-of-congestion, and so forth. Packet-overhead adds up.

The real test at the end of the day is: "How much application-data did
you deliver, vs what you would have otherwise?"

You have to go into a whole new level of network analysis if you want to
compare bytes in and bytes out.

Aborted connections can also chew up a bit. Check out quick_abort in the
conf file and the documentation.

D

-- 
-----BEGIN GEEK CODE BLOCK-----
Version: 3.1
GAT d- s++: a C++++$ UL++++B+++S+++C++H++U++V+++$ P+++$ L+++ E-
W+++(--)$ N++ w++$>--- t+ 5++ X+() R+ tv b++++ DI+++ e- h-@ 
------END GEEK CODE BLOCK------