Hi there!
I'm using Squid in a cache-mesh that consists of several ISP top level
caches - as it is now, the ISPs (being competitors on the market) are
very jealous of their bandwidth so that a strict hit-only policy is
enforced on the cooperating peers. When we had been erecting the mesh we
tested cache-digests but they yielded quite high false-hit ratio
(resulting in numerous 'Forwarding Denied' error messages) so we had to
abandon them completely and stick to the good old ICP.
As Squid 2.2 announces the ability to retry failed requests that report
504 Gateway Timeout, a question comes out whether the same mechanism
will be used to process cache-digest false-hits on siblings. This way we
could use them in spite of false-hits and thus reduce the ICP bandwidth
overhead and delays.
Comments? Opinions?
JAR (haven't made 2.2 to run yet :( )
Received on Tue Feb 02 1999 - 06:11:37 MST
This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:44:19 MST