> On Tue, 29 Sep 1998, Scott Michel wrote:
>
> > The cache digest is a cache-to-cache metadata transfer. I'm posing
> > a more general question: Do we want/need/have to define a cache
> > system-to-cache system metadata exchange,
>
> What is a "cache system"?
Assume you have a hierarchy of Squid servers, like NLANR's. Assume
ISP A decides to use a collection of Brand X cache servers.
You have a cache system. ISP A has a cache system. Maybe "cache cloud"
or "cache fabric" is a better term. Perhaps "meta-autonomous system
of caches" (MASC), but I think that acronym is taken already.
> Cache digest is a cache-to-cache metadata that describes the content of a
> cache. One probably would need other types of cache-to-cache metadata.
>
> What does system-to-system metadata describe?
For example, systems could pass activity metrics for groups of
URLs between "cache clouds". One could also pass application
level URL routing data between cache clouds.
One thing that's been bugging me about so-called "content pushing"
cache systems which "distribute popular content" is how they measure
popularity. Joe Touch's system is pretty straightforward and upfront
about how it does it. Without feedback from either the client proximate
or server proximate caches, you can't really tell. (I'm expressing
an opinion again... ;-)
> > e.g. interop between SkyCache, Inktomi, and Squid?
>
> If you refer to individual caching proxies, they can exchange cache-to-cache
> metadata.
Vide supra.
-scooter
-- Scott Michel | No research ideal ever survives UCLA Computer Science | contact with implementation. PhD Graduate Student |
This archive was generated by hypermail 2b29 : Thu Nov 18 2004 - 11:21:25 MST