Squid Cache Design

Hi All,

  I've been doing a lot of reading on squid, and various implementations.
But, i'm not sure what are the god & bad points about each, if you think
you know anything about proxy design please forward all your comments onto
me..

here is a list of designs i'm concidering.

1) single large cache, just one box with a ton of hdd & ram catering for
all users
2) multiple large caches, serveral large caches, each one catering for a
diffrent user type (perm, dialup, etc)
3) steped cache, a small fast cache handleing all users, parenting to an
other cache with a large hdd.
3) multi steped cache, same as above but with diffrent servers for
diffrent users
4) round-robin cache, a few large caches in a round robin system
5) parent r-r cache, a round-robin cache with a few small caches for
users.

so far all the multi-cache systems are all siblings & proxy-only, but,
how do the caches lower down fill up at all if it's proxy-only to it's
parent ?? or what about proxy-only/siblings amonst caches on the same
level, and then parent to the next? but is it a waste to have an object
stored twice in a parent and localy? As a side note, we are more
intrested in b/w & data savings rather than a speed improvement, so if
that then makes any changes... we also need to have it so that the
caches can be easily upgradred.. either adding another box, or buying the
newest flashy thing, puting it at the top of the tree, then moving all the
rest down one level... ?? if anyone has any comments please pass
them onto me at fatman@orac.net

Thanks

-Russ
Received on Mon Oct 13 1997 - 23:08:05 MDT