Re: Direct as last resort

From: Clifton Royston <cliftonr@dont-contact.us>
Date: Fri, 30 Jul 1999 19:38:16 -1000 (HST)

Antony writes:
> At the risk of annoying the regulars, can anyone enlighten me on how to do the
> following...
>
> We have several sibling and two parent caches.
> We want the siblings to have priority over the parents
> We want the parents to have priority over direct
> We want to go direct if the parents have failed.
>
> Why? Cost issues. It is significantly cheaper to go via our uplinks
> caches than go direct, however I want to go direct should their
> caches fail (I don't want the support calls).
>
> Currently if we use ICP to manage this, the parent caches -always-
> win (they're slightly closer and our siblings don't have the same
> links we do) so that voids the whole point of having peers.

Dumb question here: why wouldn't you *want* this to be the case? If
the parents end up responding faster to ICP, they're probably going to
deliver the hits faster too. Isn't that what you want?

It may be technically sweet to have a fully distributed network with
load going off to the siblings, but if the parents serve the hits
better, it would seem to me you ought to just let Squid use them. The
sibs would still be useful in that they'd end up picking up hits if the
parents were saturated or down for some reason.

(It's an interesting question you posed though, as to how you would
accomplish preferring the siblings. You might be able to do something
via the experimental cache-digest features?)
  -- Clifton

-- 
 Clifton Royston  --  LavaNet Systems Architect --  cliftonr@lava.net
        "An absolute monarch would be absolutely wise and good.  
           But no man is strong enough to have no interest.  
             Therefore the best king would be Pure Chance.  
              It is Pure Chance that rules the Universe; 
          therefore, and only therefore, life is good." - AC
Received on Fri Jul 30 1999 - 23:24:34 MDT

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:47:38 MST