Re: how hint caching works

Andres Kroonmaa wrote:

> Andres Kroonmaa <andre@online.ee> noted:
> Same worries here. It seems that all caches in the cloud have copy of
> metadata of every other cache. This metadata is main memory consumer
> already for single cache. What impact does pushcache add to that memory
> usage?

It's worth worrying about. I strongly suggest reading our paper with
the simulations on that subject, and detail of hint cache algorithms.

> It seems to me that it would be more memory efficient to use single
> central "hint-cache" that does nothing else but centralises hint data
> and provides it to every box in cloud via some ICP-like protocol.
> Have you considered such approach

We did, actually. Some results are in the paper. The problem there
would have been that then we would have been back to having to wait
for an ICP round trip before doing anything. And not only any ol'
round trip, but a round trip to a probably-distant server.

Also, we were very paranoid about central, root-like servers. This
algorithm was developed during the years of the DNS root crisis. We
felt that any algorithm that required White House intervention to
work was seriously flawed :-).

In fact, we have an algorithm in there, the Plaxton algorithm, that
completely distributes even 'root router'-type duties.

-- 
Jon Kay        pushcache.com                      jkay@pushcache.com
http://www.pushcache.com/                             (512) 420-9025
Squid consulting				  'push done right.'