Re: how hint caching works

Henrik Nordstrom wrote:
>
> Jon Kay wrote:
>
> > Unlike cache digests, hints have a notion of version through
> > last-modified time, and always go for the most recent known version.
> > They also have some other handy metadata.
>
> How much metadata is involved?

We ran simulations in the paper, and they suggest:

> * Data sent over the networking

We only send changes, and only about the nearest copy of such changes.
The simulation in the paper with all Digital's cache traffic amounts
to 0.4kbit/s average for the central-most location if you're running
without distributed hint hierarchies, and 2kbit/s peak. My guess is
that the average real setup would be one or two hundred bits per
second, especially if you have distributed hierachies going.

Updates are packaged up into HTTP packets pushed once every minute or
two. If there are no updates during the interval, no packet is pushed.

> * Disk space

80-160Megs of hint cache. 1% of the average new cache.

> * Memory requirement

Like cache digest, whatever part of the hint cache is cached in
memory. Plus a few additional data structures and additions to
existing data structures. IMHO, the disk is likely to be more of a
worry.

> I guess the first is dependent on the cloud activity, the other two on
> the cloud size?

Hint cache size depends on numbers of URLs in the hint cache cloud.
If the cloud has a low population, the hint cache can be small. But
the algorithms are most effective in populated clouds, so that's what
we encourage.

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