"Predictive" caching

From: Miguel A.L. Paraz <map@dont-contact.us>
Date: Tue, 7 Dec 1999 14:26:38 +0800

Hi,

First of all thanks to those responded about my Squid "data mining" question.
It was useful.

Next, I'm thinking of something - I wonder what your thoughts on it are.
It's something like "predictive caching", where you get objects in advance
in anticipation of your users getting it. How do you know? Consider the
situation where your cache has a peer whose users have the same usage patterns
as yours. You log their ICP query MISSes and queue the objects up for
fetching from your parent or the source, if they match certain criteria (like
if the extensions are cachable, like gif, jpg, html).

While this may increase the number of hits, the number of bytes downloaded
also increases. However this is OK if the fetches are done during off-peak
hours, or, there's a lot of bandwidth available but it's on a high-latency
link like a satellite.

Any thoughts? Thanks!
---m

-- 
Miguel "Migs" A.L. Paraz	            IPhil Communications Network, Inc.
http://www.iphil.net        Business Development/Network Architecture/Training 
5/F 116 Herrera St., Legaspi Village, Makati City, Philippines  +63-2-750-2288 
Received on Mon Dec 06 1999 - 23:36:32 MST

This archive was generated by hypermail pre-2.1.9 : Tue Dec 09 2003 - 16:49:45 MST