Re: Zipf vs. Uniform

From: Pei Cao (cao@Theory.Stanford.EDU)
Date: Tue Nov 30 1999 - 21:09:18 MST


I am going to shamelessly quote my own paper here :-) Our INFOCOM paper
(http://www.cs.wisc.edu/~cao/papers/zipf-implications.html)
calculated the alpha value for UC Berkeley traces, DEC Corporate traces,
as well as a few others. The values range from 0.64 to 0.83; the alpha
values are calculated using MatLab's curve-fitting tool, excluding the top
100 documents. They are slightly higher than the NLANR logs. But there
was a paper from last year's WCW conference arguing that as one goes up
the caching hierarchy, the alpha value actually decreases, which matches the
results here.

I think 0.6 is a good choice. If one is really concerned about high memory
hit ratio, I would say go with 0.4, which are below 90% of the data points
on the curve at http://www.ircache.net/Cache/Statistics/Popularity-Index/.

Any thoughts?

Pei



This archive was generated by hypermail 2b29 : Tue Jul 10 2001 - 12:00:09 MDT