CSC2231 Review of cooperative web caching

This paper studies the benefits of cooperative web proxy caching in terms
of population, and drew the conclusion that this technique is only
suitable for several small organizations if the total population is less
than the knee. Their study is based on a trace collected from UW and
Microsoft at the same time of a week period.

We agree with the authors cooperative web proxy makes no sense between
large organizations. However, we argue this is also true even for small
organizations. Since the total population of these organizations is not
big, we can just put a proxy cache above all organizations. We think
hierarchy cache proxy design is superior to cooperative web proxy design
in terms of simplicity and deployment. Although the authors claim that
proxy placement are affected by political, geographical factors, we think
these limitations also exist for cooperative web proxy, which has more
serious security problems since it is easily exploited to infer the
behavior of web characteristics of other organizations. An organization
can have several high level cache proxies and thus this inference can be
avoided in some degree.

More important, this paper shows the bottleneck of cache hit rate is
cacheable of the web objects, which implies the gain space of cooperative
web caching is minor. In addition, the authors consider the effect of
cooperative web proxy caching as a big cache with the total population.
This assumption is optimistic, since there is usually some latency for
cache requests between two cache proxies considering the two organizations
are distributed in Internet.

This paper examines the gain of the cooperative web caching through
extensive experiments. From their results, we think very minor gains as
well as security issues prevent the practical use of the cooperative web
proxy caching.
Received on Thu Oct 20 2005 - 10:01:55 EDT