Understanding Availability Paper Review

 The paper is discussing about characterisitcs of a host availability
measurement in a peer-to-peer file sharing system. The work is is providing
an additional point of view for host availability measurement : IP aliasing
imports a significant error into measurements based on probing IPs, rather
than some unique characteristic of he hosts (like their Unique ID).
 The key strength of the paper is that it finds a significant drawback in
previous measurements by means of simple yet informative measurements.
Although, the first motivation for their work is quite intuitive : In
long-terms measurements and especially in peer-to-peer systems (in which
peers are not important servers with fixed IPs), it is very likely that a
host's IP changes over the preiod of measurement.
 One argument on the measurements and their analysis is on the variation of
host-availability distribution among measurements lasting in different
times. Put in other words, the authors are claiming that the availability
distribution curve is decreasing (moveing to left) as the period of
measurements increases. However, it is clear that (1) the speed of this
degradation in availability diagram is very likely to be decreased (over
increasing measurement periods) and (2) There should be an upperbound for
the period of measurement after which our diagrams will converge. We can
measure this upperbound by looking at the speed of change in the diagrams
for different measurment periods.
Received on Mon Nov 14 2005 - 10:21:32 EST