From: Seth Cooper (scooper@cs.washington.edu)
Date: Mon Nov 08 2004 - 03:23:56 PST
This paper attempts to measure the relative use of the Internet by
different methods of content delivery. The paper mentions three main
methods for content delivery: the World Wide Web, Content Delivery
Networks, and Peer-to-Peer Networks. Then it discusses a measurement
study performed at the University of Washington, that monitored packets
to characterize incoming and outgoing traffic.
One strength of the paper is that it reaches the conclusion that most
Internet traffic is now P2P. It then discusses the possibility of
caching for P2P traffic. If P2P is truly the most prolific traffic on
the Internet, then being able to cache that traffic would have a large
positive effect in saving bandwidth. Reasons for P2P traffic causing so
much traffic are also presented. For instance, the average P2P object
is three orders of magnitude greater than the average web object.
A weakness of the paper is that the study was performed only on a
campus network. It is very likely that the various amounts of traffic
entering and leaving a campus network are in different proportions than
they would be on another network. It would be interesting to see what
kinds of traffic took up most of the bandwidth on a network that was
more business oriented, or for an ISP that provided a service to private
homes.
This paper is relevant because in oder to improve the Internet, it is
important to understand how it is being used. As in software design, it
is important to profile something before you go in and decide which
parts to optimize. This paper is one such attempt to profile the
Internet and discover which parts will give the most gain if worked on.
This archive was generated by hypermail 2.1.6 : Mon Nov 08 2004 - 03:23:51 PST