From: Erika Rice (erice@cs.washington.edu)
Date: Tue Nov 02 2004 - 21:36:39 PST
"Quantifying the Causes of Path Inflation" by Neil Spring, Ratul
Mahajan, and Thomas Anderson:
When the paths taken by packets become inflated because of the decisions
of routers network resources are underutilized. In "Quantifying the
Causes of Path Inflation" Neil Spring, Ratul Mahajan and Thomas Anderson
discuss causes of path inflation and the tests they used to determine
the relative effects of the causes. The authors identified three main
sources of path inflation: intra-domain routing policy, peering
topology, and inter-domain routing policy.
Their results show that intra-domain routing policy does not have much
of an effect on inflation. ISPs know how to efficiently route packets
within their own network. However, peering topology and inter-domain
routing policy have a large effect on path inflation. Thus we can say
that path inflation is largely an effect of interaction between ISPs.
This gives insight into path inflation. During the time of the
development of the Internet and the earlier ARPANET, the network was
more open. Thus, much more thought was given to making sure every thing
worked within a cooperative network where information about topology was
more available. The commercialized and competitive Internet shows these
implicit and explicit assumptions behind these decisions to be
problematic. When information is hidden, routing is a much more
difficult problem.
However, the paper does leave us with a ray of hope. The observations
of the authors lead them to hypothesize that competing ISPs are willing
to do what is best for packet routing when they have the information to
do so. Thus, if techniques can be developed to allow ISPs to exchange
just the information they need to make good routing decisions there is a
good chance they will use them, leading to an ultimate decrease in path
inflation.
This archive was generated by hypermail 2.1.6 : Tue Nov 02 2004 - 21:36:40 PST