From: Kevin Wampler (wampler@cs.washington.edu)
Date: Tue Nov 02 2004 - 23:20:44 PST
It has been observed that the paths in the Internet through which packets
are routed often are significantly longer than the minimum-length paths
through which they could be routed. The reasons for this path inflation
are studied in "Quantifying the Causes of path Inflation". The authors
focus on the impact of intra-domain routing policies, the policies by
which packets are routed between peers, and the policies of inter-domain
routing.
Despite the the reluctance of ISPs to disclose their topology or routing
methods the authors do a good job of inferring these qualities.
Furthermore, the decision of to consider the impact of the peering
relationships and policies related to routing traffic between peers and
inter-domain traffic seems to have given a good deal more insight into the
causes of path inflation. I also find the result that most of path
inflation comes as a result of elements outside of the control of a single
ISP (peeing policies and inter-domain traffic) to be interesting,
particularly since policies arising from commercial concerns are found not
to be a significant factor.
Although this paper does provide some insight into the reasons that path
lengths in the current Internet are inflated, it is not clear that enough
information is provided to give to give too much detail on how to remedy
this situation. In particular, it is not clear how exactly what and how
ISPs should share information with each other to achieve shorter path
lengths while still keeping certain information confidential. In any
case, such persuits are not in the aims of this paper, and is does serve
as a general outlike of path inflation problems.
This archive was generated by hypermail 2.1.6 : Tue Nov 02 2004 - 23:20:44 PST