From: Ioannis Giotis (giotis@cs.washington.edu)
Date: Tue Nov 02 2004 - 22:31:19 PST
Path inflation is the appearance of unnecessary delay introduced when
looking at end-to-end paths. The authors try to identify the sources of path
inflation by performing trace measurements throughout the internet.
The main causes of path inflation are routing policies of ISPs and
topological properties of the network. The results of the paper suggest that
intra domain traffic or even between 2 ISPs does not introduce significant
delays mainly because the intra domain ISP routing policies are optimized to
reduce delay.
On the other hand, when considering inter domain traffic, traffic
manipulation by the ISPs can result in increased delays and non-optimal
routes.
The paper results are backed up by a lot of real life measurements and
appear to be realistic. The conclusions inferred are also intuitive and the
authors do not seem to be rushing into any conclusions. On the other hand,
it seems that the collected data is not necessarily representative as it was
collected over a short period of time. Secondly, because of the results
highlighting certain aspects of ISPs' policies, more specific benchmarks
targeted to test these effects would also be nice. And of course, most
importantly, Ratul should get a haircut...
Overall, I found the results interesting and believe the authors managed to
give a clear picture on how ISPs' policies and uncontrolled "optimizations"
affect the internet. A lot of questions arise, and certainly more work needs
to be done on how to optimize internet traffic in a way that will enforce
the ISPs to be more cooperative.
This archive was generated by hypermail 2.1.6 : Tue Nov 02 2004 - 22:31:06 PST