Quantifying the Causes of Path Inflation

From: shvet <shvetank_at_gmail.com>
Date: Thu, 16 Nov 2006 02:33:57 -0500

In this paper, the authors have attempted to quantify the causes of path
inflation by studying a trace of 65 ISPs. They find that topology,
routing policy choices and lack of BGP policy controls are major
causes of path inflation.

Key Points:

1) Intra domain traffic engineering is commonplace but has only a
minimal impact on path inflation.

2) There is significant cooperation between adjacent ISPs to avoid
poor routes or load balance traffic across multiple peering links.

3) Many paths that use early-exit are inflated compared to a
hypothetical optimal exit policy.

4) Topology-insensitive load balancing can cause significant path
inflation.

5) Roughly half of all path lengths are inflated due to inter
domain routing because of using AS-path length as a routing metric.

6) 80% of the intra domain paths are not inflated at all suggesting that
intra domain traffic engineering is consistent with latency sensitive
routing.

7) There is little difference in latency between the early and late
exit routing strategies as paths taken by late-exit are usually just
 the reverse of paths taken by early-exit.

8) Peering policies often cause path inflation. The inflation decreases
quickly as number of peering points are increased from about two.

9) The underlying cause of path inflation is the lack of good tools
for ISPs to find better paths. There is an absence of mechanisms in BGP
to enable better path selection.

10) Adding geographic coordinates to route advertisements would help in
reducing path inflation as geographical separation is a good indicator
of latency for the paths that lie within most ISPs.
Received on Thu Nov 16 2006 - 02:34:15 EST

This archive was generated by hypermail 2.2.0 : Thu Nov 16 2006 - 05:01:03 EST