Review: One-Hop Source Routing

The demand of internet reliability is increasing as new real time
applications are being build. Recent studies have shown that the current
internet infrastructure falls short of expected reliability of internet
path. In recent past work has been done to increase the reliability by
recovering from failure using techniques as overlay networks, server
replication and multi-homing. This paper is a continuation of one such
effort and has purposed a scheme called One-hop source routing to
efficiently route the packets through alternate path in case of a link
failure.

Basically this paper is divided into three sections. In the first section
Author has carried out large scale measurement study to analyze the
characteristics of path failures. 67 Placelab vantage points were used to
probe 3,153 internet destinations over one week to analyze the failures.
Following are the finding of this study

- Failures are spread throughout the Internet and last hop failures
dominate other failures for broadband hosts but
for popular servers last hop failures are rare.
- During the week long trace average path to a server was down for 2,561
seconds and 33,630 seconds for broadband host translating into 99.6% and
94.4% availability respectively.
- Most paths in the internet experienced at least one failure. Only 22% of
server paths and 12% of broadband paths were failure-free.
- 66% of all failures to server and 39% of broadband failures were
potentially recoverable through at least one intermediary.

Based on these observation author has evaluated different techniques to
failure recovery using intermediate nodes. Author has purposed the One-hop
source routing algorithm based on random-4 technique to select the
intermediate node for transmission. In One-hop source routing the nodes
need to have no prior knowledge of each other and the paths that exists.
Its only when the failure occur the source selects 4 random nodes to
transmit through and the one which responds earliest is selected as
intermediate node for transmission. In order to provide proof of concept
author has also implemented the one-hop source routing technique and has
discussed the implementation details in third section.

This scheme is another overlay network approach to recover in the event of
path failures. My concerns remain the same as of RON. This technique again
has been tested on small scale and hence no results are present to
demonstrate its working once deployed in large scale internet environment.
It seems the author never seems to consider that as he has not discussed
how many intermediates are needed if one want to implement this scheme.
Also no one know how this will behave once it is exposed to adverse
internet conditions.
Received on Tue Oct 17 2006 - 07:55:42 EDT