Review - The Design and Implementation of a Next Generation Name Service

The existing DNS infrastructure is insufficient to handle the ever expanding
demand placed on it by the modern Internet. Moreover, malicious attacks,
flash-crowds and complex administration have demonstrated weaknesses in the
system. CoDoNS is an overlay network which uses DHTs and analytically
informed caching to correct these shortcomings. Also, CoDoNS incorporates
backwards compatibility with legacy DNS protocols to easy the transition,
should it be widely adopted.

Ramasubramanian et al. present a discussion on various inadequacies in the
modern DNS. They note that 30% of DNS translations result in over 1 second
latency to object requests on the Internet. CoDoNS is intended to reduce DNS
latencies, as well as alleviate bottlenecks (from DoS and flash-crowds) and
remove administrative issues. It implements Beehive for automatically
replicating DNS mappings throughout the network to match demand and provide
performance guarantees. Beehive is an elegant extension to HDTs, where
replication is based on levels.

While CoDoNS seems to be an intuitive and well evaluated alternative, there
are obvious difficulties in wide-scale deployment on the Internet.
Specifically, an entire switch over to this technology is not practical.
This is because CoDoNS does not support some of the current DNS uses, such
as load distribution in content distribution networks (Akamai). In order to
coexist with legacy DNS and easy the transition, CoDoNS has been designed to
be backwards compatible. Unfortunately, this requires inclusion of several
hacks. Also, in order to maintain the classical hierarchical naming
structure, CoDoNS relies on a public key infrastructure. A concept that is
not fully evaluated in this paper.

Although CoDoNS has several compatibility issues with current DNS
constructs, it is a remarkably elegant system. The authors have clearly
spent some effort to show that a graceful transition from current DNS
infrastructure to that of CoDoNS can be made relatively pain free. Lastly,
with the recent debate over American administrative control over the DNS
root infrastructure, it is arguably the right time to introduce change on
this scale.
Received on Thu Nov 10 2005 - 09:52:30 EST