Review - Serving DNS Using a Peer-to-peer Lookup Service

There are many issues with the current DNS model that make it attractive to
look for elegant new solutions. While some of the shortcomings discovered in
early versions of DNS are in the process of being addressed, via the DNSSEC
extensions, these fail to alleviate error-prone administrative overhead. Cox
et al. present a possible name lookup service implemented as a distributed
hash table. Their findings indicate that implementing DNS over Chord has
many disadvantages over the current system.

The implementation discussed here, referred to as DDNS, stores and retrieves
resource records using DHash, a Chord DHT. This provides load balancing and
robustness as nodes are spread randomly over the network. In order to
provide authority, the DNSSEC signing model is borrowed. Here, a parent
domain is required to sign the authenticity of joining nodes. The primary
advantage to this model is its reduced demand on administrative
configuration and routing.

Unfortunately, DDNS does not provide mechanisms for some services offered by
conventional DNS. For example, systems which depend on coupling of
administrative hierarchy and the service structure, such as Akamai.
Moreover, DDNS can result in much higher latencies, which are undesirable
for obvious reasons. Equally important, and perhaps a counter to the
administrative errors in current DNS, DDNS requires the cooperation of end
nodes. It is not clear weather or not individuals are willing to contribute
to the greater good.

Lastly, I believe there are questions of scalability in this model. The
hierarchical simplicity of current DNS has proven to be extremely scalable,
regardless of administrative error. DHTs present a fundamentally flat
design, which is difficult to imagine scaling to a every node in the
Internet. Perhaps a hybrid paradigm will allow DHTs and the current
hierarchical design to be mutually beneficial.
Received on Thu Nov 10 2005 - 09:49:57 EST