DNS Performance and the Effectiveness of Caching

Paper's Contribution: 

A more recent and thorough analysis, as compared to similar studies,
of DNS & as sociated TCP Internet traffic traces and how caching
affects the performance of DNS.

Paper's Critical Ideas:

The paper's primary motivation is to investigate modern-day
performance of DNS, given the fact the system is so widely adopted and
yet so little studied.  Especially since the studies that have been
done have typically concluded that DNS isn't operating as efficiently
as it could be.  And yet despite this research DNS cont inues to gain
in popularity, being used in new and ever-changing ways, such as to
look-up blacklisted mail spammers, by Interne t applications and
services.  The 2 questions this paper seeks to answer is how
reliable/efficient is DNS and what r ole do its scalability mechanisms
of hierarchy and caching play in ensuring performance.  Caching is
specifically studied by ru nning a battery of trace-driven
simulations.

The papers trace data acquisition methods are far superior to those
used in the previous paper, An Analysis of Wide-Area Name Server
Traffic (1992).  As this research studies traces for roughly weeks a t
a time rather than days, and looked at 2 different domains, which
spanned 2 different countries, which greatly increased the chances of
tracking international queries which might suffer from completely
different response patterns due to additional latency.  The paper
raises the concern for preserving user privacy during data
acquisition, as this sensitive point would c ertainly inhibit data
collection.

Paper's Flaws:

Just as in the previous paper, the data collection techniques seemed
overly limi ted for what is really required to give an accurate
picture of DNS usage across the entire Internet.  It would be a vast
im provement to study dozens of representative domains, across
commercial, education and government domains from different regi
ons/countries over many weeks if not months.  But I'm sure the cost of
that kind of collaborative effort far outweigh ed available grant
money in this case.  Overall this paper is quite effective and points
out several interesting data points tha t may help evolve DNS in the
future.

Relevance to modern systems:

The Domain Name System is the primary system for host name/address
mapping in us e on the Internet and yet despite its widespread use,
there aren't all that many studies that track its performance an d
effectiveness.  It would seem that periodic DNS performance analysis
would be a "must-have" for this crucial system, especia lly given the
ever-changing traffic patterns of the Internet's usage.