DNS is an important part of the Internet and this paper tries to analyze its performance and the repercussions of its design parameters. Traces were collected at two universities - MIT and Korea - and were analyzed. Focus was on the latency and failures of DNS clients, and effect of cache sharing and TTL on caching effectiveness.
Over a third of all DNS lookups were not answered successfully. 23% of the client lookups in the MIT trace failed to elicit any answer while 13% lookups gave errors in the answer. This seems a fairly high number...wonder what it would be today. What does "not elicit any answer" mean? Packet loss?
As expected, the name popularity is Zipf-distributed. Hence aggressive caching does not necessarily buy us much. That said, I wonder if traces from a university are generic enough to make this comment. Maybe corporations have a different trend? Definitely caching should be useful in scenarios when the number of clients is more varied and large (e.g., ISPs) .
This paper does seem to point out some interesting results about DNS performance and I would vote for keeping it in the reading list.