News for hackers. Stuff that matters.
Re:Introducing Google Public DNS: A new DNS resolver from Go... by Goladus on Thursday December 03, 2009 @02:58PM (#975097)
Doing a single query for each domain isn't entirely fair.

    $ dig tb4.fr @208.67.222.222 | grep Query
    ;; Query time: 274 msec 
    $ dig tb4.fr @208.67.222.222 | grep Query 
    ;; Query time: 9 msec 
    $ dig tb4.fr @208.67.222.222 | grep Query 
    ;; Query time: 9 msec 
    $ dig tb4.fr @208.67.222.222 | grep Query 
    ;; Query time: 36 msec

Also, query time only tells you how long the server took to respond, and doesn't include network latency.
Re:Introducing Google Public DNS: A new DNS resolver from Go... by pmjordan on Thursday December 03, 2009 @06:05PM (#975432)
Obviously subsequent queries are going to be faster - at that point it's cached. That's the point of aggressive/speculative caching: it's so the first (and in practice this will be your only) query is faster. If you want to accurately average across multiple measurements, you need to space those measurements hours or days apart.
Re:Introducing Google Public DNS: A new DNS resolver from Go... by Goladus on Friday December 04, 2009 @11:14AM (#976554)
What you really want to do is log performance under real conditions over an extended period of time and then evaluate the results.

Given that the differences in query time between various DNS solutions are significant but minor, you will almost certainly find differences in performance that you hadn't thought about.