“You can't optimize it until you can measure it”
As the previous pages have explained, the DNS Benchmark has been designed to both “just work” easily for the casual user, and also to serve as a powerful, flexible, customizable and expandable tool to facilitate (and encourage) the advanced measurement of many aspects of DNS performance.
The Benchmark incorporates two built-in, replaceable lists which, with their default composition, work well to characterize the performance of popular DNS servers. But those lists necessarily incorporate some (US-centric) assumptions about the domains DNS users visit frequently, and about the publicly available alternative DNS resolvers that might form practical alternatives. For some users, more appropriate lists could be developed and provided to override or augment the lists built into the Benchmark.
The two built-in lists perform the following functions:
Though this list is subject to evolution over the lifetime of the benchmark, the v1.0 edition of the DNS Benchmark incorporated and, by default, tested the following seventy (70) publicly available DNS resolvers for comparison against the system's currently configured resolvers, and for comparison to each other:
Note that a current list used by any future version of this benchmark can easily be obtained by running the benchmark, deleting all current “system” and “public” resolvers, adding back only the “public” resolvers (this must be done in this sequence because there could be an overlap between the system and public resolver lists), then using the System Menu to “Save Nameserver to .INI file”, thus saving the Benchmark's current public resolver list as an .INI file . . . which will be a text file formatted as shown above.
You can then edit the file to add or remove any resolvers. Once you have an updated list, you may:
We expect this built-in list to evolve over time as we receive feedback from DNS Benchmark users.
If our Benchmark's users provide any locale-specific publicly accessible
resolver lists, such as for specific geographic regions of the Internet,
we will use this page to share them with all other Benchmark users.
The second list incorporated into the Benchmark is the set of “dot com”, or dot-com like, domain names that are sent to each of the benchmarked resolvers for their resolution. In mid-2009, when the DNS Benchmark was being written, the Alexia database was used to obtain the one hundred globally most popular domains. After removing any that might be objectionable due to their nature, despite their popularity, the following fifty (50) domain names were selected for use by the Benchmark:
That default domain name list is contained in this simple text file: domains.txt.
Since the need to alter this default domains list should be much more rare than altering the system's default resolver list, no provision for replacing the list was built into the Benchmark's user-interface. However, as was detailed on the command-line options page, the /domains command-line option can be used to replace the Benchmark's built-in list with any alternative user-provided list.
Consequently, if the Benchmark's built-in domains list is replaced with one having many fewer than fifty (50) domains, the benchmark will run faster overall, but it will also produce lower reliability results.
And, conversely, if a domain list consisting of more domains were used instead of the Benchmark's current fifty (50) domains default list, the Benchmark's total running time would be increased, but the statistical significance of its results will be improved for increased accuracy.
Here is a list of the top 100 domains (100domains.txt) which could be used to obtain a higher level of statistical significance. Warning! This list has NOT been pruned for possibly-objectionable domain names, so any computer running the Benchmark with this list will apparently be looking up the IPs of all of these domains! This might not be appropriate for use in a corporate setting.
As with the resolver list above, if our Benchmark's users provide any interesting special-
purpose lists of domain names, such as locale-specific domains for specific geographic
regions, we will use this page to share them with all other Benchmark users.
Gibson Research Corporation is owned and operated by Steve Gibson. The contents
of this page are Copyright (c) 2014 Gibson Research Corporation. SpinRite, ShieldsUP,
NanoProbe, and any other indicated trademarks are registered trademarks of Gibson
|Last Edit: Jun 18, 2010 at 07:47 (1,722.52 days ago)||Viewed 20 times per day|