Saturday, August 9, 2014

How to improve Internet connection performance by selecting the fastest DNS servers

In a lot of cases the Internet connection performance can be improved by using DNS servers different from the ones supplied by the ISP. How do we know that using Google Public DNS will improve performance? We don’t. Not unless we test it. DNS Benchmark and Namebench are two free utilities that are available to aid us in this task. Both are great troubleshooting tools and are a must have for every network admin’s toolset. This post provides a very general overview of the two products.


DNS Benchmark is free utility from Steve Gibson of Gibson Research Corporation (GRC) fame. This is one of many great utilities that Steve has developed. It’s a tiny 164 KB standalone executable compared to the 5 MB compressed setup for Namebench. On the other side Namebench has already been ported to Linux and MacOS.


DNS Benchmark’s DNS server ranking methodology is slightly different from Namebench’s, as clearly indicated by the results of two consecutive tests using both products. DNS Benchmark recommended Level 3 ( and Speakeasy ( DNS servers. Namebench recommended UltraDNS ( and Genuity/Level 3 (


Namebench benchmarks DNS server responses for your Internet connection utilizing a number of selectable options and generates a detailed report in HTML format. I am including screenshots of the report below to give a better idea of the results. In this particular case UltraDNS returned faster results than Google Public DNS (, which are my current preferred DNS servers.


Granted the above applies mostly to small office/home office (SOHO) Internet connections, as most enterprise environments will have their own DNS server infrastructure in place. Another interesting point to note is that DNS cache plays a huge role in Internet connection performance. Given two identical DNS servers — one with few users querying it and one with many, the one with many will provide quicker responses because most of its DNS lookups will already be cached. This holds true until the number of users or DNS queries gets so large that the server is not able to respond due to network I/O or processor performance bottlenecks.

Post source