Google Public DNS benchmarking

Yesterday, Google announced a public DNS service that promises to speed up your browsing experience.

We here at x7 Labs like our interwebs speedy, so I decided to take a quick look at the service. For testing, I used my home 110 Mbit/s cable modem connection from Welho.

As background information, I wanted to know how long it took for packets to travel to Google data centers compare to my ISP.

TCP/IP packets travelledĀ several additional hops compared to the local ISPĀ  case. First they were beamed to Sweden, adding 10-20 ms to the roundtrip time. From there, they were routed to an unknown datacenter, possibly in UK or continental Europe.

All in all, the roundtrip time for Google DNS wasn’t bad, around 50 ms (see below), compared to 8-10 ms for my local ISP (not shown):

UDP roundtrip test

UDP roundtrip test

A side note: The media was in a frenzy when it was announced that Sweden is allowed to monitor all network traffic flowing through their backbones. At least until Google opens a data center in Finland (which is under construction), the Swedes can easily log DNS queries made by Finnish internet users through Google.

Next, I used the DNS Benchmarking tool developed by Steve Gibson. It does exhaustive tests to find out how long it takes for the DNS servers to return both cached and uncached records. I compared my local ISP, OpenDNS and Google:

DNS Benchmark by GRC

DNS Benchmark by GRC

My local ISP was twice as fast (~100 ms vs. ~200 ms) when serving uncached records, and around 5-6 times faster for cached DNS records (8-9 ms vs. 50-60 ms).

Next, I wanted to see if changing the system-wide DNS setting affects the browsing speed. The test case was very simple: reset all caches, open yahoo.com and record time spent in DNS. For this test, I used the excellent Fiddler HTTP Debugging Proxy.

Using my ISP’s DNS, name resolution time was around 300 milliseconds. Here’s a typical result:

Time spent in DNS (local ISP)

Time spent in DNS (local ISP)

When I changed system settings to Google’s Public DNS, I noticed a difference:

Time sent in DNS (Google)

Time spent in DNS (Google)

Using the Google DNS Public Service increased name resolution time almost by a factor of 10x during actual browsing.

Clearly, Google Public DNS isn’t competitive in performance at the moment.

2 comments

  1. IpProvider

    Nice article. It is nice to hear, that Google is not the best provider is this planet as the thought.