Quick Benchmarks of Google Public DNS v. My Default Comcast Settings: I’m Switching Back
Posted by Kevin Hanson | Posted in Misc, Technology | Posted on 03-12-2009-05-2008
5
Today Google announced Public DNS. The aim is to speed up your connection speeds and security. Lifehacker reported on it earlier today. The idea is to replace the DNS server your router uses (typically provided by default from your ISP) with a “Google Public DNS.” I decided to put their speed tests to work. Read on to see my results.I decided I use one of my favorite quick-and-easy speed tests. I live in San Francisco, CA and for these speed tests, you choose a connection city. I decided to try out Washington, D.C. and Atlanta.
Washington, D.C. with old DNS Settings:
Washington, D.C. with new DNS Settings:
Download speeds are virtually the same, and upload speeds are significantly improved for this test. I ran the same test with Atlanta as the connection city instead, and then I ran two browsing tests. The first browsing test consisted of my opening seven sites at the same time in multiple tabs, starting the stopwatch, and then stopping the watch when they were all fully loaded. I then repeated the same test with a different group of eight sites. I pasted all of my results into the table below.
[TABLE=3]
In my results, it is slightly faster in a few tests, but SIGNIFICANTLY slower in other tests. Perhaps I’ll give this another go some other time, but for now I’m sticking with Comcast’s default DNS settings.
Changing your DNS won’t make a difference to your download/upload speeds, in this case any differences are just natural variances with your connection. All the DNS does is translate the address (google.com for instance), into an IP so the request can go to the appropriate server. What you should be measuring is the response time of the DNS server to your requests.
Interesting – I could run those benchmarks in a week when I get home from business travel, but Google indicated that I should have a faster browser experience. The tabs taking nearly double the time to load isn’t faster! That doesn’t seem like natural variances. What are your thoughts?
Pretty stupid test. DNS doesn’t affect the transfer rates – it does only affect the latency, the time where the page starts to load, and we’re speaking of miliseconds differences here.
@juanjux
Thanks for the feedback. When I get home, I will update with more relevant tests. That being said, Google mentions that browsing should be snappier, and in nearly all cases, it was slower….
Wishing no offense, this test is basically invalid. The reason I say this is that DNS, for lack of a better term, is basically the Yellow Pages for the internet. DNS does nothing more than resolve “www.sitename.com” to an IP address. You internet connection and all the routers in between do all the travel.
A helpful analogy: The bandwidth test performed above is the equivalent to finding a restaurant’s address and seeing how quickly you can drive there. Whether you find the address by looking at the Yellow Pages or by asking a friend will in no way affect the amount of time it takes you to drive from one point to another.
If Google was telling your machine the route to the website, then I could see this as an issue, but that is not what DNS does; DNS tells our machines where the destination is, not how to get there or what router through which to hop.
Additionally, DNS is like learning where a restaurant is in another way: Once you know the address, you will remember it for a bit (Unless your memory is like mine). This is because DNS lookups are cached on client computers. Windows XP caches DNS lookups for 5 minutes by default, then “forgets” it. After 5 minutes, it has to ask where the site is again.