Google recently launched it’s own Public DNS offering, allowing Internet users to make use of their DNS servers for resolving address queries on the Internet. I have been a fan and user of OpenDNS for quite some time now, which offers the same service with many more options (as compared to Google’s new service). Immediately, I was curious as to how Google’s service compares with OpenDNS.
On a functional and service offering standpoint, I don’t need to really talk about anything, because the founder of OpenDNS already has excellent thoughts on this topic. But he doesn’t talk about the most important metric of all, which is performance. I was curious about how the two compare on pure performance terms. So, sitting in India, I decided to run a quick test myself to measure the performance between the two.
The method is the same as I used when I compared OpenDNS with Airtel, my local ISP in India. I basically measured lookup times of 100 random domains and compared the two DNS on that metric. Here are my findings.
The graph below plots the query response time for both the DNS services for 100 random domains. What we can see from the graph already is that neither one is clearly faster than the other.
Here’s another graph which gives clarity on which one is faster more often:
As we can see, they are almost the same. So, how do we know which is better or faster?
I took a look at the average of the measurements for the two services, and this is what I found:
- OpenDNS – 0.38
- Google Public DNS – 0.41
This means that on average, OpenDNS is slightly faster than Google. However, as we can see there are 3 spikes in the Google Data, and one in the OpenDNS data. When I remove those, the average for the two is almost the same (with OpenDNS being marginally slower).
Another thing worth noting however is the Standard Deviation – the measure which tells us how often will the measurement fall close to the average performance. This can even be surmised by the first graph. With the complete measurements, the Standard Deviation of Google was about 3 times that of OpenDNS, and even with the spikes removes, the standard deviation of Google was twice that of OpenDNS. Which means that OpenDNS on the whole is more consistent in its operation.
This means that when Google is deviating on the faster side, it will be much faster than OpenDNS, but when it deviates on the slower side, it will be much slower than OpenDNS (on average).
At this stage, I will have to say that I cannot conclude, though I would like to think that OpenDNS is a better performer for providing consistent performance, as compared to Google Public DNS. But with Google’s muscle, I am sure that it will eventually surpass OpenDNS on these parameters. Who knows, maybe Google will buy OpenDNS (although I am not sure if they will be selling).