Google’s New DNS Service Has One Major Flaw, Poor Performance

Google-logo Last week, when Google announced on their blog that they were getting into the DNS business, many people on the web wanted to debate whether or not it would give Google too much control over the Internet, by questioning how Google would use the data that's collected. While I don't disagree that it is a valid concern, since no one is being forced to use Google's DNS offering, to me, that's not the real story.

What we should be discussing as an industry is the performance of Google's Public DNS service, something I haven't seen much written about. I've been using Google's DNS over the past week and at least for me, the performance has been worse than Level 3's DNS or my local ISP, Verizon. While Google is not going into details on where their DNS servers are located, other networking companies in the industry gave me a list of their locations which are; Atlanta, Reston, Seattle, California, Brazil, Taiwan, Germany, Netherlands, Ireland and London.

With that much coverage, you'd think their service would be at least up to par with the others, but in most cases, I'm getting results where Google is 30% slower than competitors. On the Google blog, they say the reason the industry needs their Google Public DNS service is, "to make users' web-surfing experiences faster, safer and more reliable." While that sounds nice, frankly no one is buying it and so far, the results I am seeing don't back it up. Clearly Google is looking at DNS services from the business side and knows that many companies already gladly pay for these services on the open market.

But the real question is, just what does Google think they can accelerate about this, anyway? Google's claimed interest in this is to "speed up the web", but are ISP DNS proxies really the weak link in the whole process? I don't see how they can remove that much latency from the process for a large ISP, like Comcast or Verizon, who not only more than likely has a sophisticated DNS proxy infrastructure of their own, but who also has a large user population. This means that the vast majority of DNS queries they get from users are handled via cached results from a previous user query, so no benefit would be achieved by "pre-caching" DNS responses in the vast majority of cases.

Also, ISP DNS proxies are inside the ISP network, whereas Google DNS proxies have to be reached via the Internet. So if the only speed benefit is that Google will execute code faster on its servers than an ISP will, it seems like all, or at least part of that advantage gets offset by latency associated with Google being more hops away. In the end, if this ends up providing only a very modest performance improvement for only a small percentage of queries and only when the users are on small ISPs, I guess I just don't get what the big gain to the web is supposed to be.

Of course, Google clearly knows all of this and on their website, if you read through a lot of their text you'll find on the bottom of one page a notation that says, "Note, however, that because name servers geolocate according to the resolver's IP address rather than the user's, Google Public DNS has the same limitations as other open DNS services: that is, the server to which a user is referred might be farther away than one to which a local DNS provider would have referred. This could cause a slower browsing experience for certain sites."

Since most users connect to the Internet using DHCP and are automatically assigned name servers, I don't expect the Google Public DNS service to take off. Most consumers are not technical enough to want to change their networking settings, nor do they care. But if that did happen, it seems to me that the biggest company this could impact is Akamai. A big part of Akamai's sales pitch and the reason they say their network is better, is due to them having so many servers located inside user access networks. But for any user who does not use their ISP's DNS proxy, those servers are simply out-of-play, because Akamai's DNS system won't know, at DNS resolution time, that a given user is inside a particular network. As a result, Akamai would have to resolve all Google-based DNS requests to servers at peering points, not servers inside ISPs. I wanted to get Akamai's take on this, but they didn't respond to my request for a comment.

If Google wants to convince us that the reason for their DNS service is to provide a faster web experience and not to collect more data on users, its pretty hard to take them seriously when the performance is not there, as least not for me. The bottom line is that Google is looking at this as a business. Now that Google has entered the DNS space, it should be a wake up call to ISPs who still are not offering DNS solutions and those ISPs that are doing redirection, but don't do it well. Google has now become a competitor to them for DNS services overnight.