Forum Moderators: phranque
[neustar.biz...]
Yes, there's something to it, as many of the largest web sites use them. Most of us don't have the need for this level of service, though.
When you say "we have one server", if you mean "DNS server", your current configuration is probably deficient. If you simply mean "one web server", and you're using your registrar's DNS servers, your DNS service is probably adequate.
The key feature that this type of 3rd party DNS can offer is a large number of redundant DNS servers around the world using IP-anycast technology. If you search for "IP-anycast DNS", you can find other 3rd-party DNS providers providing similar service at a MUCH lower price. No, they don't have as many servers in as many places, but the price difference is huge and, IMO, they come awfully close. Certainly close enough for the average website.
I actually mean one webserver only running a basic website with the host handling DNS level issues. In my last job we had load balancers and multi-domains and multi-servers and even then I doubt this would be needed.
I just wanted some sort of come back on this guy. As the conversation currently goes like this.
Salesmen: "You could be losing lots of traffic using BIND. Amazon lost 15% and we fixed it for them."
Me: "Hmm, we just have one server with a website and everything is running fine. I've not come across any problems."
Salesman: "No, no. You don't know how much your losing. Your hosts won't tell you."
Me: "I've never come across any problems. I'll chat with my networking expert friends see what they say. Thanks"
Salesman: "We handle the $7million infrastructure for Forbes. It's all about ROI"
Me: "Ok, we're not amazon, we have 1 site which is moderately busy. I'll look at the details and get back to you. Thanks."
Salesman: "We'll run a test, I'll send you the results tomorrow"
On and on...
I guess my problem is I don't like to be rude and I didn't want to give the wrong impression of my phone manner to my boss. Anyway...thanks!
Don't really just want to hang up on the guy but I probably will. haha
We'll run a test, I'll send you the results tomorrow
The test might actually be useful. If they give you enough information, it might be worthwhile letting them run the test. Then you can tell him to go away. :)
Here is my guess as to how the test will go...
He will tell you that studies have shown that the average user will only wait n mSec for a page to load before abandonding a site. They will test the total time to load your home page from multiple locations around the world. This will include both the time to look up your site using DNS plus the time to load your homepage (or some selection of your pages). Then, using their own average statistics, they will show you how much faster the load times would be using their DNS.
There's a bit of trickery here: if your website itself is slow, or you have a large home page that takes some time to load, you are going to have some lost users if you apply the abandonment averages. If their DNS is faster than yours (and it almost certainly is), then using their DNS will save some of the edge cases. That is, using their DNS would make the total time small enough so that some percentage of the "lost" loads would then come below the average abandonment time.
The most advantageous scenario for them would be if your load times are bordering right on the edge of acceptable load time. In that case, they can show you that they could save nearly all of your abandonments.
If they are willing to give you the breakdown of the load times, this is certainly useful information for you to have, in any case.
Their tests showed that 0.002% of dns requests were too slow.
