Namebench is a program that analyses DNS lookups to see if your DNS settings are optimal. My results are here. They recommend that I use my ISP’s DNS server, but they also show the main reason I stopped using my ISP’s DNS server – that innocuous “NXDOMAIN Hijacking” notation beside the entry for that DNS server means that if you mistype a domain name, it takes you to your ISP’s search page instead of having your browser tell you that you mistyped a domain name. I HATE that, “with the power of a thousand fiery suns” as Vicki would put it, because it breaks things, usually in ways too subtle for ordinary users to notice. I run a DNS server on my Linux box (on 192.168.1.2) because it won’t do “NXDOMAIN Hijacking”, and also because I believed it would be faster. One other reason for running my own DNS server is so I could reach computers on my home netwrok via a system name rather than via an IP, something that was probably more important when I had multiple Linux boxes that I needed to be able to ssh into than now, when I basically only connect to my Linux box and (more rarely) into my MacBook Pro.
If you look down the page to the graph “Response Distribution Chart”, it shows that for the first 30% of the responses, my home DNS server is *way* faster than the competition – I guess that means that things that it’s already seen and cached, it returns at the speed of the local network. But the graph trails off pretty quickly, and by the time you reach 50% of the responses, it’s slower than most of the other ones – I don’t know why it would be slower than “Internal 192-1-1”, which is the DNS cache on my router, but I suspect that’s because the router will just ask my ISP’s DNS server when it doesn’t know something rather than reaching out to the broader internet.
What I should do now, I think, should be to set up a DNS server on my colo box and see how it compares.