Cable providers provide terrible DNS service. You may notice a big difference in performance by changing your DNS settings to http://code.google.com/speed/public-dns/ or http://www.opendns.com/. As a bonus, I've noticed that many service interruptions are DNS failures, not routing issues.

Be Careful about changing your DNS settings if you use a CDN - iTunes, in particular, offers horrible performance if you use googles DNS.

Why do you think iTunes would care what DNS server you are using? I'm genuinely curious, as I would think that the DNS lookups are abstracted by the OS, and even if they were not, I cannot see how using a different (esp. a better performing) DNS server would hinder iTunes performance — especially since the original DNS isn't provided by Apple or anything.

CDNs often rely on your DNS info to determine which node is closest to you and, thus, which node should offer the best performance.

Using a central DNS like Google's can do two things:

1) Break locality. Using 8.8.8.8 may cause CDNs to potentially direct you to a node that's nowhere near your actual physical location

2) Aggregate traffic on a node. The more people using 8.8.8.8 for DNS, the more people who may land on the CDN nodes associated with 8.8.8.8's location and thus that node may be more loaded than the one closer to you

[edit] Google even covers this in their Public DNS FAQ: http://code.google.com/speed/public-dns/faq.html#cdn. They offer 31 different locations, while Akamai may offer thousands of nodes (I can't find hard numbers, but they refer to 1,200+ points of presence here http://www.akamai.com/hdonline and 1,000 networks here http://www.akamai.com/html/technology/edgeplatform.html). You're basically constraining yourself to a small subset of at least Akamai's CDN by using Google Public DNS.

That completely slipped my mind. You are, of course, correct.

An alternative suggestion here is to use Comcast's 75.75.75.75 IP address; while it may seem static, it's actually an anycast address, and will resolve to the nearest node.

One thing many people forget when choosing a DNS server is that it's not just the ping time that counts - the time it takes the server to reply to the actual DNS request is what matters.

e.g. while 75.75.75.75 for me (Comcast Chicago) is a 15ms trip and 8.8.8.8 is a 28ms ping, the lookup on the former (dig x.com @dns_ip) can reach 120ms while on Google's DNS it never exceeds ~35ms.

> CDNs often rely on your DNS info to determine which node is closest to you and, thus, which node should offer the best performance.

Why do they do this? Wasn't DNS SUPPOSED to be something that wasn't in any way tied to locality? Are CDN's just using the side effect/fact of life that they happened to fall out that way, because it's easier?

Pretty much, yes. But it's not just easier, it's also simpler. If you do the distribution at the DNS level (which is pretty much as low as you can go), you don't have to deal with the much messier geo-distribution on a higher level.

You might want to try NameBench, which will find the fastest DNS server for you.

http://code.google.com/p/namebench/

Thanks, but I actually do that using both Google's (8.8.8.8), as well as Level 3's (4.2.2.1)

Absolutely - change the DNS _and_ put something like DD-WRT on your router with dnsmasq as the local caching proxy server. Nothing beats that per Steve Gibson's DNS benchmark!

Was this an inside joke?[1]

1. http://allthatiswrong.wordpress.com/2009/10/11/steve-gibson-...

2. http://news.ycombinator.com/item?id=2510614

It wasn't - this is just a DNS benchmark that does DNS resolution in loop and measures how long it took etc. Nothing requiring security expertise.

Besides I don't understand why few people are after Gibson - he is generally on the mark - that's not to say he doesn't make mistakes - but nothing that makes me want to totally ignore him.

Debug the connection with http://netalyzr.icsi.berkeley.edu/

Thank you.

I was wondering why the load times have been ridiculously long.

Also: I've been getting Cox'd as well the past three or so days. Maybe I should complain, or something.

fyi, cox has their own local speed tests which can be informative as to where the hangup might be. The local one to me is:

http://azspeedtest.cox.com/

It's actually somewhat frustrating to see that register phenomenal speeds when the web past their internal routers seems ... buffering ... buffering.