Does Using OpenDNS or Google DNS Slow Down iTunes Downloads?

[Jason Dunn]

<p><img src="http://images.thoughtsmedia.com//at/auto/1307241610.usr1.png" /></p><p>When I tweeted the above comment - I was quite appalled at how slow iTunes was at delivering content - someone pointed me to <a href="http://daringfireball.net/linked/2010/12/29/itunes-store-dns" target="_blank">this post on Daring Fireball</a> that says the fix is to not use Google DNS. There are <a href="http://forums.opendns.com/comments.php?DiscussionID=8541" target="_blank">similar complaints in the OpenDNS forums</a>. The tweeter speculated that if I was using OpenDNS - and I was - it would have the same impact. I downloaded <a href="http://itunes.apple.com/us/app/tapper-world-tour-hd/id421477146?mt=8" target="_blank">Tapper World Tour HD</a> from the iTunes store and it took 3 minutes 36 seconds to grab the 193 MB file. That's 890 KB per second on the download, and my connection is capable of just over 3000 KB/s. If there was one thing I'd expect/hope to deliver bandwidth-saturating results, it would be a content deliver network such as iTunes.</p><p>I went into my router, set the DNS back to my ISP (Shaw), and rebooted the router. On my computer with iTunes, I did an ipconfig /flushdns, along with a /release and a /renew just to make sure everything was reset. I verified the computer wasn't using OpenDNS, then I started up iTunes and re-downloaded the same file. It downloaded in 3 minutes 2 seconds. A slight improvement, but not much and frankly within the margin of error for a test like this.</p><p>I'd like to ask the Apple Thoughts community if they've seen similar results with OpenDNS or Google DNS, and if stopping using those services resulted in a significant speed boost on downloads. Also, if my methodology was flawed in some way, I'm happy to learn how to better test something like this.</p>

[Vincent Ferrari]

Depends on the CDN. Google and OpenDNS do something with the DNS that essentially tells a CDN that all traffic is from the same IP. Akamai figured out a way around it, but I don't know what other ones did.

Now for the real world: I use OpenDNS anyway and barely notice a difference, but I have 30/30 and imagine it's so much bandwidth I just won't see it anyway. I don't know about others and I have heard that in the past there are issues with Google and OpenDNS and the way they handle things, so you wouldn't be the first.

[Jon Childs]

I have to admit I know nothing about the iTunes download protocol, so I am speaking out of complete ignorance. That said, why would the DNS make any difference? How many name lookups does it do to download what I would assume is one file? Does apple break down large files into lots and lots of little ones? I have used openDNS in the past and it seemed to help a little with webpages that grabbed content from various different place, but I can't imagine why it would help downloading one big file. Of course it could just be my lack of imagination.

[Vincent Ferrari]

I have to admit I know nothing about the iTunes download protocol, so I am speaking out of complete ignorance. That said, why would the DNS make any difference? How many name lookups does it do to download what I would assume is one file? Does apple break down large files into lots and lots of little ones? I have used openDNS in the past and it seemed to help a little with webpages that grabbed content from various different place, but I can't imagine why it would help downloading one big file. Of course it could just be my lack of imagination.

Akamai uses the DNS requester to route your traffic from the closest location. If your DNS server isn't near your location, the downloads can slow quite a bit. This is the same for Google DNS and OpenDNS on a few different CDN's. I heard when this first surfaced a few months ago that Akamai had figured out a workaround and I don't know if others have as well, but that's the situation as it stands as far as I know.

[Jason Dunn]

Bit of a follow up: I quadrupled my downstream bandwidth (moving from 25mbps to 100mbps) and tried my test again: same result. So it's definitely not a bandwidth limitation on my end... :(