[dns-operations] What are the data?

[East Chao]

We've done some limited testing. It is as you'd expect. They have good 
locality information for their deployment footprint. Places where they 
don't have good/similar deployments have bad/not-so-bad results. It's 
pretty good in the US. It's less good in Europe and even less good in 
the Asia/AU region.

That's not surprising as we don't deploy our networks lock step w/ 
google. Depending on where the client lives they will get marginally 
faster to slower response times by using Google DNS. In some places like 
Australia users would incur a pretty hefty hit since they're being 
redirected from local servers to servers over 100ms away.

Overall, I think a lot of the things they've done w/ the product are a 
"good idea" (pre-fetching/caching/maintaining state independent of 
query) but I think the locality loss probably doesn't compensate for 
it... At least for internet presences with large deployments worldwide. 
Averaging all Akamai users worldwide, for instance, would probably get 
worse performance overall because google just doesn't know where all the 
best Akamai nodes are.  We had the same problem with Akamai trying to 
GSLB internal Yahoo stuff. We can do a better job since we know our 
network whereas they had to guess at it from what they could glean via 
external probing from limited locations.

It's not worse (and maybe better) than any other open dns service. I'm 
not sure I'd try to make the claim it would speed up your web browsing 
experience though. It may speed up your DNS lookup experience but it can 
definitely slow down your web browsing experience.

east

Florian Weimer wrote:
> * Patrick W. Gilmore:
>
>   
>> Data exists.  Maybe even someone here is using a CDN and can provide
>> data.  Then we can make an informed decision.  Until then, posturing
>> and name calling is ... silly. :)
>>     
>
> Debian uses DNS tricks to distribute mirror load.  It proved to be
> necessary because we can't be too picky about mirror offers, and the
> network operators tend to be not very upfront about choke points.  A
> solution based on redirects would likely confuse unofficial mirrors.
> (Our own client doesn't support them, but that could be fixed.)
> _______________________________________________
> dns-operations mailing list
> dns-operations at lists.dns-oarc.net
> https://lists.dns-oarc.net/mailman/listinfo/dns-operations
>
>   


-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.dns-oarc.net/pipermail/dns-operations/attachments/20091209/081bb272/attachment.html>