[AusNOG] Google launches alternative DNS resolver

[Scott Howard]

On Thu, Dec 3, 2009 at 2:40 PM, Matthew Moyle-Croft <mmc at internode.com.au>wrote:

>  I think it's a great idea if you're North America centric.
>
> But, it'll add 150ms or more to most customers and it'll hurt the locality
> based services already in Australia.   So, performance in Australia will
> suck.  eg.  Akamai becomes 300ms+ vs 25-50ms depending on where you are.
> Think how much traffic that means bringing in from O/S?
>

It's anycast, and it would seem logical that they will roll it out to more
datacenters over time.

>From both the US and Europe I'm seeing a maximum of about 20ms to get to the
nearest 8.8.8.8 server, and less than 50ms from much of Asia, so it's
neither centralised, nor North American centric.  Australia seems to be an
ugly cousin at the moment...

Presuming the IP addresses they are using to originate the requests back out
are correctly listed in the various GeoIP databases then the likes of Akamai
shouldn't suffer at all.

>From Frankfurt :
64 bytes from 8.8.8.8: icmp_seq=0 ttl=249 time=5.440 ms

>From London :
64 bytes from 8.8.8.8: icmp_seq=0 ttl=249 time=13.734 ms

>From Washington DC :
64 bytes from 8.8.8.8: icmp_seq=0 ttl=249 time=5.487 ms

>From Singaopre :
64 bytes from 8.8.8.8: icmp_seq=0 ttl=246 time=51.271 ms
(Hits Google IP space in Singapore, but the actual resolution is elsewhere)

>From Tokyo :
64 bytes from 8.8.8.8: icmp_seq=0 ttl=246 time=39.430 ms

>From Adelaide :
64 bytes from 8.8.8.8: icmp_seq=0 ttl=244 time=164.262 ms
(Hits Google IP space in Sydney)


  Scott
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ausnog.net/pipermail/ausnog/attachments/20091203/7117ec6c/attachment.html>