Comments on: Google Public DNS http://gpoul.strain.at/2009/12/06/google-public-dns/ The postings on this site are my own and don’t necessarily represent IBM’s positions, strategies, or opinions. Sun, 22 Jan 2012 12:47:25 +0000 hourly 1 http://wordpress.org/?v=3.3.1 By: wnagele http://gpoul.strain.at/2009/12/06/google-public-dns/comment-page-1/#comment-1303 wnagele Mon, 07 Dec 2009 18:54:41 +0000 http://gpoul.strain.at/?p=842#comment-1303 jup ... but that's all about definition of frequently and the user base of the cache jup … but that’s all about definition of frequently and the user base of the cache

]]> By: gpoul http://gpoul.strain.at/2009/12/06/google-public-dns/comment-page-1/#comment-1302 gpoul Mon, 07 Dec 2009 17:50:49 +0000 http://gpoul.strain.at/?p=842#comment-1302 @wnagele: But isn't there a high probability that a frequently requested site will be in the DNS cache of my ISP anyway? :-) @wnagele: But isn’t there a high probability that a frequently requested site will be in the DNS cache of my ISP anyway? :-)

]]> By: wnagele http://gpoul.strain.at/2009/12/06/google-public-dns/comment-page-1/#comment-1293 wnagele Mon, 07 Dec 2009 00:15:53 +0000 http://gpoul.strain.at/?p=842#comment-1293 This was probably exactly the wrong test. The system (reading the FAQ) is aiming to especially speed up frequently requested sites by re-caching them automatically without any client needing to request for them. That's all nothing new really and is only in the media because G is doing it. But requesting something not cached will most likely show worse results then your local resolver because of the added round-trip between the G-resolvers and you. That's however something you can scale pretty well using Anycast these days, but i guess that's something for later. Anyway, i guess you'd be surprised how many ISPs out there have a fucked up resolver running for their customers. And i guess that is the clients that this service is aiming at, besides the redirected customers of some major ISPs. :) This was probably exactly the wrong test. The system (reading the FAQ) is aiming to especially speed up frequently requested sites by re-caching them automatically without any client needing to request for them. That’s all nothing new really and is only in the media because G is doing it. But requesting something not cached will most likely show worse results then your local resolver because of the added round-trip between the G-resolvers and you. That’s however something you can scale pretty well using Anycast these days, but i guess that’s something for later.

Anyway, i guess you’d be surprised how many ISPs out there have a fucked up resolver running for their customers. And i guess that is the clients that this service is aiming at, besides the redirected customers of some major ISPs. :)

]]> By: gpoul http://gpoul.strain.at/2009/12/06/google-public-dns/comment-page-1/#comment-1291 gpoul Sun, 06 Dec 2009 20:26:56 +0000 http://gpoul.strain.at/?p=842#comment-1291 @wnagele: I did some unscientific tests with dig and for me it's not faster. I used my local NAT DNS resolver that is set by default on my local network and compared it with response times from 8.8.8.8 with domains I usually don't lookup (so there won't be any cache hits). Lookups with cache hits were faster than cache misses, but 8.8.8.8 was always slower than my local NAT DNS resolver. Don't ask me how this local NAT DNS resolver works inside of my router and how the caches will actually behave. @wnagele: I did some unscientific tests with dig and for me it’s not faster. I used my local NAT DNS resolver that is set by default on my local network and compared it with response times from 8.8.8.8 with domains I usually don’t lookup (so there won’t be any cache hits).

Lookups with cache hits were faster than cache misses, but 8.8.8.8 was always slower than my local NAT DNS resolver. Don’t ask me how this local NAT DNS resolver works inside of my router and how the caches will actually behave.

]]> By: wnagele http://gpoul.strain.at/2009/12/06/google-public-dns/comment-page-1/#comment-1288 wnagele Sun, 06 Dec 2009 16:10:31 +0000 http://gpoul.strain.at/?p=842#comment-1288 It's quite simple - officially you try to improve the response time by reducing the time for lookups. Under the table this is also fighting the increased DNS redirect mechanisms implemented by large ISPs. Controversial, to say the least. It’s quite simple – officially you try to improve the response time by reducing the time for lookups. Under the table this is also fighting the increased DNS redirect mechanisms implemented by large ISPs. Controversial, to say the least.

]]>