PDA

View Full Version : Google Rolls out New DNS Service



kahawai chaser
04-12-2009, 08:30 PM
To help improve stability or speed into the Internet, Google has launched their own public DNS service for a faster, safer, and more reliable internet experience. Hence you can change your DNS settings as described at Google code (http://code.google.com/speed/public-dns/docs/using.html) with more details at Google code blog (http://googlecode.blogspot.com/2009/12/introducing-google-public-dns-new-dns.html).

Apparently for those that own/manage websites/blogs there is talk that a site's loading times (http://searchengineland.com/google-releases-page-speed-report-in-webmaster-tools-31036) may be used in ranking web sites listings in search results next year; Perhaps the need for Google to offer their own DNS...

xyz823
04-12-2009, 08:49 PM
I'm using it now, its seems slightly faster but I havn't done any side to side test against open DNS.

Chilling_Silence
05-12-2009, 10:17 AM
Very cool :D

xyz823
05-12-2009, 10:19 AM
Very cool :D

LOL was waiting for you to post :p

Chilling_Silence
05-12-2009, 01:55 PM
Google or OpenDNS ... Hmm ...

Sweep
05-12-2009, 06:29 PM
Has been tested here:-

http://www.geekzone.co.nz/freitasm/6980

the_bogan
05-12-2009, 06:36 PM
Has been tested here:-

http://www.geekzone.co.nz/freitasm/6980

I appreciate the story, but geez the grammar is rubbish in that article.

I thought that Telstraclear users shouldn't use opendns etc. since Telstraclear caches their stuff differently.

Sweep
05-12-2009, 06:56 PM
I appreciate the story, but geez the grammar is rubbish in that article.

I thought that Telstraclear users shouldn't use opendns etc. since Telstraclear caches their stuff differently.

You can't blame me for the story but I thought it may help.

I'm with Xtra and using OpenDNS. I wasn't planning to try GoogleDNS just yet anyway.

Chilling_Silence
06-12-2009, 09:13 AM
I thought that Telstraclear users shouldn't use opendns etc. since Telstraclear caches their stuff differently.

Same with Telecom, it can completely miss a fair bit of their caching apparently.

Agreed, the article could have been a little more in-depth, no mention of the connection type, no mention of the number of iterations tested, no mention of the time of day tested at, as all can have very different effects on the response times, especially when we're talking about the average difference being 50ms between Telstraclear & Google.

Of course, there's a high chance your own ISPs DNS servers will be fastest (Considering Telstraclear now own Paradise and the old Paradise DNS servers are hardly used, there's the possibility that for that reason they're faster than Telstraclears own main DNS servers), and that's shown for him using his TelstraClear connection.

Just running 3 tests on a Telecom ADSL2+ connection at 8AM in the morning I managed to find the top #1 spot varied between Paradise, Google & Telecom's nameservers. Interesting, next time I thought I should run something a little more complete and include all the international nameservers that DNS Benchmark uses ;)

Naturally, I'm posting my results, and I'd encourage any others to do-so too! Grab DNS Benchmark: http://www.grc.com/dns/benchmark.htm

At approx 8:45AM on a Sunday, Telecom (Big Time) ADSL2+ connection with DNS QoS'd to above everything else on my line (My connection basically unused anyway - I made sure there was no other DNS requests during the benchmark), I ran 3 iterations with the standard DNS Benchmark nameservers + the following added:
8.8.8.8 (Google #1)
8.8.4.4 (Google #2)
202.27.158.40 (Telecom #1)
202.27.156.72 (Telecom #2)
203.96.152.4 (Rachel - Paradise)
203.96.152.12 (Kirsty - Paradise)
203.97.78.43 (Telstraclear #1)
203.97.78.44 (Telstraclear #2)


Test #1 complete results in text form: http://pastie.org/729353
Attached file: Chill-DNS_Benchmark_1.JPG (http://www.imagef1.net.nz/files/Chill-DNS_Benchmark_1.JPG) (99 KB)

Test #1 Summary: My ISP (Telecom) came in 5th in the results. Whilst Google's are nowhere to be found (They're right at the bottom in-fact, search the text results for "8. 8. 8. 8"), Level 3 Communications (Same name as Googles come up as) have come in at #11. Verizon cleaning up the Top 10, with 3 of theirs at the top of the list!


Test #2 complete results in text form: http://pastie.org/729367
Attached file: Chill-DNS_Benchmark_2.JPG (http://www.imagef1.net.nz/files/Chill-DNS_Benchmark_2.JPG) (99 KB)
Test #2 Summary: iiNet placed first this time. TelstraClear came in 10th with their other DNS Server a little further down in the JPG list, and one of Paradise' also. Verizon again with 2 in the Top 10, no sign of Telecom this time though, they're half way down the list. Google again right at the bottom.


Test #3 complete results in text form: http://pastie.org/729373
Attached file: Chill-DNS_Benchmark_3.JPG (http://www.imagef1.net.nz/files/Chill-DNS_Benchmark_3.JPG) (99 KB)
Test #3 Summary: TelstraClear all right at the top, one of Paradise up there also, and a few of Verizons too, though SpeakEasy won this time.


Overall summary: Google's thus far have been consistently slowest, right at the bottom when looking at the cached results.
The time between the fastest and slowest is <25ms for uncached lookup average.
However, if you notice in the last results SpeakEasy won at the top, look at the time taken for a .com lookup and compare that with Telstraclear, Telecom or Google. All are faster.
What does this mean for an end user? You're potentially not going to notice any difference in the resolution speeds when using any of these, they're all approx 190 -> 260ms. The difference between the min, max & average for all of them is so close, I'd suggest that it's not even really worth bothering changing from your ISP's DNS servers unless they're having issues, or slow down under load, which is potentially a good reason to use something such as OpenDNS or Googles (They don't just have two servers, but a whole farm of them).
There are, however, other factors which you may appreciate, such as OpenDNS domain filtering. Interestingly enough even though TelstraClear / Telecom have scored better in the Benchmark, in real-world "feel" tests, I'm certain I'm not alone in saying that Google / OpenDNS "feel" like they're resolving everything and the pages completing loading faster?


Please, post back your results, and also add in your ISP's nameservers if you know them, when doing the test.

fred_fish
06-12-2009, 09:37 AM
Hmmmm, how long before "Sponsored listings" to pay to redirect requests for your competitors sites to your own? :eek:

Seems that resistance is indeed futile.....

jcr1
06-12-2009, 12:26 PM
Here's the "conclusions", for me, from that GRC link (they do good stuff).
I'm just wondering, if what it says, I should be adding in the DNS's of Telecom, my ISP?


System has only ONE (router based) nameserver configured.
It appears that only one local (router gateway) DNS nameserver, with the IP address of [192.168.1.2], is currently providing all DNS name resolution services to this system. This configuration is not recommended because most consumer-grade routers provide inefficient and under-powered DNS resolution services.

Unless the DNS resolvers your router is using is under your control, it may not be providing the best or complete name resolution services. For example, is it using multiple redundant DNS nameservers?

Users of GRC's DNS Spoofability system have determined that consumer-grade routers can be crashed by the receipt of specific DNS reply packets from the Internet. This opens the possibility that Internet-based criminals could acquire access to your router from the Internet as well as to the private network in controls.

Many consumer-grade routers fail to provide the full range of DNS lookup services. This may have been detected by the benchmark and noted below.

Recommended Actions:

Unless you have some specific reason not to, you should give serious thought to disabling your router's provisioning of DNS services (which it is providing for all computers on your local network). After this is done, a fresh reboot of your computers will likely reveal the multiple DNS nameservers provided by your ISP. This is a superior configuration, without an under-powered router acting as a incompetent middleman and impeding all DNS access.

Note that if you can determine the IP addresses of your ISP-provided nameservers (which may be visible in your router's web configuration) you could manually add them to the nameservers being tested by this benchmark, while also leaving your router providing DNS. This would allow you to compare the performance when running through your router versus "going direct".


System's sole nameserver is alive and replying to queries.
Although this system has only one DNS resolving nameserver, at least it is alive and replying to DNS queries. (If it were not, you would likely be painfully aware, since it would be difficult to accomplish anything requiring Internet access.)


System nameserver is SLOWER than 14 public alternatives!
This benchmark found 14 publicly available DNS nameservers that are reliably faster than the slowest nameserver currently being used by this system. If you were to adjust your system's configuration to use the faster of these nameservers instead of what it is currently using, your DNS lookup performance, and all use of the Internet, would be improved.

Recommended Actions:

With at least 95% certainty: Based upon a statistical analysis of the spread in timing value samples received during the benchmark, there is at least a 95% certainty that the performance conclusions stated above are correct. But even so, since changing DNS nameservers requires thought and effort, it's something you want to be sure about. Therefore, since these results represent a single snapshot in time, you may wish to confirm that the faster alternative nameservers are consistently faster than your system's currently configured nameservers, and that those public alternatives don't have any negative characteristics such as being colored orange to signify that they redirect mistaken URLs to an advertising-laden search page rather than returning an error (which will be a concern to some users).

You may also wish to check the relative performance at different times of day to make sure that the performance improvement over your system's current nameservers is reliable throughout the day.

And you may wish to make sure that the alternative nameservers are enough faster than what you are currently using for the improvement to be worth changing away from what you're currently using. (This test is only saying that it's 95% sure they are any amount faster.)


This system's nameserver is 100% reliable.
DNS reliability is extremely important, since lookup requests that are dropped and ignored by nameservers cause significant delays in Internet access while the querying system waits for a reply. The system is then finally forced to reissue the query to the same or to backup nameservers. While your system is patiently waiting for a reply, you are impatiently waiting to get on with your Internet access.

During this benchmark test, the system's nameserver tested returned a reply for every request sent. It doesn't get any better than that. Very nice.


This system's nameserver intercepts name errors.
One or more of this system's nameservers intercepts errors and redirects web browsers to a custom page in response to an invalid DNS lookup request. (This is shown with an orange coloring of the nameserver IP address and descriptive text on the benchmark's "Nameserver" page.) This behavior is typically used as a marketing maneuver to redirect mistaken web browser URL entries to the DNS provider's own advertising-laden marketing-related pages. The major ISPs Earthlink, Roadrunner and Comcast are known to be doing this. While this may be regarded as a useful service by some users, others object to the idea of not receiving an error in response to an erroneous request. Some free DNS server providers, such as OpenDNS, allow this behavior to be customized so that erroneous queries can be configured to return an error. Many responsible ISPs are also offering "opt-out" options to prevent advertising interceptions.

Recommended Actions:

If you feel that this marketing-driven behavior is unacceptable from a DNS nameserver, you may be able to configure the service to return errors. Otherwise, you are free to switch to any alternative high performance and high reliability nameservers that are properly returning errors in response to erroneous queries.

If you choose to configure the existing nameserver(s) to return errors, you can use this benchmark utility, at any time, to easily verify that the DNS behavior is what you expect and desire.


System nameserver is replying to all query types.
During the development of this DNS Benchmark we discovered that the routers used by some pre-release testers were not returning results for the benchmark's Uncached and/or Dotcom testing queries. Even though these queries are admittedly unusual, they are completely valid. So the only conclusion was that those few routers were inherently defective. The good news here is that your nameserver is replying to these unusual but valid queries.


__________________________________________________ __________________

kahawai chaser
06-12-2009, 09:15 PM
Same with Telecom, it can completely miss a fair bit of their caching apparently.

Agreed, the article could have been a little more in-depth, no mention of the connection type, no mention of the number of iterations tested, no mention of the time of day tested at, as all can have very different effects on the response times, especially when we're talking about the average difference being 50ms between Telstraclear & Google.

Of course, there's a high chance your own ISPs DNS servers will be fastest (Considering Telstraclear now own Paradise and the old Paradise DNS servers are hardly used, there's the possibility that for that reason they're faster than Telstraclears own main DNS servers), and that's shown for him using his TelstraClear connection.

Just running 3 tests on a Telecom ADSL2+ connection at 8AM in the morning I managed to find the top #1 spot varied between Paradise, Google & Telecom's nameservers. Interesting, next time I thought I should run something a little more complete and include all the international nameservers that DNS Benchmark uses ;)

Naturally, I'm posting my results, and I'd encourage any others to do-so too! Grab DNS Benchmark: http://www.grc.com/dns/benchmark.htm

At approx 8:45AM on a Sunday, Telecom (Big Time) ADSL2+ connection with DNS QoS'd to above everything else on my line (My connection basically unused anyway - I made sure there was no other DNS requests during the benchmark), I ran 3 iterations with the standard DNS Benchmark nameservers + the following added:
8.8.8.8 (Google #1)
8.8.4.4 (Google #2)
202.27.158.40 (Telecom #1)
202.27.156.72 (Telecom #2)
203.96.152.4 (Rachel - Paradise)
203.96.152.12 (Kirsty - Paradise)
203.97.78.43 (Telstraclear #1)
203.97.78.44 (Telstraclear #2)


Test #1 complete results in text form: http://pastie.org/729353
Attached file: Chill-DNS_Benchmark_1.JPG (http://www.imagef1.net.nz/files/Chill-DNS_Benchmark_1.JPG) (99 KB)

Test #1 Summary: My ISP (Telecom) came in 5th in the results. Whilst Google's are nowhere to be found (They're right at the bottom in-fact, search the text results for "8. 8. 8. 8"), Level 3 Communications (Same name as Googles come up as) have come in at #11. Verizon cleaning up the Top 10, with 3 of theirs at the top of the list!


Test #2 complete results in text form: http://pastie.org/729367
Attached file: Chill-DNS_Benchmark_2.JPG (http://www.imagef1.net.nz/files/Chill-DNS_Benchmark_2.JPG) (99 KB)
Test #2 Summary: iiNet placed first this time. TelstraClear came in 10th with their other DNS Server a little further down in the JPG list, and one of Paradise' also. Verizon again with 2 in the Top 10, no sign of Telecom this time though, they're half way down the list. Google again right at the bottom.


Test #3 complete results in text form: http://pastie.org/729373
Attached file: Chill-DNS_Benchmark_3.JPG (http://www.imagef1.net.nz/files/Chill-DNS_Benchmark_3.JPG) (99 KB)
Test #3 Summary: TelstraClear all right at the top, one of Paradise up there also, and a few of Verizons too, though SpeakEasy won this time.


Overall summary: Google's thus far have been consistently slowest, right at the bottom when looking at the cached results.
The time between the fastest and slowest is <25ms for uncached lookup average.
However, if you notice in the last results SpeakEasy won at the top, look at the time taken for a .com lookup and compare that with Telstraclear, Telecom or Google. All are faster.
What does this mean for an end user? You're potentially not going to notice any difference in the resolution speeds when using any of these, they're all approx 190 -> 260ms. The difference between the min, max & average for all of them is so close, I'd suggest that it's not even really worth bothering changing from your ISP's DNS servers unless they're having issues, or slow down under load, which is potentially a good reason to use something such as OpenDNS or Googles (They don't just have two servers, but a whole farm of them).
There are, however, other factors which you may appreciate, such as OpenDNS domain filtering. Interestingly enough even though TelstraClear / Telecom have scored better in the Benchmark, in real-world "feel" tests, I'm certain I'm not alone in saying that Google / OpenDNS "feel" like they're resolving everything and the pages completing loading faster?


Please, post back your results, and also add in your ISP's nameservers if you know them, when doing the test.

Thanks for your tests Chilling. I had a good read of the results; Really as you note, their is no significant difference among the DNS results. Guess I won't change yet...

But why are the std deviation results high relative to the mean? They all calculate to quite high at about 38 to 40% coefficient of variation (% cov), where I think ideal % cov's are 5 % or less.
Would results be different form other locations globally - say in the USA - presumably closer to Google's main location H/O?

I note some expressed concern at Google code blog about privacy/reliability - are their tests/analytics to assess such factors?

Chilling_Silence
06-12-2009, 10:19 PM
I've switched to Telecoms DNS servers to see if that fixes YouTube streaming on Big Time, and I notice an almost instant "feel" in the load-speed of pages. So does my wife, and I didn't tell her, she just said "Why do pages seem to take longer to open tonight?". She didn't say that for Googles which we've been on for 24 horus.

Unsure about the tests to be honest?

Chilling_Silence
06-12-2009, 11:13 PM
OK some more test @ 11PM Sunday 6/12/2009
First off using Googles:
http://pastie.org/730151
Attached file: 11pm_google_1.jpg (http://www.imagef1.net.nz/files/11pm_google_1.jpg) (139 KB)
Attached file: 11pm_google_2.jpg (http://www.imagef1.net.nz/files/11pm_google_2.jpg) (87 KB)

Then using Telecom:
http://pastie.org/730153
Attached file: 11pm_telecom_1.jpg (http://www.imagef1.net.nz/files/11pm_telecom_1.jpg) (132 KB)
Attached file: 11pm_telecom_2.jpg (http://www.imagef1.net.nz/files/11pm_telecom_2.jpg) (114 KB)

Interesting yes? :D

Speedy Gonzales
06-12-2009, 11:21 PM
And (http://www.pcworld.com/article/183671/google_public_dns_and_your_privacy.html). It lagged a bit here, when I tried it. So no thanks

Chilling_Silence
07-12-2009, 06:20 AM
I fail to see what's wrong with it logging that kind of information, no doubt OpenDNS would also be doing likewise:
http://www.opendns.com/privacy/

Also, if you read the blog post from the founder of OpenDNS, it seems generally supportive, at least 3 of the 5 (2,3 & 4) appear to be quite positive:
http://blog.opendns.com/2009/12/03/opendns-google-dns/

Naturally I understand it's again not for the masses, but having it available is a good thing :)

gary67
07-12-2009, 06:37 AM
It's odd using Ihug's DNS and having now used our data cap up so we should be throttled back until next weekend yet my browser is loading like a rocket this morning really odd

Chilling_Silence
07-12-2009, 07:41 AM
What does speedtest.net say about your throughput?