DNSBench "stable" released

http://www.grc.com/dns/benchmark.htm

Test your DNS servers and choose the fastest

[]'s
GRC's DNS Benchmark performs a detailed analysis and comparison of the
operational performance and reliability of any set of up to 200 DNS
nameservers (sometimes also called resolvers) at once. When the
Benchmark is started in its default configuration, it identifies all
DNS nameservers the user's system is currently configured to use and
adds them to its built-in list of publicly available “alternative”
nameservers. Each DNS nameserver in the benchmark list is carefully
“characterized” to determine its suitability — to you — for your use
as a DNS resolver. This characterization includes testing each
nameserver for its “redirection” behavior: whether it returns an error
for a bad domain request, or redirects a user's web browser to a
commercial marketing-oriented page. While such behavior may be
acceptable to some users, others may find this objectionable.

In article <(E-Mail Removed)>, (E-Mail Removed)
says...
>
> http://www.grc.com/dns/benchmark.htm
>
It is a good tool, proving my ISP's DNS servers
have superior response ( especially for cached results )
to any other free DNS services.

--
Amateurs have built Noah's Arc and Linux,
professionals Titanic and Windows.

Shadow wrote:

> http://www.grc.com/dns/benchmark.htm
>
> Test your DNS servers and choose the fastest
>
> []'s
> GRC's DNS Benchmark performs a detailed analysis and comparison of the
> operational performance and reliability of any set of up to 200 DNS
> nameservers (sometimes also called resolvers) at once. When the
> Benchmark is started in its default configuration, it identifies all
> DNS nameservers the user's system is currently configured to use and
> adds them to its built-in list of publicly available “alternative”
> nameservers. Each DNS nameserver in the benchmark list is carefully
> “characterized” to determine its suitability — to you — for your use
> as a DNS resolver. This characterization includes testing each
> nameserver for its “redirection” behavior: whether it returns an error
> for a bad domain request, or redirects a user's web browser to a
> commercial marketing-oriented page. While such behavior may be
> acceptable to some users, others may find this objectionable.

Also of note is that it does not install. You simply run the downloaded
..exe file rather than have to "install" it into Windows. So download
and run. Since it is self-contained, it's portable, too.

What it doesn't do is show statistical measurements for many
measurements made over a long time. It'll show which DNS server is
faster than another but that is just for that particular test. The next
test might show a different DNS server is faster. Also, some of the
"Conclusions" it lists are misleading or wrong. For example, it'll
complain if you point your host's TCP settings at your router because
you've listed only 1 DNS server (the router's) whereas the router
probably can have 2 or 3 DNS servers listed for it. The router really
doesn't have a DNS server but instead will immediately fail the lookup
and pass the request to its configured upstream DNS servers.

Oddly, the benchmark for my router (configured to use OpenDNS' DNS
servers) is faster than going direct to OpenDNS' DNS servers. My ISP's
DNS servers are faster than OpenDNS. So I configure my TCP settings to
use my ISP's best DNS server, my router's DNS server (which then goes to
OpenDNS), and my ISP's 2nd best DNS server as those are what the
benchmarks return for performance. If my ISP's DNS servers are down
then my TCP settings have me use the router's DNS server which goes to
OpenDNS. I used to have DynDNS' DNS servers included in the mix in my
TCP settings for DNS servers list but gave up because they miscategorize
too many sites as spyware sites which results in getting blocked to
those sites (if you happen to end up using their DNS server). They even
had Verisign miscategorized.

In article <i8ak70$2ka$(E-Mail Removed)>, (E-Mail Removed) says...
>
>
> Oddly, the benchmark for my router (configured to use OpenDNS' DNS
> servers) is faster than going direct to OpenDNS' DNS servers. My ISP's
> DNS servers are faster than OpenDNS. So I configure my TCP settings to
> use my ISP's best DNS server, my router's DNS server (which then goes to
> OpenDNS), and my ISP's 2nd best DNS server as those are what the
> benchmarks return for performance.

I have also realized accessing my ISP DNS servers directly
gives better results than accessing them indirectly
through my router, set to act as DNS proxy.

It was also adviced against using router DNS proxy abilities,
as far as it was said to have inferior quality of DNS operations.

But I am not expert in DNS.

--
Poutnik

Amateurs have built Noah's Arc and Linux,
professionals Titanic and Windows.

poutnik wrote:

> In article <i8ak70$2ka$(E-Mail Removed)>, (E-Mail Removed) says...
>>
>> Oddly, the benchmark for my router (configured to use OpenDNS' DNS
>> servers) is faster than going direct to OpenDNS' DNS servers. My ISP's
>> DNS servers are faster than OpenDNS. So I configure my TCP settings to
>> use my ISP's best DNS server, my router's DNS server (which then goes to
>> OpenDNS), and my ISP's 2nd best DNS server as those are what the
>> benchmarks return for performance.
>
> I have also realized accessing my ISP DNS servers directly
> gives better results than accessing them indirectly
> through my router, set to act as DNS proxy.
>
> It was also adviced against using router DNS proxy abilities,
> as far as it was said to have inferior quality of DNS operations.
>
> But I am not expert in DNS.

And yet their own benchmarks showed that using my Linksys router's DNS
server was faster than directly using the same DNS servers specified for
my router to use. So it probably depends on what router you use. Maybe
some have a small amount of caching. Just using their benchmarks showed
router+DNS was faster than DNS.

In article <i8an6u$76c$(E-Mail Removed)>, (E-Mail Removed) says...
>

>
> And yet their own benchmarks showed that using my Linksys router's DNS
> server was faster than directly using the same DNS servers specified for
> my router to use. So it probably depends on what router you use. Maybe
> some have a small amount of caching. Just using their benchmarks showed
> router+DNS was faster than DNS.

Also it can depend on ISP DNS response time,
my ones returns cached request in 1-4 ms.
All four of them were faster in fresh check
than my Linksys router set to them.

Router DNS proxy do use caching
otherwise iw would not make much sense to have an extra step.

BTW faster than OpenDNS servers seem to be Google servers
8.8.8.8 and 8.8.4.4

--
Poutnik
It can be fast, good or simple. Choose 2 of 3.

On Sun, 03 Oct 2010 11:18:50 -0300, Shadow <(E-Mail Removed)> wrote:

>http://www.grc.com/dns/benchmark.htm
>
>Test your DNS servers and choose the fastest
>
>[]'s
>GRC's DNS Benchmark performs a detailed analysis and comparison of the
>operational performance and reliability of any set of up to 200 DNS
>nameservers
Hey, people, disable your system dns servers(that includes
routers), any caching by routers or hosts files will skew your
results.
The whole idea is to discover the FASTEST server, not to
compare it to your cache/hosts/router.
Then set your system/router to that DNS server(s), and re-run
the test.
[]'s

Shadow wrote:

> On Sun, 03 Oct 2010 11:18:50 -0300, Shadow <(E-Mail Removed)> wrote:
>
>>http://www.grc.com/dns/benchmark.htm
>>
>>Test your DNS servers and choose the fastest
>>
>>[]'s
>>GRC's DNS Benchmark performs a detailed analysis and comparison of the
>>operational performance and reliability of any set of up to 200 DNS
>>nameservers
> Hey, people, disable your system dns servers(that includes
> routers), any caching by routers or hosts files will skew your
> results.
> The whole idea is to discover the FASTEST server, not to
> compare it to your cache/hosts/router.
> Then set your system/router to that DNS server(s), and re-run
> the test.
> []'s

Huh? If you configure DNS servers in the TCP properties then you go
directly to those DNS servers, not to your router. Also, when DNS
queries are sent, they are sent to a specific host. If I send a DNS
query to OpenDNS, it does NOT use the router's DNS servers nor any DNS
functionality within that router. It's just another packet getting sent
through the router to some destination host.

You do NOT need to alter any of your TCP configuration regarding to
whose DNS server you normally use to resolve hostnames to IP addresses.
If your TCP is set up for dynamic assignment then you will be using your
router's DHCP server and it will give you itself as the DNS server. If
you use a static IP address for your host then you also have to specify
the DNS servers in your TCP settings - and you will use *those* hosts
(and only the router if you so include). Connections made by any
program, like GRC's DNSBench, are not using anything in your router with
regard to anything inside of it for DNS resolution. Requests made by
DNSBench go to whatever site is specified (which is by IP address).

This is complicated to you? There is a web browser inside your router,
too, used to configure it settings. You really think your router's web
server is in anyway involved with the traffic you generate with the web
server at some site? No changes are required in your TCP settings for
your host to make a direct connection to a off-network DNS server.

Meanwhile, at the alt.internet.wireless Job Justification Hearings, Shadow
chose the tried and tested strategy of:

> On Sun, 03 Oct 2010 11:18:50 -0300, Shadow <(E-Mail Removed)> wrote:
>
>>http://www.grc.com/dns/benchmark.htm
>>
>>Test your DNS servers and choose the fastest
>>
>>[]'s
>>GRC's DNS Benchmark performs a detailed analysis and comparison of the
>>operational performance and reliability of any set of up to 200 DNS
>>nameservers
> Hey, people, disable your system dns servers(that includes
> routers), any caching by routers or hosts files will skew your
> results.
> The whole idea is to discover the FASTEST server, not to
> compare it to your cache/hosts/router.

If using one's cache/hosts/router yields faster results than going directly,
then why *wouldn't* you use it?

--
<http://ale.cx/> (AIM:troffasky) ((E-Mail Removed))
19:27:46 up 11 days, 1:47, 9 users, load average: 0.01, 0.02, 0.17
"I am utterly appalled at how I have been treated like a criminal"
-- Andrew Crossley, ACS:Law, 13 August 2010

za kAT wrote:

> VanguardLH wrote:
>
>> Shadow wrote:
>>
>>> Shadow wrote:
>>>
>>>> http://www.grc.com/dns/benchmark.htm Test your DNS servers and
>>>> choose the fastest
>>>
>>> Hey, people, disable your system dns servers(that includes
>>> routers), any caching by routers or hosts files will skew your
>>> results. The whole idea is to discover the FASTEST server, not to
>>> compare it to your cache/hosts/router. Then set your system/router
>>> to that DNS server(s), and re-run the test.
>>
>> Huh? If you configure DNS servers in the TCP properties then you go
>> directly to those DNS servers, not to your router.
>
> I think you mean IP settings, not TCP properties.

Control Panel -> Network Connections
Right-click on your connectoid
Properties
Internet Protocol (TCP/IP)
*PROPERTIES*
There is where you configure the DNS servers used by your host. Yeah,
saying "IP properties" would've been more accurate.

>> Also, when DNS queries are sent, they are sent to a specific host.
>> If I send a DNS query to OpenDNS, it does NOT use the router's DNS
>> servers nor any DNS functionality within that router. It's just
>> another packet getting sent through the router to some destination
>> host.
>
> Yep, with the caveat that the built in resolver looks in its cache first.

Why would the local DNS cache be involved when connecting via IP
addresses? DNS resolution is only needed when you specify a hostname.
That is not the case here. DNS Bench is *not* using hostnames. It uses
IP addresses. No DNS lookup is required or even applicable if you are
already using an IP address.

My mistake was saying "they are sent to a specific host" without
mentioning that the host is specified by an IP address, NOT a hostname.
However, why would anyone specify a DNS server by its hostname which
would itself require a lookup? I've never seen anyone configure a DNS
server by hostname plus the input fields to do the configuration force
you to specify an IP address. In the IP properties for your host and in
DNS Bench, you can't enter the hostname. You can only enter the IP
address - and using an IP address doesn't involve DNS (local cache,
router cache, or anything DNS related).

> Most simple routers run DNS forwarders, not full blown DNS servers,

Yep, know that. They fail the lookup and then pass the request on to
their configured upstream provider. However, since IP addresses are
being used by DNS Bench, neither the local DNS cache or the router's DNS
functions are involved.