noob question

[Frantz]

I am currently using Google's DNS servers which I imagine are reasonably performant. If I run DNS Benchmark and find alternative servers that respond a few milliseconds faster, will it actually make a noticeable difference?

I guess I am trying to understand what is the point of this. Is it geeking out for the sake of geeking out or will changing DNS servers make things better for me?

 

[DanR]

I am trying to understand what is the point of this. Is it geeking out for the sake of geeking out or will changing DNS servers make things better for me?

The point of this is to see how your chosen system, resolvers compare to what is out there to choose from.

A Custom Build allows comparison with the fasted resolvers available to you at your location.

If your chosen system resolvers rank up there with the fastest, you need do nothing - other than running DNSB occasionally to verify all is still OK.

However, if the system resolvers are notably slower than the faster ones out there, then changing the system resolvers to the faster resolvers will very likely improve browser performance via vaster page downloading.

 

[coffeeprogrammer]

I just got the new version tonight and have not run it yet, however based on what was said during the podcast, it can find the fastest DNS over HTTPS as well as the normal unencrypted DNS servers. So you could use it just to find the fastest DNS over HTTPS to use in your browser and then use a different one for your operating system (although I think Win11 support DNS over HTTPS as the system resolver). The point being, these days people might be using more than one DNS server, if you want some kind of filtering on your browser for example, maybe the benchmark would be more useful). Just figured I put in my two cents.

 

[PHolder]

@Frantz Yes, I hear you, and when Steve started on this project, I asked the same question of him. And I guess his answer is pretty fair... he doesn't necessarily understand why the free DNSB is so popular, but the fact that it was motivated him to see if a paid version was called for. And here we are. So I think the answer to your question is: If you're happy, and things are working for you, and you're not feeling especially nerdy to get stats for the sake of stats, then probably just stay happy and don't worry. As you already suspect, you're not missing out on any "secret" or "feature". On the other hand, the free version remains free, so it would be free to use it to check how your config (your router, or whatever you've assigned as your DNS gateway) is performing. If it turns out that doing that gets your nerd on, then you can step up to V2.

 

[Schrodinger's Cat]

I am currently using Google's DNS servers which I imagine are reasonably performant. If I run DNS Benchmark and find alternative servers that respond a few milliseconds faster, will it actually make a noticeable difference?

It depends on your specific scenario. DNSB is a tool that benchmarks and gives you a quantitative (that is numbers) measurement and options.

In my case, DSNB did validate that my ISP provided udp53 DNS servers to the fastest, out ranking clouflare google and others (udp/doh/dot alike), but i chose not to use the ISP provided ones as they run a government mandated blocklist. With the dnsb servers ranked, i could pick a different one (the next fastest) with the features i care about, eg. privacy, encryption, custom ad/blocklists etc

It may improve your page load times (but then it depends on your usage patterns) if sites you visit have lots of assets from different domain names (roll eyes at modern e-commerce giant sites, often with ultra low TTL dns entries), that would make many dns queries which may multiply the time shaved off using a dns resolver with 1 or 2 ms faster average query response time.

tl,dr: At best it could make pages load slightly snappier, at worse, not noticeable; but speed isnt everything, there are other options out there and dnsb can help measure performance of a provider you didnt know about that have the features you care about (encryption/security/privacy etc)

 

[Frantz]

Thanks, I appreciate everyone sharing their thoughts on this.

 

[GreenWine]

As Schrodinger's Cat suggested, I'm interested in using dns bench 2 to find out which is the fastest private, secure and full featured DNS provider for me.

I'm currently paying for nextdns, but there are others with similar offerings.

For now, I just like playing with it and bought it to support GRC. I have listened to SN since 0001, so it was about time I put my hand in my pocket.

 

[peterblaise]

If we only have IPv4 we can test with the free DNSBench 1.

If we want to test IPv6 - and DoH and DoT - DNSBench 2 can help.

Either can confirm we're getting what we expect.

Plus a variety of Internet speed tests also provide a view into our
performance choices.

For me, I spread the services, having a dozen resolvers on each
protocol on each network interface, just so that if any one of them
pauses for a moment, ever, my computers automatically move on to
the next resolver.

Periodic DNSBench tests confirm that all my chosen resolvers are
essentially equivalent - culled from periodic DNSBench runs of Build
Custom List and Benchmark.

What I haven't done thoroughly is test filtering resolvers to see if I
can insulate myself from risk and still get good performance.

There's always more testing to do.
If anyone has a big collection of filtering resolvers in any INI files,
please do share - I'm always on the hunt.

Things like:

1.1.1.2 security.cloudflare-dns.com
86.54.11.1 protective.joindns4.eu
185.228.169.9 security-filter-dns2.cleanbrowsing.org
185.228.168.168 family-filter-dns.cleanbrowsing.org

... but how do we test the qualities of the filtering?

Thanks.

 

[AlanD]

If we only have IPv4 we can test with the free DNSBench 1.

But bear in mind that the text conclusions panel is biased towards cached performance in V1, whilst V2 looks at overall performance. Looking at the raw numbers in V1 will give you a better picture.

 

[peterblaise]

Yes, DNSBench 2 ranks by the average of three queries:

cached
uncached
dot com lookup

DNSBench 1 ranks primarily by cached, and only ranks by the other
measurements in case of a tie in cached performance.

But both programs report on all three query measurements.

The results should correspond well.

Their Benchmark results can be manually compared 1:1 in
the Tabular Data tab and or in the saved Tabular data file, and or in
a CSV file.

But, yes, the conclusions may differ, where DNSBench 1 leans
toward the best cached performance, and DNSBench 2 leans
toward the best average performance.

 

[Steve]

If we only have IPv4 we can test with the free DNSBench 1.

If we want to test IPv6 - and DoH and DoT - DNSBench 2 can help.

As @AlanD said, what you posted is NOT correct, Peter.

The world has changed dramatically since the release of v1, with the result that caching performance is no longer the overriding determiner of overall resolver performance. The connectivity of a remote resolver to the rest of the Internet — where the uncached answers lie — has become at least as important as the remote resolver's connection to the user.

• Cached performance reflects the user's connection to the remote resolver.
• Uncached and DotCom performance reflects the remote resolver's connection to the rest of the Internet.

With websites now making scores of requests for domains that all need to be looked up, the remote resolver's ability to obtain those answers for its user has become at least as important as its ability to reply to domains that are already in its cache.

For this reason, v1's "cache priority" approach no longer make sense and is mis-ranking resolvers for today's much more distributed world. We've been seeing feedback from v2 users who are shocked that their local residential NAT router which used to be #1 in the list under v1 doesn't even make it onto the screen without scrolling down under v2. The reason is that their residential NAT router is very poorly connected to the rest of the Internet compared to the big CDN routers from Cloudflare, Google, Quad9, etc. THEY are the better choice today, so v1 is no longer serving anyone well.

If someone wished to use v1 properly for IPv4 only, they could manually average the performance of all three query types themselves (or they could just grab a copy of v2 which will do that for them, along with producing much more stable results due to 5x the sampling by default.