What to make of DNSB v2 r4?

[Texas Flyer]

Long time fan of Steve's, paid user of his products and user of most of the freeware he has created. So you understand this is not a "hit" post against Steve. I am running a Protectli Vault (Steve's recommendation and love it), with an ASUS AI Mesh network behind it. In the original DNSB, I had multiple DNS that always showed fastest. In the current iteration, I understand Steve's comments about the subtleties of how today's internet works and why a combo score is more effective at determining a "best" DNS selection. I have collectively run the latest version/release many, many hours, day and night, on a quiet network. Every time it runs, it provides a different ranking of bests. If I update my DNS settings, refresh my DHCP issued addresses to my PC, verify they are the new ones, and run the test again, it shows those are often no longer the best servers. I can imagine that many folks may be frustrated by the ever changing best. I always evaluate my choices by clicking and scanning down the results to see the differences in overall scores. Often the #1 to #10 are a few thousands of a second different at most. I traditionally have sticked to Quad 9, Cloudflare, OpenDNS as known quantities with or without some level of malware blocking. One test they will be near/at the top of the test, the next time they aren't even visible on the screen without scrolling. Clearly, activity on the internet, beyond our control affects the timing and scoring of individual servers.

Honest question - how does one effectively make use of these results? Why use these results at all? Why not just run with a well known brand (OpenDNS, Quad 9, Cloudflare) and call it good? All of them will be better than a local router based DNS. If I have missed some additional explanatory information as Steve is updating the support pages for v2 then I'd appreciate a point out to that information.

 

[peterblaise]

Agreed, I have collected ~100 DNS Servers in my 'fastest' results from
DNSBench 2 over time, and I juggle applying them and seeing how my
systems 'feel'.

Those top ~100 DNS Servers are variable BUT they all are also
essentially equivalent, such that their variability seems inconsequential.

I think the fastest DNS Servers out there are so close to each other in
performance that we're going to see many, many worthy offerings.

Perhaps we are at least getting confirmation of our choices, especially
versus ISP defaults, and we also get offerings of more-responsive
alternatives, and we can test again whenever we want to confirm again,
and make tweaks again, and confirm again, whenever, over time.

Analogy: I have a tire gauge that confirms my tires are properly inflated
- I wouldn't be without it, and I use it often, even if just to confirm.

And, yes, DNSBench 1 scheme of ranking DNS Servers according to
their cached responses may no longer match how complex,
multi-sourced web pages are delivered nowadays.

DNSBench 2 ranks by average of all three measurements: cached,
uncached, and dot-com lookup,

I look forward to more people's experiences with DNSBench 2.

 

[PHolder]

Honest question - how does one effectively make use of these results? Why use these results at all? Why not just run with a well known brand (OpenDNS, Quad 9, Cloudflare) and call it good?

When Steve announced to his testers on his newsgroups that he was embarking on a new [paid] version of DNSB, I asked the same question. Personally, I can't tell any difference between any of the top contenders for "best speed" of DNS Server, so I choose based on reputation, and I chose Quad 9 because it theoretically offers some extra security. Steve's answer to me was, that he didn't understand why DNSB V1 was so popular on his site (implying that he doesn't fully understand how people make use of it) but since it was so popular, he figured he needed to know if a paid version was called for. And now, a year later, here we are. In the end it's just a tool... use it however you feel best suits your needs.

Personally, I run it all the time because I am a tester of it, but I have never once let it influence my decision about how my network should be configured. I was never using my ISP's chosen DNS since Quad 9 was announced, and I was probably even then using Google or something else before that. I think it's good to use the tool to be aware of how things look/work on your network (as a snapshot in time) and it can let you become aware of some other choices if you happen to have a sub-optimal one. Once you've got things "optimized" for your needs... maybe use it now and then to check back in to make sure no weird problems have developed.

 

[DanR]

Clearly, activity on the internet, beyond our control affects the timing and scoring of individual servers.

Exactly. That is something that we as users can neither predict or control. Thus DNSB results will vary somewhat during any given day, from day to day, etc.

Honest question - how does one effectively make use of these results? Why use these results at all?

As a long time tester I have observed that, with multiple runs over time, certain DNS servers will most often (but not always) be at or near the top of the list. These, therefore, are the optimum DNS servers for me at my location. As long as that generally includes my chosen system DNS servers, I feel I am good and no changes are necessary.

However, if a system DNS server fairly consistently ranks down in the list, that would be an indication to consider removing it and perhaps replacing it with a faster one as indicated by multiple DNSB run results.