DNS Benchmark V2 - Lost connectivity

[Info]

Sorry, just a few questions:

First, I'd like to thank DeanR for the discovery. My ISP must have some limitation when many queries are made. Thanks to him, and with a maximum list of 60 resolvers, no more errors appear. Above 60 resolvers, I get a connection error.

Based on the results, the DNS server cleanbrowsing 185.228.168.9 is superior to the Quad9 9.9.9.9 that I currently use, correct?

A silly question: when I ping cleanbrowsing I get 7ms compared to 6ms for Quad.

When I do a tracert to cleanbrowsing I get 13 hops, compared to 11 for Quad.

Isn't this relevant? Or should I consider Short by cached Performance?

Best Regards,

 

[Steve]

@Info: The black link shown in the bar chart is the average of the three DNS query types — Cached, Uncached and DotCom. That's probably what matters most. We see that you have some extremely fast (red) cached response times. That must be due to the extremely high quality of your Internet connection and good hardware that's getting the packets to and from your machine.

The (green) uncached response times are significantly longer, which is to be expected, since that represents the resolver you're querying needing to, in turn, query that other domain's resolver for its IP.

The original v1 DNS Benchmark would ONLY take the cached responses into consideration. That might have been the right thing to do 16 years ago. But the Internet is a far different place today with web pages being assembled from scores of different domains as ads and code libraries and (annoyingly) trackers are being referenced. So today's v2 Benchmark averages all three query types since they have become equally important.

 

[Steve]

@Info: The “ping time” is probably very equivalent to the resolver's cached performance since most of the time will be spent with packets in transit. And we see exactly that in your DNSB chart. Quad9's ping response (the red bar) at 6ms is just a tiny bit shorter than the red bar for the CleanBrowsing at 7ms. (Pro Tip: If you LEFT-CLICK and hold on a resolver, you can directly read the timings for each of those bars and for the "black line" which is their average.)

But the reason the CleanBrowsing resolver might make more sense, is that Quad9 is slower (and a LOT slower for the dotcom queries) which means that anytime you ask for a domain name that Quad9 doesn't already have in its cache it needs to, in turn, ask other resolvers for the domain's IP address and we see that takes more time for Quad9 than for the CleanBrowsing resolver.

Now, in fairness, Quad9 might have a larger cache, or its popularity might mean that there's a greater likelihood that the domain you want to lookup is in its cache. At the moment there's no way to tell. Someday, I might tackle arranging to emit simultaneous queries to multiple resolvers so that real-world statistics could be obtained. But I have a few other goodies to create first!

Thanks again for your interest and support... And feel free to spread the word!

 

[Steve]

@Info:

One other thing I noted was that you had your Sideline Threshold set to 100msec. That would normally not be any problem. But in an instance where your ISP appears to be sensitive to the amount of DNS you're doing, setting a lower threshold — perhaps even 25msec — will allow DNSB to quickly and automatically disqualify slower resolvers to prevent their subsequent benchmarking. So you'll automatically wind up testing fewer resolvers and the ones you do test will be among the fastest.

 

[Info]

@Info:

One other thing I noted was that you had your Sideline Threshold set to 100msec. That would normally not be any problem. But in an instance where your ISP appears to be sensitive to the amount of DNS you're doing, setting a lower threshold — perhaps even 25msec — will allow DNSB to quickly and automatically disqualify slower resolvers to prevent their subsequent benchmarking. So you'll automatically wind up testing fewer resolvers and the ones you do test will be among the fastest.

@Steve
Everything is working now with my list, my limit is 60 DNS servers, I don't have any more problems.

Another round of testing, now with 50x - 50msec, took 2:30 hours.

Thank you, very happy with the result and all the work.

@DanR
Thank you for all your help and patience in helping me figure out my problem. There really was no need for a list of 287 resolvers; that was my problem. My ISP limit is 60.

Best Regards

 

[DanR]

Thank you for all your help and patience in helping me figure out my problem. There really was no need for a list of 287 resolvers; that was my problem. My ISP limit is 60.

You are very welcome! I'm glad that things are working well for you now.

I would suggest you consider building a Custom List. This process is very much more benign than a normal benchmark process. I would not expect your ISP to have any issue with it. When it is done, you will have a list of the 50 or so fastest resolvers available to you at your location, plus your system resolvers.

This will typically be a better list than the pruned down Built In list, which is not optimized for your location.

After a benchmark run or two, you may prune off the bottom 1/3 or so as too slow, or problematic, etc. The resulting list will benchmark very fast!

 

[Info]

You are very welcome! I'm glad that things are working well for you now.

I would suggest you consider building a Custom List. This process is very much more benign than a normal benchmark process. I would not expect your ISP to have any issue with it. When it is done, you will have a list of the 50 or so fastest resolvers available to you at your location, plus your system resolvers.

This will typically be a better list than the pruned down Built In list, which is not optimized for your location.

After a benchmark run or two, you may prune off the bottom 1/3 or so as too slow, or problematic, etc. The resulting list will benchmark very fast!

Could you tell me what the process of creating it would be like building a Custom List.?

I imagined it was just a matter of adding the desired DNS servers to DNSBench.ini.

 

[DanR]

Could you tell me what the process of creating it would be like building a Custom List.?

I imagined it was just a matter of adding the desired DNS servers to DNSBench.ini.

Nope! It's much simpler than that.

Click the red icon in the UL corner of the DNSB main window to access the drop down main menu.

Almost halfway down there will be the item [Build Custom Nameserver List] - click it.

In the popup window you have one choice to make: Check [DNSSEC only]. Or not.

Then click [Build Custom List] - that's it!

DNSB downloads a data base of 4,909 resolvers from GRC and proceeds to do a quick speed test on each one from your location. It consistently takes 16 minutes for me at my location.

When it is done it will present you with the 50 or so fastest resolvers for your location. After benchmarking this list a time or two, too slow resolvers, unresponsive resolvers, etc. may be pruned from the list. The resulting shortened list will benchmark quite fast.

You may repeat this exercise as often as you wish.

Note: I have IPv6 disabled (no connectivity). This reduces the Custom Build list to 4,807 resolvers for me.

 

[Info]

Nope! It's much simpler than that.

Click the red icon in the UL corner of the DNSB main window to access the drop down main menu.

Almost halfway down there will be the item [Build Custom Nameserver List] - click it.

In the popup window you have one choice to make: Check [DNSSEC only]. Or not.

Then click [Build Custom List] - that's it!

DNSB downloads a data base of 4,909 resolvers from GRC and proceeds to do a quick speed test on each one from your location. It consistently takes 16 minutes for me at my location.

When it is done it will present you with the 50 or so fastest resolvers for your location. After benchmarking this list a time or two, too slow resolvers, unresponsive resolvers, etc. may be pruned from the list. The resulting shortened list will benchmark quite fast.

You may repeat this exercise as often as you wish.

Note: I have IPv6 disabled (no connectivity). This reduces the Custom Build list to 4,807 resolvers for me.

Thank you so much, I didn't realize that option existed, it worked too.

 

[johnawolf]

Hello everyone, I am very pleased to have acquired my DNSBench license, however I am experiencing the error below.

View attachment 1840


I'm running in standard 5x mode, I don't have IPv6 enabled, only IPv4.
Is there something I should do?

Best Regards,

I am having the same issue and changing the various options does not seem to fix the issue. I changed the benchmark speed to 100ms and still no luck. I would note that v1 still works.

 

[DanR]

I changed the benchmark speed to 100ms and still no luck.

100 msec is too fast for your ISP. Does the default of 20 msec work?

 

[johnawolf]

No, it does not I still get the "internet connectivity was lost while benchmarking" error. My Internet connection was not lost so I am not sure what the issue is. v1 of the benchmark still works.

 

[Info]

No, it does not I still get the "internet connectivity was lost while benchmarking" error. My Internet connection was not lost so I am not sure what the issue is. v1 of the benchmark still works.

Remove the resolvers to a maximum of 40 DNS servers; your provider shouldn't handle too many resolvers.

 

[DanR]

@johnawolf What Info said!

And you might try Custom Build for resolvers optimized for your location. See my post above for instructions.

 

[Info]

@johnawolf What Info said!

And you might try Custom Build for resolvers optimized for your location. See my post above for instructions.

I created a custom build, it took 16 minutes, and even then I analyzed the list and chose to remove about 10. After having the customized list, I went ahead and tested it at 100x speed, and the result was what you see in the attachment.

Now I can remove the worst ones.

 

[DanR]

@johnawolf You might also try reducing the speed further to 150 or 200 msec. This means longer benchmark times, but if it keeps your ISP happy . . .
Note: Less msec is faster. In my first post to you my thinking was backwards.

 

[Steve]

I am having the same issue and changing the various options does not seem to fix the issue. I changed the benchmark speed to 100ms and still no luck. I would note that v1 still works.

Hi John.

During testing we discovered that some ISPs (and many VPNs) were throttling UDP or, in some cases, blocking it outright if they saw too much UDP DNS activity. However, it's suspicious that v1 is working without trouble for you.

Does the interruption notice appear immediately? Or after some time into the run of the benchmark? That screen shot you quoted as being the same as yours ( from @Info ) showed that only 2 resolver queries had been sent, thus an immediate loss notice. Are you really seeing the same?

Thanks!!

 

[johnawolf]

Hi John.

During testing we discovered that some ISPs (and many VPNs) were throttling UDP or, in some cases, blocking it outright if they saw too much UDP DNS activity. However, it's suspicious that v1 is working without trouble for you.

Does the interruption notice appear immediately? Or after some time into the run of the benchmark? That screen shot you quoted as being the same as yours ( from @Info ) showed that only 2 resolver queries had been sent, thus an immediate loss notice. Are you really seeing the same?

Thanks!!

Steve, with the default set up and most of the adjustments suggested it is popping up immediately. It will run (sort of) with the list pared down to 49 resolvers and the benchmark speed set to 200 ms. I say sort of because the pop up does not come up immediately and if you click ignore it will run for a bit before it comes up again, typically 10s of queries. Not very practical to babysit it through a full run.

v1 loads the names of the DNS servers (which v2 does not) and runs to conclusion with no errors. At the conclusion, it did report a high number of unreliable resolvers which I don't recall seeing before. FYI my ISP is Spectrum and my router is an Eero 6+ via wire without the Eero+ stuff active.

Any thoughts would be appreciated.

 

[Info]

Steve, with the default set up and most of the adjustments suggested it is popping up immediately. It will run (sort of) with the list pared down to 49 resolvers and the benchmark speed set to 200 ms. I say sort of because the pop up does not come up immediately and if you click ignore it will run for a bit before it comes up again, typically 10s of queries. Not very practical to babysit it through a full run.

v1 loads the names of the DNS servers (which v2 does not) and runs to conclusion with no errors. At the conclusion, it did report a high number of unreliable resolvers which I don't recall seeing before. FYI my ISP is Spectrum and my router is an Eero 6+ via wire without the Eero+ stuff active.

Any thoughts would be appreciated.

Please reduce the number of DNS resolvers to only 40 or 30 and test again, regardless of the time frame. Time is never a problem for me; the issue is the number of DNS servers.

 

[johnawolf]

At 39 resolvers and 20 ms benchmark speed it appears to be working. However, it still lists the resolvers as ... determining ownership ... but at least it is running.