Re: [GTALUG] DNS benchmarking

[Evan Leibovitch via talk]

What was happening is that:

   1. Mobile Firefox refused to display some pages when I used my home
   wifi, claiming bad certificate or to SSL or something else. Switching off
   the home wifi to the cellular provider solved that problem.
   2. Some addresses had been hijacked, when my Economist app tried to call
   home to go beyond headlines, I would get danger notices that would go to
   phishing sites if I ignored the warnings.
   3. On the desktop, some web destinations that I knew were good
   *sometimes* were blocked by FF's security mechanisms.
   4. A site that had changed its name-to-IP info a month ago refused to
   resolve properly, no matter how many times I tried to clear cache
   5. An automated program that fetched downloads via HTTPS often (but not
   always) failed with error 111 (connection refused), eventually it would
   succeed after many retries

The solution I tried, even before doing the benchmarks, was changing the
DNS on the home DSL modem/router (which also did DHCP for the house) to
Google's public DNS (8.8.8.8, 8.8.4.4). All the above problems went away.

I subsequently ran namebench (have to invoke it with `python2`, won't run
with the current version) which said that my existing (Google) setup was
faster than any of the alternatives, including Teksavvy's outsourced CIRA
server.
FWIW, next fastest was Primus(!) at 216.254.141.2

I toyed with the idea of trying the Cloudflare 1.1.1.1 app but decided it
wasn't worth the bother. Has anyone else tried it?

- Evan

On Thu, Nov 3, 2022 at 1:30 PM Scott Allen via talk  wrote:

> On Thu, 3 Nov 2022 at 12:13, Evan Leibovitch via talk 
> wrote:
> > Part of the reason for my questions is that Teksavvy's own DNS is both
> slow and occasionally hijacked.
>
> Where are you configuring your DNS server(s) and what are you using
> them for? If it's mostly for web browsing, your browser might be
> configured for DNS over HTTPS (DoH) or you may wish to consider doing
> so. In this case the DoH server is probably configured in the browser
> settings and likely isn't Teksavvy's (I don't think they even provide
> DoH).
>
> --
> Scott
> ---
>
---
Post to this mailing list talk@gtalug.org
Unsubscribe from this mailing list https://gtalug.org/mailman/listinfo/talk

Re: [GTALUG] DNS benchmarking

[Stewart Russell via talk]

On Thu, Nov 3, 2022 at 1:29 PM Scott Allen via talk  wrote:

> ... If it's mostly for web browsing, your browser might be
> configured for DNS over HTTPS (DoH)
>
>
TIL that Firefox had already set this for me, and was using "CIRA Canadian
Shield" (the default). As for DNS, never explicitly fiddled with that
myself.

 Stewart
---
Post to this mailing list talk@gtalug.org
Unsubscribe from this mailing list https://gtalug.org/mailman/listinfo/talk

Re: [GTALUG] DNS benchmarking

[Scott Allen via talk]

On Thu, 3 Nov 2022 at 12:13, Evan Leibovitch via talk  wrote:
> Part of the reason for my questions is that Teksavvy's own DNS is both slow 
> and occasionally hijacked.

Where are you configuring your DNS server(s) and what are you using
them for? If it's mostly for web browsing, your browser might be
configured for DNS over HTTPS (DoH) or you may wish to consider doing
so. In this case the DoH server is probably configured in the browser
settings and likely isn't Teksavvy's (I don't think they even provide
DoH).

-- 
Scott
---
Post to this mailing list talk@gtalug.org
Unsubscribe from this mailing list https://gtalug.org/mailman/listinfo/talk

Re: [GTALUG] DNS benchmarking

[Evan Leibovitch via talk]

On Thu, Nov 3, 2022 at 9:27 AM Stewart C. Russell via talk 
wrote:

Looks like it got forked to this project, and the Go port was updated to
> at least replicated the original Python 2 functions:
>
> mrwiora/NAMEinator: NAMEinator DNS Benchmark tool (namebench successor)
> — https://github.com/mrwiora/NAMEinator


Thanks! I'll give it a try. Hopefully the hoops you describe won't be too
onerous. Never worked with Go before.



> This on about the slowest Teksavvy DSL around.
>

Part of the reason for my questions is that Teksavvy's own DNS is both slow
and occasionally hijacked.

I love them as a data pipe but their nameserver is ... sub-optimal.

- Evan
---
Post to this mailing list talk@gtalug.org
Unsubscribe from this mailing list https://gtalug.org/mailman/listinfo/talk

Re: [GTALUG] DNS benchmarking

[Stewart C. Russell via talk]

On 02/11/2022 13.32, Evan Leibovitch via talk wrote:

Hi all.

Can anyone recommend a modern alternative to "namebench 
"?


Looks like it got forked to this project, and the Go port was updated to 
at least replicated the original Python 2 functions:


mrwiora/NAMEinator: NAMEinator DNS Benchmark tool (namebench successor) 
— https://github.com/mrwiora/NAMEinator


Note that even it has hit some bitrot with the build instructions. 'go 
get' isn't a thing anymore, seemingly (hey, why keep something so 
obvious and useful?), so the build process is now described here:


 https://github.com/mrwiora/NAMEinator/issues/46#issuecomment-789247422

You might want to substitute

 $GOPATH/src/github.com/mwiora

for the various paths in that description.

Also, it seems to require go >= 1.16.1, which is newer than Debian 
stable. I tried to build it on a Raspberry Pi and it all went very 
pear-shaped. go's build errors are not very compelling: who knew that 
the quiet response "package embed is not in GOROOT" was a total build 
failure?


The results are terse, but possibly useful to someone:

 ./NAMEinator
 starting NAMEinator - version custom
 understood the following configuration: {numberOfDomains:100
 debug:false contest:true nameserver:}
 -
 NOTE: as this is an alpha - we rely on feedback - please report bugs
 and feature requests to https://github.com/mwiora/NAMEinator/issues
 and provide this output
 OS: linux ARCH: amd64
 -
 trying to load nameservers from nameserver-globals
 trying to load domains from alexa-top-2000-domains
 LETS GO
 900 / 900 [] 100.00%
 9 p/s 1m43s

 finished - presenting results:

 1.1.1.1:
 Avg. [82.974024ms], Min. [10ms], Max. [818.257401ms]

 1.0.0.1:
 Avg. [87.01652ms], Min. [10ms], Max. [1.331284997s]

 8.8.4.4:
 Avg. [87.607785ms], Min. [10ms], Max. [716.72469ms]

 8.8.8.8:
 Avg. [91.865296ms], Min. [10ms], Max. [817.38634ms]

 208.67.222.222:
 Avg. [114.947277ms], Min. [10ms], Max. [1.384931226s]

 127.0.0.53:
 Avg. [119.101241ms], Min. [10ms], Max. [718.128141ms]

 2001:470:20::2:
 Avg. [133.263523ms], Min. [10ms], Max. [614.40377ms]

 156.154.71.1:
 Avg. [136.4174ms], Min. [10ms], Max. [614.731507ms]

 216.146.35.35:
 Avg. [172.974867ms], Min. [10ms], Max. [1.33162978s]

 Au revoir!

This on about the slowest Teksavvy DSL around.

 Stewart






---
Post to this mailing list talk@gtalug.org
Unsubscribe from this mailing list https://gtalug.org/mailman/listinfo/talk

[GTALUG] DNS benchmarking

[Evan Leibovitch via talk]

Hi all.

Can anyone recommend a modern alternative to "namebench
"?

That's an old Google Summer of Code project that determines the
best-responding DNS servers for any particular installation. Python 2
based, so it doesn't even run without mods on a modern Ubuntu system which
defaults to Python 3. Also, it hasn't been updated since 2010 and certainly
some of the more-popular public DNS servers have changed since then. What
techniques or apps are used now to do this?

Evan Leibovitch, Toronto Canada
@evanleibovitch / @el56
---
Post to this mailing list talk@gtalug.org
Unsubscribe from this mailing list https://gtalug.org/mailman/listinfo/talk