isp dns question

[jim trudel]

hi,


I made a test (dns benchmark) and my 3 isp dns are the fastest so I wonder if I would be better to keep those one... faster than google, cloudflare and open dns..
what do you think?
thank you.

 

[dave14305]

If it works well, stay with the ISP DNS. You will likely get optimized CDN content using your ISP. If you have any privacy concerns with your ISP collecting your DNS data (check a privacy policy), then consider a third party DNS (with or without encryption).

How good/bad were the other DNS servers? Was it significant difference?

 

[jim trudel]

How good/bad were the other DNS servers? Was it significant difference?

Very close to cloudflare and quad9.



Envoyé de mon SM-G960W en utilisant Tapatalk

 

[MichaelCG]

Were you comparing just ping latency? Or overall response latency?

I actually just found that in the past week or two, routing must have changed because from a network latency perspective, OpenDNS is now the same latency (~15ms) as Google...who is my local ISP. So to get OpenDNS filtering, I have switched back to using them.

Determine what you want here. My priorities were basic filtering and then performance. I have been flipping between CloudFlare and OpenDNS to get the most basic malware/badguy filtering while still keeping my overall performance as high as I can without doing detailed testing and tuning. Privacy....well....not really on my radar. Google (Fiber) and ATT (mobile) are my Internet providers. Both are notorious for not really being private. Google already knows me, my habits, and such due to my use of many of their services. When on my cell phone off of WiFi, I do tend to use my VPN to help "hide" some from ATT, but that really depends on how annoyed I get by the generally slower respond over the VPN.

 

[jim trudel]

Were you comparing just ping latency? Or overall response latency?

I actually just found that in the past week or two, routing must have changed because from a network latency perspective, OpenDNS is now the same latency (~15ms) as Google...who is my local ISP. So to get OpenDNS filtering, I have switched back to using them.

Determine what you want here. My priorities were basic filtering and then performance. I have been flipping between CloudFlare and OpenDNS to get the most basic malware/badguy filtering while still keeping my overall performance as high as I can without doing detailed testing and tuning. Privacy....well....not really on my radar. Google (Fiber) and ATT (mobile) are my Internet providers. Both are notorious for not really being private. Google already knows me, my habits, and such due to my use of many of their services. When on my cell phone off of WiFi, I do tend to use my VPN to help "hide" some from ATT, but that really depends on how annoyed I get by the generally slower respond over the VPN.

Thanks for your reply.
I tested with DNS benchmark, you can take a look, pretty complete test.

What is your firmware/ router?

I have quad 9 DNS, I know cloudflare is supposed to be faster, btw, if you use dns with filtering, what is the dns addresses you use with open DNS?

Also,, how you test the DNS latency?

Thanks

Envoyé de mon SM-G960W en utilisant Tapatalk

 

[MichaelCG]

Router: SophosXG on x86
OpenDNS IPv4:
- 208.67.222.222
- 208.67.220.220
Cloudflare IPv6:
- 2606:4700:4700::1112

I lied, I am using OpenDNS for IPv4 settings and Cloudflare for IPv6.

As for my testing, "ping -c 10 208.67.222.222" and using the AVG out of that. Far from perfect, but at least gives me some clue on overall network latency.

 

[RMerlin]

Ping latency for DNS is mostly irrelevant. You will never see the difference between 15ms and 30ms, plus once you've sent one query, it gets cached locally, so that means you end up seeing that 10-15 ms difference only on the initial query.

 

[jim trudel]

Ping latency for DNS is mostly irrelevant. You will never see the difference between 15ms and 30ms, plus once you've sent one query, it gets cached locally, so that means you end up seeing that 10-15 ms difference only on the initial query.

Do you know the dns benchmark test?

Is it a good test?

Envoyé de mon SM-G960W en utilisant Tapatalk

 

[MichaelCG]

Ping latency for DNS is mostly irrelevant. You will never see the difference between 15ms and 30ms, plus once you've sent one query, it gets cached locally, so that means you end up seeing that 10-15 ms difference only on the initial query.

While I mostly agree with this...keep in mind with modern CDNs, DNS results have very short TTL so your client should not cache responses for very long. However, 15ms vs 30ms....yeah, for the most part, not a huge deal. Now comparing 15ms to 100ms, you might start to notice things especially when loading larger more complex web sites. A good example would be something like cnn.com. The number of DNS queries just to load the main page will surprise many.

Then again...many of us on here tinkering are rarely happy with "good enough" and like to tune/tweak things more than we should.

 

[RMerlin]

Do you know the dns benchmark test?

I don't know, I never kept track of them. Check what it actually tests. If all it does is test the ping latency, then it's worthless. If it also test resolution time for various well-known domain, then it can be useful.

 

[RMerlin]

While I mostly agree with this...keep in mind with modern CDNs, DNS results have very short TTL so your client should not cache responses for very long.

I would expect ISP edge caches records to have longer TTLs however, in addition to being possibly almost permanently within their own servers cache, since there are constantly customers querying for those specific records, causing them to be re-cached very frequently.

It's another benefit of using your ISP's servers instead of, for example, implementing your own resolver on a VPS (which would be the way to go if people are being paranoid about their DNS queries - setup DoT and a recursive resolver on a VPS under your control, and point your router at it). Your own resolver will rarely have results within their own cache, so more often a recursive query would be required, which is when DNS lookup performance can drop significantly.