Choosing DNS Servers for Asus AC68U

[Wutikorn]

I used DNS benchmark software to test the reliability and speed of 6 main DNS servers(2 of the same companies). I wonder how Asus router use DNS servers; does it use the primary one first, then the second if it's not working? or does it use both at the same time? I also wonder if it is okay if I put 8.8.8.8(Google) as primary DNS Server and 208.67.222.222 as secondary DNS Server. Will it do any good compared to using both of the same company. What I found out is that even I used 8.8.8.8 and 8.8.4.4, for example, the reliability is not the best of the two(my router is set to use 8.8.8.8 and 8.8.4.4 as DNS Servers). Even if 8.8.8.8 reliability is 100% for uncached, sometimes the router uncached reliability is not 100%. I don't need any protection from DNS Servers, but having that does not matter either. What should I choose to be my router's primary and secondary DNS Servers then? Based on benchmarks below. The longer red tab on the left, the less reliability. For the 3 tabs, the longer the slower. Red is for cached benchmark, green is for uncached, and blue is for dot com. Is it possible to use 3 DNS servers in total? Will that do any good?

Add: My ISP DNS servers are less reliable and slower than Google DNS Servers(Thailand).

 

[L&LD]

Most native, ISP's DNS servers are the best for each individual to use. Even if there are a few where other DNS servers are better (for speed).

Cached results are not a good indication of relative performance between different servers.

The best benchmark is to simply use a set of DNS servers for a day or two (or even a week) and see how responsive your network is using them. Real world? The ISP's servers are usually the preferred choice.

 

[Wutikorn]

Most native, ISP's DNS servers are the best for each individual to use. Even if there are a few where other DNS servers are better (for speed).

Cached results are not a good indication of relative performance between different servers.

The best benchmark is to simply use a set of DNS servers for a day or two (or even a week) and see how responsive your network is using them. Real world? The ISP's servers are usually the preferred choice.

Does real world mean how I feel? Or does it mean the uncached and dotcom parts(green and blue)? My ISP's DNS servers is less reliable than Google DNS most of the time I tested(set the router to automatically find DNS Servers and compared router to 8.8.8.8 in benchmark).

 

[L&LD]

Real world isn't a feeling. I do not know what the green and blue parts are actually measuring either.

But I was at a customer that has 50/10 Mbps ISP service level and light browsing on the internet was obviously slower than a customer that had a much slower 15/0.5 ISP service. After a little digging, I found google's DNS servers were setup in the router and checking my notes, was an 'addition' to what had setup originally. An employee had used a similar utility like you have and made the change on the router. It did not cause any issues for a few days, but when the ISP's DNS servers were restored (and key network devices rebooted), browsing speed was much faster and back to normal.

An example of a website that was slow was tomshardware.com (2 to 4 seconds to load the home page).

RMerlin has a couple of posts that also support keeping the ISP's DNS servers in use and basically ignoring utilities such as the one you're using to benchmark this aspect of your network, use search.

 

[Wutikorn]

Real world isn't a feeling. I do not know what the green and blue parts are actually measuring either.

But I was at a customer that has 50/10 Mbps ISP service level and light browsing on the internet was obviously slower than a customer that had a much slower 15/0.5 ISP service. After a little digging, I found google's DNS servers were setup in the router and checking my notes, was an 'addition' to what had setup originally. An employee had used a similar utility like you have and made the change on the router. It did not cause any issues for a few days, but when the ISP's DNS servers were restored (and key network devices rebooted), browsing speed was much faster and back to normal.

An example of a website that was slow was tomshardware.com (2 to 4 seconds to load the home page).

RMerlin has a couple of posts that also support keeping the ISP's DNS servers in use and basically ignoring utilities such as the one you're using to benchmark this aspect of your network, use search.

The blue one is dotcom, which I don't understand and the green one is uncached. So should I try how fast it is from using? Can you show me the link that RMerlin says those stuffs or please let me know where I can find them. Thanks

 

[L&LD]

The blue one is dotcom, which I don't understand and the green one is uncached. So should I try how fast it is from using? Can you show me the link that RMerlin says those stuffs or please let me know where I can find them. Thanks

You can try, but from my experience it is wasted time.

http://www.snbforums.com/threads/what-dns-do-you-use-with-your-asus-merlin.31473/#post-249763


The above link is an example of what I remember RMerlin stating and I agree completely with (based on actual experience).

https://www.google.ca/?gfe_rd=cr&ei...erlin+dns+benchmarking+software+snbforums.com


The link above is an example of the search terms used to get to that post.

 

[RMerlin]

Those benchmark tools test the wrong thing. This study shows the REAL performance impact in real-life usage, and it's generally worse.

 

[Wutikorn]

Those benchmark tools test the wrong thing. This study shows the REAL performance impact in real-life usage, and it's generally worse.

So is there a way to test real performance in my country or using my PC? I live in Thailand, so I wonder if it will be different. Would having my ISP DNS Server as Primary server and Google DNS Server as secondary be a good idea for reliability?

 

[L&LD]

Use the ISP's DNS servers for a week or so and if you find anything lacking try another DNS server (and I would keep primary and secondary DNS servers matched to the same provider too, for all tests) for the same time period under the same network use.

You cannot do real world testing with synthetic tests.

 

[Wutikorn]

Use the ISP's DNS servers for a week or so and if you find anything lacking try another DNS server (and I would keep primary and secondary DNS servers matched to the same provider too, for all tests) for the same time period under the same network use.

You cannot do real world testing with synthetic tests.

Let's say that ISP DNS servers are the best after all DNS Servers I had tested. Should I use both ISP DNS servers as primary and secondary? Or should I use main ISP DNS server as primary and use the other provider DNS server as secondary for reliability?

 

[L&LD]

Let's say that ISP DNS servers are the best after all DNS Servers I had tested. Should I use both ISP DNS servers as primary and secondary? Or should I use main ISP DNS server as primary and use the other provider DNS server as secondary for reliability?

Already indicated in the post you're quoting. "and I would keep primary and secondary DNS servers matched to the same provider too". Whichever DNS servers you happen to choose.

 

[Phoenix]

Changing your DNS server to anything other than your ISP's native DNS will make things slower due to more time needed to connect to a server that is far away. it's common sense, no point in wasting any calories on this

People in the US or UK may have better results since there are Google Servers there but for someone in Asia like me (I live in Dubai), it makes no sense

 

[Wutikorn]

Already indicated in the post you're quoting. "and I would keep primary and secondary DNS servers matched to the same provider too". Whichever DNS servers you happen to choose.

Oh, okay, your quote was after you were saying try another DNS server, so I was not sure if you mean that use same provider only for testing or for both testing and after testing.

Changing your DNS server to anything other than your ISP's native DNS will make things slower due to more time needed to connect to a server that is far away. it's common sense, no point in wasting any calories on this

People in the US or UK may have better results since there are Google Servers there but for someone in Asia like me (I live in Dubai), it makes no sense

Don't you have Google Servers in your country? There is at least one Google server in Thailand hosting www.google.co.th(or extremely close to Thailand as it has a very good ping), so that is why I was wondering. The benchmark shows me that I get Google DNS Server's responses faster than my ISP one as I believe they are exactly in the same province. However, because they say that the benchmark can't really be trusted, so I will keep my ISP DNS Servers.

 

[Phoenix]

Oh, okay, your quote was after you were saying try another DNS server, so I was not sure if you mean that use same provider only for testing or for both testing and after testing.

Don't you have Google Servers in your country? There is at least one Google server in Thailand hosting www.google.co.th(or extremely close to Thailand as it has a very good ping), so that is why I was wondering. The benchmark shows me that I get Google DNS Server's responses faster than my ISP one as I believe they are exactly in the same province. However, because they say that the benchmark can't really be trusted, so I will keep my ISP DNS Servers.

I live in Dubai bro, I don't know what do you think is the closest Google server to me?

When I tested (not in benchmarks) I felt that website loading was faster using my local ISP DNS vs Google's DNS servers

 

[RMerlin]

The time it takes to send a single DNS query matters little, because once you send one for a domain, it gets cached and doesn't need to be resent again. What matters is for that DNS response to point you at the closest server when dealing with Akamai (like when downloading anything from Microsoft) or Netflix/Youtube (who often have local caches running direct at your ISP's data center). Using a remote/random DNS will defeat that. You will save 100 ms querying the nameserver, but your download/video stream will be slower throughout its duration.

 

[Wutikorn]

The time it takes to send a single DNS query matters little, because once you send one for a domain, it gets cached and doesn't need to be resent again. What matters is for that DNS response to point you at the closest server when dealing with Akamai (like when downloading anything from Microsoft) or Netflix/Youtube (who often have local caches running direct at your ISP's data center). Using a remote/random DNS will defeat that. You will save 100 ms querying the nameserver, but your download/video stream will be slower throughout its duration.

So do you mean that even I get the query faster, it still take longer time time to load the web page? That would make a lot more sense. Thanks

 

[RMerlin]

So do you mean that even I get the query faster, it still take longer time time to load the web page? That would make a lot more sense. Thanks

That's correct, provided that webpage is distributed through a content distribution network and not from a single server located at a fixed location.