dslreports logo
 
    All Forums Hot Topics Gallery
spc
Search similar:


uniqs
5533

d4m1r
join:2011-08-25

d4m1r

Member

[Cable] For those that use OpenDNS or Google DNS...

If you are experiencing slow downloads but your overall connection is fine, you might wanna look into something called "Akamai". It is basically a server TSI has where they can store popular downloads (itunes files, game patches, etc) locally so it saves them external bandwidth and you get the files quicker too.

However, if you are using a 3rd party DNS service, you are not able to access your ISPs Akamai server, which is usually a lot closer to you geographically (at the POI?). Using OpenDNS or Google DNS means you are downloading from them meaning it depends on where they have DNS servers, and it's usually in the U.S or E.U...

ghostly57
join:2011-08-22
QC, Canada

ghostly57

Member

How can I access it?

Inssomniak
The Glitch
Premium Member
join:2005-04-06
Cayuga, ON

Inssomniak

Premium Member

Simply using TSI DNS will access it when required.

I didn't know TSI had one of these servers.

drjp81
join:2006-01-09
canada

drjp81 to d4m1r

Member

to d4m1r
I'd get my facts straight before saying stuff like that. It's a mirroring service that has nothing to do with Teksavvy, really. And they don't do it for free. These people do it in order to gather usage statistics to sell to marketing firms, among other things. So if privacy is a concern keep using your third party DNS.

»en.m.wikipedia.org/wiki/ ··· nologies
Eug
join:2007-04-14
Canada

Eug to d4m1r

Member

to d4m1r

I stopped using TSI's DNS servers a few years back...

I stopped using TSI's DNS servers several years back because they were wreaking havoc on my Macs. Sites simply wouldn't load, or else would load after several minutes or incompletely.

I switched to public DNS servers and never looked back.

I wonder what they're like now. I suspect they're fine, or else there'd be lots of complaints here, given that Apple products are near ubiquitous these days.

SimplePanda
BSD
Premium Member
join:2003-09-22
Montreal, QC

SimplePanda to d4m1r

Premium Member

to d4m1r

Re: [Cable] For those that use OpenDNS or Google DNS...

Generally speaking the OP is mostly correct in principal... using your ISP's assigned DNS will always give you optimal performance as the IP resolutions you receive will be accurate for your routing.

8.8.8.8 (Google Public) and OpenDNS servers are indeed anycasted to nearest server, but nearest isn't always near enough to get an accurate resolution.
34764170 (banned)
join:2007-09-06
Etobicoke, ON

2 recommendations

34764170 (banned) to drjp81

Member

to drjp81
said by drjp81:

I'd get my facts straight before saying stuff like that. It's a mirroring service that has nothing to do with Teksavvy, really. And they don't do it for free. These people do it in order to gather usage statistics to sell to marketing firms, among other things. So if privacy is a concern keep using your third party DNS.

»en.m.wikipedia.org/wiki/ ··· nologies

Then get your facts straight before making such a completely bogus post.
34764170

34764170 (banned) to Inssomniak

Member

to Inssomniak
said by Inssomniak:

I didn't know TSI had one of these servers.

Not surprising since it's already known that TSI has a Google Global Cache cluster on their network. With TSI being as big as they are and Akamai being the largest CDN it's only logical to have an Akamai cluster on their network as well. The Akamai clusters consist of 4-8 servers.
BikeHelmet
join:2010-04-15

BikeHelmet

Member

My router has 3 DNS servers set.

8.8.8.8:53 - Google
8.8.4.4:53 - Google
206.248.154.170:53 - Teksavvy


I was experiencing some packet loss, so I set my router to send DNS requests to all three. Whichever comes back first is the one that it goes with.

If you have a Tomato router, go to:
»192.168.1.1/advanced-dhcpdns.asp

To apply the same tweaks, enter in:
cache-size=2048
all-servers


The default cache size is 150 (!?) and then entries are kicked from the cache. If you have 2+ avid websurfers, this could be an issue, hence my increasing it.

You may also want to increase the number of simultanious DNS requests that can be made:
dns-forward-max=256

But not too high. Heck, maybe even 256 is too high.

Just for the record, I am experiencing many slow single-connection downloads. I'll try removing the Google DNS and see if it changes.

d4m1r
join:2011-08-25

d4m1r

Member

I never experience slow downloads (slower during weekday nights but nothing major) but once doing some technical research into OpenDNS (which is what I use because of Googles privacy policies) I found out about these Akamai servers. For Ontario cable customers, they are Rogers and not TSI's as far as I know.

I always download from HTTP direct links and they are not usually popular files so I personally won't notice a difference but I've seen several people post that their downloads are slow but overall web traffic is fine which I found weird..
BikeHelmet
join:2010-04-15

BikeHelmet

Member

Click for full size
priyen asked me to post more info.

These are the pages to be edited. They'll look slightly different because I'm using Toastman's Tomato rather than TomatoUSB or stock Tomato.

More info on the options I chose here:
»www.thekelleys.org.uk/dn ··· man.html

log-async is a performance option. It needs to be somewhere between 5-25 (apparently) or large numbers of DNS requests can stall while it waits to log them.

no-negcache prevents your router from remembering that there's no domain. It checks every single time. Every once and a while overclock.net would timeout for me - some weird DNS/website glitch. After it happened I'd have to reboot my router to clear the DNS table. (until I discovered this tweak)

It happened with both Telus and Teksavvy, so it's some weird issue with where their servers were located and how they were setup. Any which way, now they're with a different host, so it's not strictly necessary.

dhcp-authoritative says that this router is the only DHCP router on the network, and must be obeyed. It apparently speeds up DHCP negotiation slightly.

Reduce packet size shrinks DNS packets to prevent ornery routers between you and the DNS servers from dropping them. Near as I can tell that one should be enabled for everyone, so I'm not sure why it's optional/visible. Unticking it caused DNS timeouts for me. (Long 20-30 second delays before webpages started to load)
highwire2007
join:2008-05-17
Nepean, ON

highwire2007 to d4m1r

Member

to d4m1r
With TSI's default DNS servers configured in my router, Google's Namebench program tells me that TSI's are the fastest, followed by UltraDNS and then OpenDNS. Guess who's last? Bell, of course, with very slow response times, and lots of time-outs.

Try it out:

»code.google.com/p/namebe ··· ads/list
BikeHelmet
join:2010-04-15

BikeHelmet to d4m1r

Member

to d4m1r
Does anyone have a full list of Teksavvy's DNS servers? I was surprised to find out they had some out in BC as well.

This site makes no mention of the BC ones:
»fixppp.org/tsi.php

Google:
8.8.8.8
8.8.4.4

Teksavvy:
76.10.191.198
76.10.191.199
206.248.154.170
206.248.154.22
Eug
join:2007-04-14
Canada

Eug to highwire2007

Member

to highwire2007
I ran namebench and it said that for the fastest lookups, TSI was faster, but for average speed, Google DNS was faster. ie. Google DNS was more consistent.

TSI Marc
Premium Member
join:2006-06-23
Chatham, ON

TSI Marc

Premium Member

Right that's because googles cache will hold more queries then ours.. When ours need to go out to the web to get answers that google already has you will see that gap.

XNemesis
join:2002-11-16
Kitchener, ON

XNemesis to d4m1r

Member

to d4m1r
sorry for the thread jack but is it better to use TSI's or Googles dns? And for whatever one, do I just input the numbers BikeHelmet posted in the DNS fields in my E2000 router?

fluffy
@teksavvy.com

fluffy

Anon

use the following instead :
8.8.8.8
4.2.2.1
91.191.136.152
its better using these instead of google or tekavvy dns only. works much better.

Sandroid
BSD geek
Premium Member
join:2002-08-08
Anjou, QC

Sandroid to d4m1r

Premium Member

to d4m1r
It's amazing the amount of bull$h1t on here sometimes. Don't call yourself a geek if you don't know what akamai does.

Kudos to brad for calling out the guy trying to call out the OP.

The speed improvement with DNS is not just because of the correctly geolocated IP for the response to a query. At any rate, as someone pointed out, your request, as it is, is being diverted to a "better" (ie: closer) DNS server.

The main reason to use your ISP's is because A: Typically, the cache on the ISP server is more relevant than what may be in some public DNS, outside what is cached for sites like google or twitter. B: the perception of slowness is sometimes due to DNS performance, as most tuned websites are made in such a way that once you connect to the server, the responses that make up the content of the page you see, making you feel like something is happening, is pretty damn quick (for those sites where the devs are clued into creating a high performance site, and not just a fancy one to show off that they can code).

If I have to do a DNS lookup, and it takes 60 extra ms to reach the remote DNS server, and then it takes another 30-40 seconds to do the iterative query on the google dns for teksavvy.com (which may not be cached there) you end up with that annoying delay where you browser says "Looking up www.teksavvy.com" and once it's done, everything is lightning quick. Even if everything is cached on both, every lookup will still require a 60ms RTT to googleDNS, whereas it may be as little as 10-12ms to your ISP's DNS server.

Here's an example below. It's from a colo server mind you, and it's pretty "close" by way of peering somewhere nearby upstream to the google DNS, but it's still many orders of magnitude longer than the local DNS, which is in the same DC (hence the sub 1ms response).

[$ME@mercury ~]$ ping -c 2 8.8.8.8
PING 8.8.8.8 (8.8.8.8): 56 data bytes
64 bytes from 8.8.8.8: icmp_seq=0 ttl=49 time=16.375 ms
64 bytes from 8.8.8.8: icmp_seq=1 ttl=49 time=15.941 ms
 
--- 8.8.8.8 ping statistics ---
2 packets transmitted, 2 packets received, 0.0% packet loss
round-trip min/avg/max/stddev = 15.941/16.158/16.375/0.217 ms
[$ME@mercury ~]$ ping -c 2 216.46.1.2
PING 216.46.1.2 (216.46.1.2): 56 data bytes
64 bytes from 216.46.1.2: icmp_seq=0 ttl=253 time=0.349 ms
64 bytes from 216.46.1.2: icmp_seq=1 ttl=253 time=0.335 ms
 
--- 216.46.1.2 ping statistics ---
2 packets transmitted, 2 packets received, 0.0% packet loss
round-trip min/avg/max/stddev = 0.335/0.342/0.349/0.007 ms
 

d4m1r
join:2011-08-25

d4m1r

Member

I can't take namebench seriously :/

Ran it yesterday it told me 295.ca's (dialup company) DNS was 7% faster than OpenDNS....
Walter Dnes
join:2008-01-27
Thornhill, ON

Walter Dnes

Member

said by d4m1r:

I can't take namebench seriously :/

Ran it yesterday it told me 295.ca's (dialup company) DNS was 7% faster than OpenDNS....

Since most DNS queries are 512-byte UDP packets, it's perfectly possible if you're dealing with a local query on 295 versus a long-distance query to Google.

cork1958
Cork
Premium Member
join:2000-02-26

cork1958 to d4m1r

Premium Member

to d4m1r
Personally, and I'm not a TSI customer, but OpenDNS and Googles DNS servers are so far from the top of the list, for me, using dns benchmarking tools, I wouldn't use either in some one's computer that I WANTED to slow down on purpose!!

When it comes to DNS servers, the ONLY thing you can do is experiment and not go by what some one else is using.

FWIW,
Here's the 2 that work best for me.
69.54.70.15 69.7.192.2
priyen
join:2009-12-23
North York, ON

priyen to BikeHelmet

Member

to BikeHelmet
Thanks, helped a lot. Is there anyway i can "test" if there is a performance increase or anything?
BikeHelmet
join:2010-04-15

BikeHelmet

Member

said by priyen:

Thanks, helped a lot. Is there anyway i can "test" if there is a performance increase or anything?

Not really. All of our connections are subject to congestion and other line issues, so if there is a difference, there's no way to know if it's because of the DNS changes or just how good your internet is on that given day.

It mostly comes down to what you notice - and I notice a lot less 20 second delays while it resolves hosts. But to graph that I'd have to run some sort of benchmarking software 24/7 for several weeks.

pstewart
Premium Member
join:2005-10-12
Peterborough, ON

pstewart to drjp81

Premium Member

to drjp81
said by drjp81:

I'd get my facts straight before saying stuff like that. It's a mirroring service that has nothing to do with Teksavvy, really. And they don't do it for free. These people do it in order to gather usage statistics to sell to marketing firms, among other things. So if privacy is a concern keep using your third party DNS.

»en.m.wikipedia.org/wiki/ ··· nologies

The OP is correct and as some others have pointed out, please make sure you understand how CDN's like Akamai work before posting bogus information.

3rd party DNS will work at some point in the future with Akamai's on-net node clusters .. but it's not there yet. So for maximum performance to sites powered by Akamai (and other CDN's) you should use your ISP's DNS servers.
pstewart

pstewart to cork1958

Premium Member

to cork1958
said by cork1958:

Personally, and I'm not a TSI customer, but OpenDNS and Googles DNS servers are so far from the top of the list, for me, using dns benchmarking tools, I wouldn't use either in some one's computer that I WANTED to slow down on purpose!!

When it comes to DNS servers, the ONLY thing you can do is experiment and not go by what some one else is using.

FWIW,
Here's the 2 that work best for me.
69.54.70.15 69.7.192.2

Further to the OP's original comments - if you want to see what various DNS servers work well with CDN's like Akamai then just do a traceroute to www.akamai.com

If you're using your ISP's DNS servers and if they have on-net Akamai clusters you should see something like a few hops away at most and a few milliseconds typically. If using 3rd party, it will vary greatly depending on which public servers you are using.

Example (on-net cluster):

traceroute to www.akamai.com (216.168.105.87), 30 hops max, 60 byte packets
1 dis1-rtr-mb-vl10.nexicom.net (216.168.115.177) 0.241 ms 0.224 ms 0.209 ms
2 xe0-0-0.dis2.millbrook1.nexicom.net (98.124.46.18) 0.238 ms 0.242 ms 0.233 ms
3 xe5-3-0-1.core2.toronto1.nexicom.net (98.124.49.234) 1.954 ms 1.954 ms 1.954 ms
4 a---87.deploy.akamaitechnologies.com (216.168.105.87) 5.020 ms 5.000 ms 5.494 ms

Example (3rd party DNS - in this case OpenDNS)

traceroute to www.akamai.com (204.245.162.42), 30 hops max, 60 byte packets
1 dis1-rtr-mb-vl10.nexicom.net (216.168.115.177) 0.250 ms 0.228 ms 0.213 ms
2 xe2-3-0.core1.toronto1.nexicom.net (98.124.46.201) 2.403 ms 2.420 ms 2.447 ms
3 xe-7-0-0-101.tor10.ip4.tinet.net (77.67.71.81) 2.440 ms 2.446 ms 2.454 ms
4 xe-9-1-0.chi10.ip4.tinet.net (89.149.187.78) 13.850 ms 13.841 ms 13.828 ms
5 xe-0.equinix.chcgil09.us.bb.gin.ntt.net (206.223.119.12) 14.817 ms 18.079 ms 14.788 ms
6 ae-4.r23.nycmny01.us.bb.gin.ntt.net (129.250.2.41) 84.141 ms 68.994 ms 68.798 ms
7 ae-1.r20.asbnva02.us.bb.gin.ntt.net (129.250.2.9) 73.290 ms 70.870 ms 74.532 ms
8 ae-1.r02.asbnva02.us.bb.gin.ntt.net (129.250.2.183) 44.845 ms 34.204 ms 41.141 ms
9 a204-245-162-42.deploy.akamaitechnologies.com (204.245.162.42) 34.530 ms 37.888 ms 41.238 ms

JC_
Premium Member
join:2010-10-19
Nepean, ON

JC_ to d4m1r

Premium Member

to d4m1r
I'm using the following DNS servers ;
8.8.8.8 - Google
4.2.2.1 - Level3
208.67.222.222 - OpenDNS
206.248.154.22 - TSI
206.248.154.170 - TSI

Just default settings without TSI DNS.
Tracing route to a152.g.akamai.net [216.246.75.145]
over a maximum of 30 hops:
 
  1    <1 ms    <1 ms    <1 ms  192.168.1.1
  2    14 ms     6 ms     7 ms  10.124.0.1
  3    16 ms    16 ms    19 ms  69.63.242.25
  4     8 ms     7 ms     7 ms  fallowfield1.cable.teksavvy.com [24.246.55.22]
  5    14 ms    16 ms    15 ms  fallowfield2.cable.teksavvy.com [69.196.175.185]
  6    18 ms    30 ms    16 ms  tge11-3.fr4.yyz.llnw.net [208.111.134.241]
  7    29 ms    24 ms    33 ms  ve5.fr3.yyz.llnw.net [69.28.171.141]
  8    16 ms    18 ms    20 ms  torix.tge2-3.ar1.tor1.ca.nlayer.net [206.108.34.53]
  9    39 ms    42 ms    54 ms  xe-0-1-1.cr1.mtl1.ca.nlayer.net [69.22.142.80]
 10    32 ms    32 ms    36 ms  xe-10-1-0.cr1.nyc2.us.nlayer.net [69.22.142.109]
 11    32 ms    36 ms    40 ms  vlan-77.ar1.nyc3.us.nlayer.net [69.31.34.132]
 12    31 ms    31 ms    29 ms  unknown.scnet.net [216.246.75.145]
 
Trace complete.
 

With commands without TSI DNS.
Tracing route to a152.g.akamai.net [131.103.137.41]
over a maximum of 30 hops:
 
  1    <1 ms    <1 ms    <1 ms  192.168.1.1
  2     8 ms     7 ms     7 ms  10.124.0.1
  3    16 ms    18 ms    19 ms  69.63.242.25
  4     8 ms     9 ms     8 ms  fallowfield1.cable.teksavvy.com [24.246.55.22]
  5    27 ms    17 ms    27 ms  fallowfield1.cable.teksavvy.com [24.246.55.17]
  6    43 ms    77 ms    28 ms  tge11-3.fr4.yyz.llnw.net [208.111.134.241]
  7    17 ms    15 ms    15 ms  xe-2-3-0-350.tor10.ip4.tinet.net [77.67.79.65]
  8    32 ms    32 ms    33 ms  xe-11-0-0.chi10.ip4.tinet.net [89.149.182.89]
  9     *        *        *     Request timed out.
 10    40 ms    39 ms    40 ms  ae-1.r05.chcgil09.us.bb.gin.ntt.net [129.250.2.25]
 11    39 ms    38 ms    39 ms  dialup-131.103.137.41.omnilec.com [131.103.137.41]
 
Trace complete.
 
----------

With TSI DNS and no commands
Tracing route to a152.g.akamai.net [184.84.239.17]
over a maximum of 30 hops:
 
  1    <1 ms    <1 ms    <1 ms  192.168.1.1
  2     9 ms     5 ms     8 ms  10.124.0.1
  3    24 ms    18 ms    19 ms  69.63.242.25
  4     7 ms     7 ms     7 ms  fallowfield1.cable.teksavvy.com [24.246.55.22]
  5    15 ms    16 ms    14 ms  fallowfield1.cable.teksavvy.com [24.246.55.17]
  6    14 ms    14 ms    19 ms  Paix-tor.netarch.akamai.com [198.32.181.127]
  7    19 ms    17 ms    41 ms  a184-84-239-17.deploy.akamaitechnologies.com [184.84.239.17]
 
Trace complete.
 

With TSI DNS and all commands
Tracing route to a152.g.akamai.net [184.84.239.17]
over a maximum of 30 hops:
 
  1    <1 ms    <1 ms    <1 ms  192.168.1.1
  2    18 ms    18 ms     6 ms  10.124.0.1
  3    17 ms    19 ms    23 ms  69.63.242.25
  4     7 ms     6 ms     9 ms  fallowfield2.cable.teksavvy.com [69.196.175.186]
  5    16 ms    25 ms    14 ms  fallowfield1.cable.teksavvy.com [24.246.55.17]
  6    19 ms    15 ms    17 ms  Paix-tor.netarch.akamai.com [198.32.181.127]
  7    24 ms    16 ms    14 ms  a184-84-239-17.deploy.akamaitechnologies.com [184.84.239.17]
 
Trace complete.
 

Guspaz
Guspaz
MVM
join:2001-11-05
Montreal, QC

1 recommendation

Guspaz to d4m1r

MVM

to d4m1r
I wonder if using third-party DNS would bypass TekSavvy's google caching server, which seems to be responsible for TekSavvy's neverending youtube performance issues...

Those caching servers might save TekSavvy bandwidth, but being stuck buffering all the time isn't very fun.
timo
join:2010-03-17
Nepean, ON

timo

Member

said by Guspaz:

but being stuck buffering all the time isn't very fun.

Agreed.

d4m1r
join:2011-08-25

d4m1r to Guspaz

Member

to Guspaz
I use OpenDNS and have NEVER had buffering issues with youtube so maybe?

JC_
Premium Member
join:2010-10-19
Nepean, ON

JC_ to Guspaz

Premium Member

to Guspaz
said by Guspaz:

I wonder if using third-party DNS would bypass TekSavvy's google caching server, which seems to be responsible for TekSavvy's neverending youtube performance issues...

Those caching servers might save TekSavvy bandwidth, but being stuck buffering all the time isn't very fun.

With the DNS servers that I use I've never had any issues with playing videos off Youtube.

8.8.8.8 - Google
4.2.2.1 - Level3
208.67.222.222 - OpenDNS
206.248.154.22 - TSI
206.248.154.170 - TSI