 roccaStart.caPremium
join:2008-11-16
London, ON
kudos:12 | reply to Davesnothere
Re: HOW to Do It You have to remember that latency to a particular path includes _all_ the hops to that point, not just the latency _of_ that point.
So when you show hop 5 and say "that's odd", there are two reasons for seeing 101ms there, a) your computer did something at that moment, ie a background update, a web page ad refreshed, your instant messenger got or checked for a message, your mailbox downloaded, etc. or b) that router doesn't prioritize packets to its control plane (ie routers are designed to pass packets, not respond to requests directly against the router - packets destined _to_ a router are called control plane packets, whereas packets destined _through_ a router are traffic planes and have _very_ different handling). So we continue along and you see hop 6 and a dropped packet and probably think "dropped packet, must be a problem there" ...however again the control plane on that router probably limits how many responses that target that router are processed. Continuing you see hop 7 and say "oh good, everything is normal from here" but you have to remember that getting to hop 7 means that exact packet that replied went through all the other hops (and back again) that you were concerned about.
In the end, traceroute is good for seeing where there is a problem if the end node is responding poorly, ie if you saw 100+ for _all_ the hops after 5, then you could conclude hop 5 was a problem, but the fact that the end point is quite fine, then therefore the entire path must be quite fine also despite what a control plane on a specific router in the middle of the path said. |
|
DavesnothereNo-BHELL-ity DOES have its Advantages
join:2009-06-15
START&Cogeco
kudos:6 | said by rocca:You have to remember that latency to a particular path includes _all_ the hops to that point, not just the latency _of_ that point.
So when you show hop 5 and say "that's odd", there are two reasons for seeing 101ms there....
Right, so I ought to repeat the test to rule out that sort of stuff.
But Hop 11 seems to be a consistent slowdown ?
BTW, did you re-read my last post ? - I revised it. |
|
roccaStart.caPremium
join:2008-11-16
London, ON
kudos:12 | reply to Davesnothere
said by Davesnothere:That's been told to me occasionally before, but it can be hard to take to heart sometimes. I wouldn't lie to you. 
Here's a recent story from our rural wireless side. We have a customer about a 1/2 km from one of our towers and wants to upgrade their speed, but the path is poor (tree in the fresnel) and that path can't support higher speeds. So we setup a dedicated link from another tower about 5km away but has a different path. The customer is initially very concerned "How can a link 5km away be faster than the one a tenth of that distance" - valid 'gut' feel, but that's not the way networks work, particularly when packets are travelling at roughly the speed of light. The extra hop on the router adds about 0.5ms latency and the customer gets full speed.
Moral of the story: the shortest physical distance between A and B is often not the most optimal route.
Hope that helps....
...and now I'm done for the day, hockey tournament.  |
|
|
|
DavesnothereNo-BHELL-ity DOES have its Advantages
join:2009-06-15
START&Cogeco
kudos:6
1 edit | reply to Davesnothere
Nexicom Retested
Tracing route to nexicom.com [216.168.96.37]
over a maximum of 30 hops:
1 - 1 ms 1 ms 1 ms 192.168.67.1
2 - 9 ms 12 ms 7 ms 10.205.36.1
3 - 42 ms 43 ms 43 ms asr-link1-burlington1.net.start.ca [64.140.112.154]
4 43 ms 43 ms 44 ms core1-toronto1-te3-3.net.start.ca [64.140.112.153]
5 - 43 ms 43 ms 44 ms gi2-27.ccr01.yyz03.atlas.cogentco.com [38.122.70.81]
6 - 42 ms 42 ms 54 ms te7-5.mpd02.yyz02.atlas.cogentco.com [154.54.45.117]
7 - 47 ms 43 ms 44 ms te0-3-0-0.ccr22.yyz02.atlas.cogentco.com [154.54.40.165]
8 - 56 ms 53 ms 55 ms te0-3-0-6.ccr22.jfk02.atlas.cogentco.com [154.54.47.129]
9 - 59 ms 56 ms 56 ms te0-5-0-0.ccr21.jfk07.atlas.cogentco.com [154.54.80.178]
10 - 54 ms 57 ms 55 ms tiscali.jfk07.atlas.cogentco.com [154.54.10.90]
11 - 71 ms 74 ms 71 ms xe-4-3-0.tor10.ip4.tinet.net [89.149.184.110]
12 - 89 ms 75 ms 79 ms xe-4-2-0-101.core1.toronto1.nexicom.net [77.67.71.82]
13 - 77 ms 75 ms 75 ms dis1-rtr-mb-ge8-15-vl3.nexicom.net [76.75.120.81]
14 - 83 ms 80 ms 80 ms dis2-rtr-mb-ge9-5.nexicom.net [216.168.98.137]
15 - 76 ms 79 ms 78 ms ge6-0-23.dis3.millbrook1.nexicom.net [98.124.59.130]
16 * * * Request timed out.
17 * * * Request timed out.
18 * ^C
Hop 5 looks better now.
I guess it's just the sheer number of hops in the Cogent and other peerings to there.
They cannot ALL be invisible. |
|
roccaStart.caPremium
join:2008-11-16
London, ON
kudos:12 | reply to Davesnothere
Re: HOW to Do It said by Davesnothere:I don't think I'll find the 'Phantom Ping' It's at hop 3. |
|
roccaStart.caPremium
join:2008-11-16
London, ON
kudos:12 | reply to Davesnothere
said by Davesnothere:But Hop 11 seems to be a consistent slowdown ? That's a question for tinet. |
|
DavesnothereNo-BHELL-ity DOES have its Advantages
join:2009-06-15
START&Cogeco
kudos:6 | reply to rocca
said by rocca:....Moral of the story: the shortest physical distance between A and B is often not the most optimal route.
Hope that helps....
...and now I'm done for the day, hockey tournament.
Yep, I get it.
Have fun there, and BTW, you should offer that analogy to the hockey players regarding the PUCK and scoring.  |
|
DavesnothereNo-BHELL-ity DOES have its Advantages
join:2009-06-15
START&Cogeco
kudos:6 | reply to rocca
said by rocca:said by Davesnothere:I don't think I'll find the 'Phantom Ping' It's at hop 3.
Yes, but I meant the reason for it.
WHAT ? - Not gone yet ?
EDIT : I didn't mean 'Gone', just gone |
|
 GonePremium
join:2011-01-24
Fort Erie, ON
kudos:3
 ·Start Communicat..
| reply to rocca
said by rocca:In any case, the new 20ms latency that popped up last week is being investigated but as everyone seems to be reporting good speeds and no impact, 20ms (we are talking milliseconds here, ie 1/50th of a second) really shouldn't affect day-to-day in the meantime. Browsing and downloading, no.
Ask me the same thing when I'm trying to play Team Fortress, though. |
|
DavesnothereNo-BHELL-ity DOES have its Advantages
join:2009-06-15
START&Cogeco
kudos:6
1 edit | reply to Gone
Re: Speed Tests, April 13th, Near Peak said by Gone:said by Davesnothere:The Toronto test server responds just fine when I'm on my Cogeco account and modem - in fact the lowest ping of them all.
Did I miss something obvious ? Yes. Cogeco has drastically different peering arrangements than Start. Cogeco doesn't peer at Torix, for example. Who knows what kind of route you're getting to get to that server, or what or what port you're using if it's multihomed. You're posting decent speed tests from other servers, so who the hell cares if but one server has lower performance than the other ones you're posting? Seriously, when it comes to stuff like this there are so many variables that were this conversation in person it would be nothing more than an exercise in oxygen/CO2 exchange.
But I THRIVE on learning stuff, even at my advanced age.
The tests I have performed since you posted that last post have illustrated very well the differences in peering among the 2 test sites with START, and so just for kicks, before Cogeco unprovisions me, I'm going to repeat those 2 cities thru Cogeco account to see what peering that THEY use.
BTW, Peter posted something which explained why the Toronto test reads poorly.
Even though they both peer at TorIX, Nexicom has not yet turned on peering to Start thru TorIX, but plans to soon. |
|
GonePremium
join:2011-01-24
Fort Erie, ON
kudos:3 Reviews:
·Start Communicat..
| said by Davesnothere:Nexicom has not yet turned on peering to Start thru TorIX, but plans to soon, apparently. I know, I read that when rocca said that.
If that speed test server needs to be routed through Cogent, that explains why the speed tests are out of whack relative to the actual performance, at least based on my own past experiences. Once they're talking to each other directly at Torix you'll see it improve drastically. |
|
DavesnothereNo-BHELL-ity DOES have its Advantages
join:2009-06-15
START&Cogeco
kudos:6 |
Lurking then, I see.
So it's not that Montreal was that much better in my tests, it was the Toronto test that was out of range, and the other user noticed that too, when comparing Toronto to Michigan.
A-HAH !
And there sure were a lot of hops on the way to Nexicom from START !  |
|
DavesnothereNo-BHELL-ity DOES have its Advantages
join:2009-06-15
START&Cogeco
kudos:6 | Cogeco to Nexicom in Toronto
Tracing route to nexicom.com [216.168.96.37]
over a maximum of 30 hops:
1 - 1 ms 1 ms 1 ms 192.168.67.1
2 - 7 ms 7 ms 7 ms 10.71.128.1
3 - 22 ms 19 ms 19 ms d226-5-81.home.cgocable.net [24.226.5.81]
4 - 18 ms 27 ms 19 ms d226-6-130.home.cgocable.net [24.226.6.130]
5 - 24 ms 17 ms 21 ms xe4-3-0-56.core1.toronto1.nexicom.net [206.108.34.66]
6 - 19 ms 19 ms 19 ms xe3-3-0.core2.toronto1.nexicom.net [98.124.50.241]
7 - 33 ms 20 ms 22 ms ge1-2-9.dis2.millbrook1.nexicom.net [98.124.46.162]
8 - 24 ms 22 ms 23 ms ge6-0-20.dis3.millbrook1.nexicom.net [98.124.50.2]
9 * * * Request timed out.
10 * * * Request timed out.
11 ^C
It's a quick hand-off to Nexicom, et VOILA !
Deux Points ! |
|
DavesnothereNo-BHELL-ity DOES have its Advantages
join:2009-06-15
START&Cogeco
kudos:6 | Cogeco to Fibrenoire in Montreal
Tracing route to fibrenoire.net [208.84.104.2]
over a maximum of 30 hops:
1 - 1 ms 1 ms 1 ms 192.168.67.1
2 - 14 ms 8 ms 7 ms 10.71.128.1
3 - 20 ms 22 ms 19 ms d226-5-81.home.cgocable.net [24.226.5.81]
4 - 25 ms 20 ms 19 ms d226-6-130.home.cgocable.net [24.226.6.130]
5 - 18 ms 19 ms 20 ms bvi1.asr01.tor151f.fibrenoire.ca [206.108.34.141]
6 - 25 ms 25 ms 28 ms te4-3.crt01.mtlcx03.fibrenoire.ca [68.67.63.197]
7 - 26 ms 25 ms 25 ms gi1-19.core02.mtl-1250.fibrenoire.ca [208.88.108.226]
8 - 26 ms 25 ms 27 ms srv01.mtl-1250.fibrenoire.ca [208.84.104.2]
Trace complete.
Very similar peering.
One network transfer and done. |
|
Reviews:
 ·Cogeco Cable
 ·voip.ms
·Start Communicat..
| reply to Davesnothere
Re: Cogeco to START - Changeover Logistics Quite awesome today.
Why would it be GRADE B? |
|
GonePremium
join:2011-01-24
Fort Erie, ON
kudos:3 | Latency is a bit high and the speed is still low in comparison to the fastest stuff out there. That's why it's B rather than A. |
|
DavesnothereNo-BHELL-ity DOES have its Advantages
join:2009-06-15
START&Cogeco
kudos:6
1 edit | reply to Last Parade
Anticipation said by Last Parade:said by Davesnothere:
Is Last Parade up and running on START yet, with the fancy D3 Cisco beast ?
No, I can't wait for my 9 hour install window on the 18th though.
Ironic that the only one of us GOUGEco lemmings with a proper Kosher D3 modem has to endure latency in his INSTALL DATE !
4 more Sleeps, then....
--
We have only 2 things about which to worry :
(1) That things may never get back to normal
(2) That they already HAVE !
|
|
DavesnothereNo-BHELL-ity DOES have its Advantages
join:2009-06-15
START&Cogeco
kudos:6 | reply to spikernum1
Re: Cogeco to START - Changeover Logistics said by spikernum1:Quite awesome today. - Why would it be GRADE B?
One of my best with START was almost exactly as yours, at 2 AM today to Montreal, and was also judged that way.
I'm with Gone on that. |
|
Reviews:
·Cogeco Cable
·voip.ms
·Start Communicat..
| reply to Davesnothere
I'm not at home today, but my wife is complaining about pretty poor connection when streaming YouTube. I told her to unplug the modem and router and try again. Still had complaints after.
Just wondering if anyone else is getting slower d/ls right now. |
|
DavesnothereNo-BHELL-ity DOES have its Advantages
join:2009-06-15
START&Cogeco
kudos:6 |
Try other stream sites, as may be just YT.
Speed Tests are anywhere from 7 to 12 Mb on various servers right now.
Pings R similar to our other recent tests.
Specific DLs ? - not done any today so far. |
|
