New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Comments
Yep, comcast takes me through cogent as well
However, on my VPS w/ DigitalOcean, I go thru Telia.
What's utterly idiotic about these routes out of Denver is that:
Denver to San Jose is 1282 driving miles.
To get to Buffalo, NY via San Jose, CA, you must travel 1282 miles west, then 1282 miles back east = 2500 miles (maybe less with how fiber is ran, but still insane distance).
Denver to Buffalo is 1522 driving miles or 240 miles more than Denver to San Jose.
A packet from Denver to San Jose must travel 2500 miles just to get back to Denver + 1500 miles more for 4000 miles total. Doing so is well over double the distance of Denver to Buffalo directly.
And... as you might notice, Denver is probably < 200 miles from due center of the United States. Meaning routing east coast traffic via California is IDIOTIC.
Routing is messed up from Toronto and Montreal also...
1 216.187.120.170 (216.187.120.170) 0.469 ms 0.558 ms 0.455 ms
MPLS Label=299920 CoS=0 TTL=1 S=1
2 10ge.xe-1-0-0.tor-151f-cor-1.peer1.net (216.187.114.197) 7.803 ms 7.860 ms 7.824 ms
MPLS Label=299920 CoS=0 TTL=1 S=1
3 10ge.xe-0-2-0.tor-151f-dis-1.peer1.net (216.187.114.146) 7.658 ms 7.637 ms 7.657 ms
4 ae0-112.tor10.ip4.tinet.net (77.67.68.9) 7.739 ms 7.744 ms 7.682 ms
5 xe-0-0-0.chi12.ip4.tinet.net (89.149.187.82) 21.926 ms xe-1-0-0.chi12.ip4.tinet.net (89.149.187.90) 22.434 ms xe-0-0-0.chi12.ip4.tinet.net (89.149.187.82) 42.745 ms
6 chi-bb1-link.telia.net (213.248.82.249) 21.907 ms chi-bb1-link.telia.net (213.248.75.177) 25.532 ms chi-bb1-link.telia.net (213.248.82.249) 32.196 ms
7 cle-b1-link.telia.net (80.91.246.32) 29.281 ms 29.128 ms 29.245 ms
8 buf-b1-link.telia.net (80.91.246.35) 33.982 ms 34.035 ms 40.863 ms
9 giglinx-ic-155660-buf-b1.c.telia.net (213.248.96.42) 21.555 ms 34.505 ms 34.482 ms
Toronto and Buffalo are pretty close. 98 road miles.
That route goes:
Toronto --> Chicago --> Cleveland --> Buffalo.
I am not doing the air or road miles here, but far exceeds a local route it should be. 27ms for two end point 100 miles apart.
Okay, the migraine is going away - I don't feel like I'm in the twilight zone anymore.
Anything to East Coast from here goes... EAST and is 60-65 ms.
On the other hand... Because of the unique way CC picks up Telia, it hauls Denver, CO to Palo Alto through TATA, then joins up with Telia in SJ... When Telia is present / has a PoP in Denver.
Cheaper?
All of the bigger networks have peering requirements in order to establish a peering session with another network in a POP. If there isn't sufficient traffic to justify a peering session in that POP, the networks will backhaul it to the next closest POP where they do peer, or will hand it off to one of their transit providers to do the handoff (if they aren't settlement free).
While it would be nice if every network peered with every other network in every city where they were commonly located, that's just not the way the Internet works.
Nor does it work where a mid-west location hauls traffic to San Jose, CA to get the other end of the country. At least not on actual data center networks. Maybe Low end "data centers" - Aka cages with an ASN.
As much folks complain about HE+Cogent dual stack networks, like the famous one in Kansas City, you just don't see these routing to the wrong coast issues. Sure Cogent is known to overshoot and double back, but not 2500 miles worth.
I'd stop thinking of Colocrossing in Buffalo as a premium network based on these changes.
Afterall, as the ad said:
+1
nlayer also has strange/unstable routes.
from CC buffalo to iperweb/prometeus.
3 96.44.180.33.internal.quadranet.com (96.44.180.33) [AS22298] 0.270 ms
4 xe-5-0-1.ar1.lax2.us.nlayer.net (69.31.127.41) [AS26627] 2.303 ms
5 ae1-80g.cr1.lax1.us.nlayer.net (69.31.127.129) [AS26627] 0.210 ms
6 xe-2-2-1.cr2.ord1.us.nlayer.net (69.22.142.255) [AS4436] 48.768 ms
7 ae3-40g.cr1.ord1.us.nlayer.net (69.31.111.153) [AS4436] 48.850 ms
8 xe-5-0-0.cr1.lhr1.uk.nlayer.net (69.22.142.83) [AS4436] 140.921 ms
9 ae2-70g.cr1.nyc2.us.nlayer.net (69.31.95.174) [AS26627] 64.902 ms
10 xe-4-1-0.cr1.lhr1.uk.nlayer.net (69.22.142.63) [AS4436] 136.084 ms
11 te1-3.milano52.mil.seabone.net (195.22.196.183) [AS6762] 240.298 ms
12 cdlan.milano52.mil.seabone.net (195.22.192.3) [AS6762] 184.581 ms
13 217.171.46.241 (217.171.46.241) [AS20836] 168.397 ms
14 pm30.prometeus.net (192.71.245.3) [AS34971] 173.942 ms
it seems too busy router in UK send back to nyc?
RTT is also unstable.
(us and la lines are in CC Los Angels DC)
https://www.dropbox.com/s/5korg92e5irc1q3/ping-pinpoint=1359850522,1360541722.png
My URPad VPS did the same at around 7AM, but didn't come back up until I manually booted it in SolusVM.
It also could be that a power feed was overloaded. You may have 2 20 amp power strips, but in reality you can only safely load them to 9amps peak load. Because if you lose one side everything draws off the other side at full power. So that 1amp server is drawing a half amp across 2 feeds.
Jon passed me an RFO about the power outage since I had contacted him about the networking stuff.
Right from the horses mouth:
Thankfully we didn't get hit by this but other providers weren't as lucky.
Word on the network changes is that, at least for the Buffalo area, Cogent/Telia are giving better routes than L3. Comcast is loaded pretty heavy in a few spots and they're favoring Telia on those just to get around it. L3 is still in the mix but I'm assuming they're just taking carrier routes. They're planning on adding XO but I kinda wished they wouldn't. XO is terrible and the only place I know that actively uses them is like...Burst?
I can agree to L3 being dumb sometimes. It was common for them to backhaul things to Chicago before going anywhere.
I wish I had been given a little heads up but I can understand Jon being "a little" pissed off that I'm painting him and boy wonder Chris with the same brush.
Francisco
Adding them to the mix isn't an issue as long as it's used where it's better. This is the point of having more providers in the mix and being able to choose.
I have seen XO being in providers' mix besides Burst.
The vast majority of routing/network issues burst gets that land on WHT are XO related.
I've never had a good experience with them, especially back in the day when FDC had a lot of them. Granted it's FDC so you have to blame some of the crap on that factor, but still.
Francisco
XO has really stagnated. US main pops are a few (check out their looking glass).
Here's XO's asset map:
http://www.xo.com/SiteCollectionImages/about-xo/xo-network/maps/map_complete_1600.gif
Right now, CC is bubblegum fixing issues at best. Haven't checked for route snafus yet this morning. Hoping they've cleaned up from yesterday. Really bad random routes yesterday with providers upstream playing hot potato with the packets.
I saw all Level3 before to Buffalo before. Tracing from Comcast Florida.
From BGP.he.net, Level3 is now a part of the group that amounts for less than 3% of traffic...
Looks like someone "optimized" the routes yesterday... for maximum cost efficiency.