We're looking to extend some services into Canada. While our lawyers dig into it, I thought that I'd ask the hive mind about border restrictions.
For traffic routing, is anyone constraining cross-border routing between Canada and the US? IOW, if you are routing from Toronto to Montreal, do you have to guarantee that the path cannot go through, say, Syracuse, New York?
I'm asking network operators about packet routing; data storage is a very different matter, of course.
Guarantee to whom?
Back a few years ago when I looked into it, most of the traffic within
Canada went through the US, e.g., since Bell didn't want to peer with
anyone in Canada, you'd go something like YYZ - ORD - YYZ, clearly
visible through the traceroute.
Possibly somewhat better nowadays — there's been quite a few new IX
POPs that popped up — but I doubt the scenario is a thing of the past.
P.S. Just for the giggles — checked http://lg.he.net/ routing from
core1.tor1.he.net to www.bell.ca — still goes through Chicago, to
Montreal, from Toronto. Going straight to Montreal,
core1.ymq1.he.net, will route you to www.bell.ca (still in Montreal)
through the peering at NYC.
P.P.S. In other words — if someone wants guarantees, they better
explicitly ask you for it.
No. In fact, Bell Canada / Bell Aliant and Telus guarantee that you _will_ go through Chicago, Seattle, New York, or Ashburn, since none of them peer anywhere in Canada at all.
Last I checked (November of last year) the best-connected commercial networks (i.e. not CANARIE) in Canada were Hurricane Electric, MTS Allstream, Primus, and Zip Telecom, all of which peer at three or more Canadian IXes. So, they’re capable of keeping traffic in Canada so long as the other end isn’t on Bell or Telus, which only sell U.S. bandwidth to Canadians.
In November, only 27% of intra-Canadian routes stayed within Canada; 64% went through the U.S. That’s way worse than five years ago, when 60% stayed within Canada, and 38% went through the U.S.
As has been pointed out, Canada has been building IXPs… Just not as fast as the rest of the world has. They’re behind the global average growth rate, and behind the U.S. growth rate, which is why the problem is getting worse. Bandwidth costs are falling faster elsewhere, so they’re importing more foreign bandwidth.
What does Bell buying MTS do? Does it change your statement or will the MTS portion of Bell still peer locally?
I’d have to go back and look at the actual ASNs in our analysis. I think what we called “MTS Allstream” in the chart is actually the Zayo-owned Allstream, not the Bell-owned Bell MTS.
MTS Allstream is no longer a combined entity. MTS was purchased by Bell Canada and Allstream was purchased by Zayo.
With the peering policies of the major Canadian ISPs, you're virtually guaranteed to hairpin through the US on most paths.
Robellus (Rogers, Bell & Telus) will peer with you at any of their major Canadian peering points, such as NYC, Chicago or LA.
To be fair, Rogers does peer in Toronto. Along with New York, Chicago, Seattle, and Ashburn.
The major national networks (Bell, Rogers, Telus, Shaw, Zayo/Allstream) do peer with each other and some other large / old Canadian networks (e.g. MTS, SaskTel, Peer1) within Canada. While they do practice peering protectionism and only purchase transit out of country, the situation is not *quite* so bad that all traffic round-trips through the US.
Of course if neither side of the conversation has at least one of those major networks as a transit upstream - which is most of the eyeballs and most of the important Canadian content - you'll see that hop through Chicago or Seattle (or worse). Which is exactly the way they like it.
OK, Maybe I was a bit overly dramatic. One of the big 3 peered with us in a US location, but refused to peer in Canada.
I can't recall if we actually did specifically ask Rogers at one point or not. I know we haven't asked recently.
No, not all, 64%. By comparison, only 0.27% of intra-US traffic goes through Canada.
It is worth noting, however, that the former AllStream ASN (formerly AT&T
Canada) AS15290 is a completely different thing, and has distinct
infrastructure and routing from the AboveNet ASN which is operated by Zayo.
Although they are probably using "Free" Zayo transport by now.
If I am grossly wrong and anybody from layer 3 network operations at Zayo
wants to chime in and tell us about the 40,000 ft view of their plans to
combine AS15290 and AS6461, I am sure the community would be very
Generally speaking, when customers have concerns about their traffic
crossing borders, they do ask upfront.
As a multinational operator you can only guarantee traffic if customers
asks and offcours pays the fee for special class of service.
Allstream fiber goes counterclockwise from Toronto to Buffalo along the lake. Just like the rest of them. And at several places all these sysgtems are probably in the same conduit.
Finally, all fiber is exhausted Toronto/Buffalo. Existing players could not sell if they wanted to and no one selling dark on this route today.
I would bet that most British Columbia traffic gets routed to Vancouver>Seattle. Just a hunch, but I suspect that connectivity capacity across Canada from British Columbia to the Eastern part of the country is pretty limited.
I am not sure that this answers your question, but carriers are looking for more diversity on cross border traffic. Right now virtually all Toronto/NYC runs through Buffalo and 350 Main, Buffalo. May be Montreal/NYC has more options.
There have been huge network builds in the US Northeast by Unite and Firstlight, which are providing optical backhaul from cell towers. So more routing options may be available than in the past.
An issue that no one discussed in the age of the fiber. The big carriers and the Web Giants do not want to 1998 manufactured fiber. And some of the Web Giants are only putting 4x 100 gig waves per fiber pair. So the incentive to create new diverse routes with new large fiber builds is growing.
Sorta, kinda. The various ASs operated by Zayo are more interconnected than that description would imply. The traditional mode of operation on an "acquired AS" has been to turn down any upstream transit as quickly as contractually possible and upgrade NNI capacity between that AS and 6461 to compensate. Over time, legacy devices are overbuilt or replaced with ones directly on 6461. The net-net of it is that most traffic will end up egressing to other providers via 6461's peering after a fairly short period, although this isn't universally true, especially for "local" traffic (e.g. traffic originating on the Neo AS staying in France, etc.).
You mean ROBALLOFUS right?