Hacker News Reader

Comment by Scoundreller | Hacker News Reader

It's kinda a strange thing in Canada.

The incumbent telecoms generally refuse to peer domestically with independent ISPs and smaller transit providers that cheaper DCs use.

One ISP reported that Bell would only connect to them in Equinix in Chicago, even though Equinix has a PoP in Toronto. https://www.peeringdb.com/net/1550

This means if you subscribe to internet with the local telephone monopoly (Bell) in Toronto, your traffic to toronto[1-8].voip.ms goes through NYC or Chicago. Have seen some routes that add Washington DC to the mix too. A 5km trip made into several thousand km.

The incumbents even don't peer much at our big IX: http://torix.ca (they buy ports, but don't peer freely).

In a lot of ways, because the independents will peer with everyone and love IXs, your traffic often stays a lot more local in Canada.

View on HN

The US is the same. All major ISPs refuse to peer with any independents at all; refuse to be on routeservers; and demand some minimum of like 50-100 Gbps for a peering session, with insane requirements.

I've seen numerous ones also demand things like RPKI/ROA and good filtering - while themselves having everything ROA and IRR invalid, announcing customer space with no IRR records whatsoever, etc.

So how do independents interconnect their routes?

I think the Canadian issue is that incumbents aren't even peering with transit providers.

For this context, when I say independents, I mean smaller providers - content providers, independent ISPs that aren't large enough yet. They will obviously peer with Google, Akamai, etc.

Independents more or less just buy transit/blend and deal with it.

Here's an example of a large incumbent's peering policies: https://www.spectrum.com/content/dam/spectrum/residential/en...

The first requirement completely removes the possibility of ever getting a peering session with a small ISP or independent provider. So does the fourth.

Another one: https://www.xfinity.com/peering

> minimum of four mutually agreeable geographically diverse points in the US. Interconnection points must include at least one city on the US east coast, one in the central region, and one on the US west coast

This makes "launch an ISP from garage" bordering on "never" because they do not want regional providers getting settlement free peering - they just want them to buy IP transit from them.

If you look at virtually every major US/CA ISP on peeringdb, they will be Restrictive / Selective or Private peering only, most likely.

Does enough independent isp demand exist to aggregate the traffic to provide breakouts and satisfy Spectrum's 8 point term?

This reminds me of writing a bit to catch the BT SIN Supplier Information Notes ephemeral notices: BT and later OpenReach were required by law to advertise every sweetheart deal to all comers on this domain. This might take place for only long enough for the page to come and go from the google cache (which BT had some way of automating content exclusion from, the cache page link in search results leading to a 404 was a dirty little trick to claim the advertising and universal offering components of statutes, which didn't last long for a ruse because the entire uk industry seemed to automatically file in court to obtain even the simplest of provisions, I know we weren't alone) anyway that was how we discovered BT would lease twisted copper pairs from CPE to exchange and you could hop onto commercial backhaul from there, so we had 4 wire SHDSL modems either end and the tent was a enormous £495pq British pounds or about $3kpa for a symmetric 10Mb/s last mile to clec

> Does enough independent isp demand exist to aggregate the traffic to provide breakouts and satisfy Spectrum's 8 point term?

You're very close to describing standard IP transit. The regional-ness of small ISPs and indies make it such that interconnections throughout the entire US make it infeasible, especially considering the cost of such numerous transport and XCs.

I can get a blend from the DC - they aggregate enough capacity to peer with Telus for example, so the ASPath would go me -> DC -> Telus over IX or me -> DC -> Transit -> Telus.

If I got on the exchange directly, Telus peers selectively and does not do routeservers, so my traffic to them would not go over the exchange anyway, so what is the point?

Another annoyance is that in many facilities cross connects cost more than bandwidth itself - I get spam from Cogent daily offering cheap bandwidth, but the XC monthly cost costs more than several gigabit of Cogent bandwidth, so why bother unless I need a 100G port at once?

What are the major iSP's incentives for not peering?

I am starting to think may be Government should run a minimal Ix every x km2 and force all ISP to connect with it.

> What are the major iSP's incentives for not peering?

In the US, each peering agreement is a private agreement (ie. a contract) between 2 parties, usually with roughly equal bandwidth traffic.

If one party doesn't like you, perceives you as a competitor, or you don't have similar bandwidth, then no deal.

Examples of typical rejections are Cogent, who compete on price, and Netflix, who need last-mile acess to consumers.

https://en.wikipedia.org/wiki/Peering

For Canada, specifically Bell, they both an eye ball network and a transit provider. Bell has no incentive to peer at local IXs and "give away" part of their service (the eyes, as that is the lucrative part) when they can sell on the fact they are a large national network AND you can get access to their end customers as well.

Edit: clarification on the part that is more valuable.

Reply on news.ycombinator.com