In 2011 the Canadian Internet Registration Authority commissioned research from the team at the Packet Clearing House (PCH) to examine patterns of Internet traffic movement across Canada. We were becoming increasingly concerned that Canadian Internet performance was being hurt by the slow adoption of Internet exchanges in Canada and wanted to see how the architecture of Canada's networks was affecting the movement of Canadian users' data.
The 2011 study provided us with the data that would support the construction of six new Internet exchanges in Canada, creating a viable path for Internet traffic to flow efficiently across the country with hubs in cities from Halifax to Montreal, to Calgary, to Vancouver.
Overall PCH analyzed a total of 2,207,228 traceroutes between 2013 and the present. The study found that routes between Canadian sources and Canadian destinations crossed into the US 64 per cent of the time. By comparison, less than one per cent of routes between US sources and US destinations crossed into Canada. Nine per cent of Canadian-to-Canadian traffic crossed into other third countries.
If we drill-down further and look at the flow of packets on a hop-by-hop basis, meaning taking into account each move they make and not just the overall trip, traffic between Canadian destinations spends 27 per cent of its time in the US. 14.2% of this traffic passed through three or more countries before returning to Canada.
So, why do we care about this? The Internet is global and interoperable. This is intentional right? If the best traffic route is through the US, this all makes sense, correct?
It's not that simple and, as we get into the possible reasons for these traffic patterns, the plot thickens. The PCH study points to three explanations for why these traffic patterns occur.
Firstly, the global routes between Canada and the US have been upgraded at a greater rate than domestic routes within Canada. When you want to move traffic between Canadian cities, the route with the most capacity may be through New York or Seattle. This is an investment choice.
Secondly, the nature of hub-and-spoke routing topology should favour the use of Canadian exchange points wherever possible. Generally, networks are economically advantaged by using bandwidth created at these exchanges. This is certainly true for smaller networks, because the cost savings come from limiting transit costs, however it is also true for larger ones as well. The difference is the relative importance of the economic benefit to smaller versus larger providers.
Finally (and this is where I get most frustrated) larger incumbent networks in Canada can essentially force their competitors to spend money on transit by peering at global exchanges. Small and medium sized ISPs in Canada need to pay the incumbent to route their customers' data to and from these global exchanges.
All of this adds cost and reduces domestic performance.
The good news is that we have the solution to this problem. Over the last three years CIRA has worked with the Canadian Internet community to facilitate the scaling-up of our national Internet exchange infrastructure.
We now face a new challenge. Canadian incumbents are not peering. PCH neatly summarized the state of peering at Canada's ISPs in this chart.
Essentially, if incumbent ISPs were to peer at Torix in Toronto, nearly every other smaller ISP would save on transit costs and could provide better services at lower price points.
I see this as a win/win. In an effort to provide their customers with the best-possible performance, ISPs should be peering at points across the country. More peering benefits Canada's Internet and we hope that this is something that all ISPs can agree on.
Let's not let short-sighted business decisions keep us from having the best possible Internet performance. We can do better than this. Both for the performance of the domestic internet ecosystem and for the benefit of end users.
An initial summary of the study is available on our website and we will post the complete report once it is published later this year. There is much more to the results, including a look at the locations of Canadian government websites. I encourage you to have a look.