Kansas? Canada? Congo? Where should your next datacenter be built?
While technicians most often discuss siting in terms of energy availability, network geometry, and perhaps workforce skills, business decisions at the highest levels more often focus on government subsidies, legal regimes and occasionally organizational strategy. Here’s a bit of the context and trends in datacenter geography:
Not just the US coasts
Energy is a major expense in datacenter operation, if not as dominant as widely believed. And there are real reasons to site racks next to financial trading centers, or such concentrations of inexpensive talent as colleges.
Increasingly, though, datacenters are planned and managed as other industrial properties, that is, with considerable regard to government subsidies and cultural factors. Kansas City, Omaha, and Des Moines, for instance, boast combinations not only of fiber-optic-based broadband achieving gigabit-per-second connections at negligible prices, but “entrepreneurial density“, modest costs-of-living, a useful range of technical colleges and universities, and “people that will work like they worked on the farm.”
Moves even farther from traditional computing centers in Silicon Valley and the Coasts more generally also make sense in many situations. The Data Centre Risk Index puts Canada just behind the US as the two countries with the best combination of factors favorable to datacenter operation. Canadian energy prices are often lower than those in the US, water (important in some designs for cooling) is abundant, the workforce skilled, and natural and man-made disasters infrequent. For many organizations, the contrast between the Canada Privacy Act and the US Patriot Act makes north-of-the-border locations compelling. Finally, Canada is a large enough market itself to warrant its own datacenters, if only to “… save milliseconds of customer load and wait times …”
Plenty has already been written on where to put, or not put, new datacenters. The trend that I see as strongest now is convergence at the business level of planning for datacenters and other industrial installations. “What is the price of electricity?” is becoming less important than, “What incentives can we contract?”, “How predictable and transparent is local government?”, “Is a skilled workforce available?”, and “What strategic reasons are there to be in this location?” For large-enough organizations, that almost certainly means spreading out datacenters, to minimize coincidental natural disasters and maximize flexibility in dealing with taxing and other legal authorities.