A quick exchange tonight on Twitter with Steve Duplessie (@stevedupe) regarding latency and datacentre location got me thinking again about the whole datacentre location issue which I’d mused over some time ago.
It goes like this; given no latency restrictions, where in the world would you place your datacentres?
Clearly the choice would be based on a number of parameters:
- ready availability of (cheap) power (i.e. electricity)
- low cooling costs (Death Valley would probably be unsuitable)
- geographic stability (no earthquakes please)
- political stability (nobody reclaiming my datacentre as their property)
- Cheap land and build costs
- access to cheap and skilled labour to run the operation
It isn’t difficult to extrapolate these points and come up with a number of locations on the globe that would fit this criteria. Companies are already doing it, so how hard is the latency problem?
Firstly, we should ask where exactly does the latency occur? With decent telecomms, customer->server latency isn’t much of an issue, especially if the communication is web based. From a storage perspective, what’s more difficult is providing application/data resiliency in geographically disparate locations. We know about technologies like Axxana that can assist in the distance replication problem, and having an application->storage understanding helps design infrastructures that can provide simultaneous diverse access to data.
Perhaps the age of datacentres located wherever we choose isn’t that far away after all…