Paul Miller: Price cuts are nice, but certainly not the whole story
Sun 20 Apr 2014

The leading providers of cloud infrastructure appear committed to following one another in a race to the bottom on pricing. Customers should welcome the latest round of price cuts says Paul Miller but also recognise that there’s more to choosing a cloud provider than what it costs.
Microsoft did as everyone expected earlier this week, joining Google and Amazon in the latest round of cloud price cuts. The headline figures from all three are certainly impressive, with prices dropping as much as 70% in some areas, but each has the scale and efficiency for plenty more of the same. This race to the bottom is far from over, although some of the industry’s smaller players will find their margins squeezed uncomfortably thin if they try to compete on price alone.
And yet, despite buyers’ quite understandable love of a bargain, price is far from the only consideration when selecting a cloud provider for the long haul. As public cloud infrastructure becomes an ever-more important piece of good IT strategy, it’s increasingly impractical to shift workloads in never-ending pursuit of today’s lowest price.
With short bursts of computing activity (processing a set of images, analysing a new data set, spinning up a VM to test some code, etc.), there’s a lot to be said for shopping around to locate the cloud which might be cheapest, quickest or greenest right this second. A growing proportion of today’s business use of the cloud bears no relation to these short and self-contained bursts of activity, though.
Instead, they are increasingly running complex applications in the cloud for the long haul. These applications are tightly connected back to on-premise systems with virtual private networks, secure networking tunnels, and even dedicated physical connections between data centres. The cloud providers themselves are layering ever-richer features and services on top of their basic compute and storage offerings, and every efficiency gain or cost saving derived from a customer’s use of these must be weighed against a corresponding increase in the pain of moving elsewhere.
You can spin up a virtual machine on any cloud, and expect it to behave in a pretty similar fashion. You can store your data on any cloud, and expect it to be there when you need it. But once you start integrating queuing services and data warehouses and analytics services, the differences in approach very quickly outpace any similarities. You could still move your computing from one provider to another, but it’s becoming a pretty major undertaking.
Not the sort of thing you’d embark upon, just to chase a short-term cost saving which will probably get undercut tomorrow.