Do Cheap Power and Massive Data Centers Really Matter?

June 2, 2011 Off By David
Object Storage
Grazed from GigaOM.  Author: Derrick Harris.

GoGrid CEO John Keagy wrote on his blog Thursday that when it comes to cloud computing, there are a couple of things that have been overrated in the industry, like cheap hydroelectric power and massive-scale data centers. What Keagy says does matter for making cloud computing financially compelling to both providers and users are things like pay-per-use pricing, automation, shared platforms and commodity hardware. His theory doesn’t make sense for every cloud provider, but he does make some good points.

Actually, power and massive data centers aren’t the only factors he calls out as being over valued — Keagy also says focusing on data center containers, blade servers, super-efficient cooling and VMware has been over hyped — but they’re probably the most controversial. After all, few would argue that containers, blades and/or VMware are less expensive — or necessarily better-performing — than the current alternative measures. But huge data centers and cheap power have become hallmark concerns when talking about the economics of cloud computing, so Keagy is treading on hallowed ground when he questions their importance.

Here’s Keagy’s stance on power:

At GoGrid, power represents less than 5% of our cost of goods sold. We’re a nicely profitable company despite buying some of the most expensive power and cooling on the planet. Theoretically, we could knock a few points out of our COGS if we used a datacenter next to the Columbia River in Eastern Oregon. . . . But then we’d have new costs that we don’t have now such as the costs of managing people who are native to Eastern Oregon and paying people from Silicon Valley to travel to Eastern Oregon to manage those people and lots of networking costs to take traffic to and from there which increases latency . . . .

Here’s his stance on massive data centers:

These are critical to Google, which has been rumored to own 2% of the World’s servers. . . . However, the efficiencies that this scale provides just aren’t relevant in today’s SaaS and IaaS markets yet. Margins are super high and businesses pay well for complex infrastructure. Super low-cost PaaS offerings aren’t yet seeing traction from power users. Nobody is giving away free complex infrastructure (yet) on an ad supported model.

His points should be well taken when it comes to cloud computing, at least for small- to mid-sized providers such as GoGrid. If margins are high and profits are acceptable, there might be little financial value in building and managing large, geologically-sited data centers that trim costs in some places while adding costs in other places. As long as customers aren’t paying more because of the decision to use expensive power, who’s to complain?

Heck, even Salesforce.com — which serves nearly 100,000 customers and is operating at a $2 billion run rate — seems to be on board with Keagy’s theory. The SaaS leader only runs about 3,000 servers worldwide, and has said that if it’s forced to build its own data centers, it will site them based on network connectivity and not cheap land and power. When cloud providers have multi-tenancy and automation down cold, they don’t need huge server footprints and they don’t need to worry about operating efficiently enough to keep profit margins up.

Cloudscaling Founder and CTO Randy Bias, who helped build GoGrid’s cloud along with many others, noted the importance of automation last year. In a blog post challenging the proposition that only massive-scale clouds can achieve economies of scale, he wrote, “one major economy of scale is the ability to have significant resources deployed for software development purposes. The outcome of most cloud software development is generally automation or technology that enables the business to scale more efficiently.”

Of course, this all depends on the type of business you’re running. Companies like GoGrid and Salesforce.com have a pretty good idea what their computing demand will be at any given time, and they deal in longer-term commitments than does a cloud provider like Amazon Web Services. It’s really one-of-a-kind in terms of its reach and the adoption of its API, which means AWS might care more about energy costs and scale than does GoGrid. It has to operate an infrastructure truly capable of handling huge traffic spikes coming from individual developers, large customers or third-party platforms (such as RightScale) that spin up AWS instances for their customers, so every wasted penny matters. Although, AWS does charge a higher rate for resources delivered from its Northern California region.

AWS also has the Amazon.com connection, and webscale companies such as Amazon definitely have to care about cheap power and scale. We’ll hear more about this at Structure 2011 when data center experts from Facebook, Netflix, LinkedIn and Comcast sit down for a panel discussion, but the general idea is broadly accepted. When you’re dealing with huge amounts of traffic coming from all over the world, you need lots of servers in lots of locations. Further, as companies such as Facebook, Google and Yahoo certainly will attest, storing all that customer data and building systems for analyzing data also require huge infrastructural investments. When you start adding up the cost of being energy-inefficient at that scale, investing in efficiency starts to make a lot more sense.

I think Keagy, who also will take the stage at Structure 2011, is right that everyone need not get preoccupied with mimicking Google, Amazon and Facebook when building clouds or when evaluating cloud providers. There are plenty of companies doing just fine and delivering quality services by following the web giants’ lead in terms of automation and homogeneity without falling victim to the siren song of massive scale. I also think, however, that if GoGrid were to start driving a relatively high percentage of web traffic, Keagy probably would change his tune in a hurry.