In Silicon Valley, the true sign of success has been to build one’s own data center. As the theory goes, once you hit a certain size, owning your infrastructure becomes essential to achieving the design and deployment flexibility for outpacing rivals. Importantly, very few companies get to play in this Data Center Club and, hence, few companies get to define the future of the Internet.
But what if owning a data center is a liability, not an asset?
Zynga’s zCloud Experiment
This seems to be one lesson from Zynga’s recent battering. Yes, Zynga CEO Mark Pincus was referring to head count, not server count, in a recent all-hands employee meeting when he said, “We want to get back to being a nimble company that moves quickly to innovate, iterate, and find the player heat.” But he might as well have been talking about the company’s big investment in its own infrastructure.
As InformationWeek‘s Charles Babcock reported in 2011, Zynga has long had a strategy of moving successful games from the cloud – Amazon’s EC2, part of the Amazon Web Services cloud platform – to its own data center. But in mid-2012, Zynga seriously accelerated this, pushing 80% of its workloads to its own servers and off AWS to improve efficiency. Along the way, Zynga built out its own zCloud to take the load of its increasingly popular games.
But “efficiency” can be read more than one way, especially with the waning of popularity in some Zynga game titles.
The Inefficiency Of Ownership
As Netflix cloud architect Adrian Cockroft notes, commenting on Zynga’s cloud strategy, there may be better ways to save money than by building a data center:
If we need to save money, we optimize usage, tune code and spend less. No lock-in to big datacenter spend. I’d rather spend $100M capex on the next original series or a new country launch than a datacenter. How much did [Zynga] spend on ZCloud, and what’s the utilization level now? Opportunity lost to spend on new product.
‘s Jack Clark put it:
The more you build your own infrastructure, the thinking goes, the more you’re tied to the possibilities that infrastructure affords. Sometimes that may be good, as it seemed to Zynga when it was on the ascendancy. But even then, spending money on expensive data centers is money that can’t then be spent on other areas of the business. And it’s debateable that any one company has the money to keep up with Amazon’s pace of innovation, as Cockroft has argued, and as the company encapsulates in a presentation on business and development flexibility:
The Bottom Line
All this aside, it’s not clear that Netflix is right, and that Zynga is wrong to build its data center. What is clear is that Zynga doesn’t seem to be getting the business flexibilty from owning its infrastructure that it needs. Perhaps the new sign of startup success will be operational excellence on someone else’s infrastructure, à la Netflix, rather than the need to go buy infrastructure in bulk.