Zynga depends on Amazon's cloud at a critical juncture--as it launches a new game.
Zynga has filed for an IPO that cites how its fate is intertwined with Facebook. Zynga games, such as FarmVille and CityVille, sit atop the Facebook platform. But as everyone frets about Zynga's dependence on Facebook, there's another dependency people should pay close attention to: Amazon.
No price per share has been set yet, but the Zynga filing is expected to result in a valuation of the company at $15 billion to $20 billion. Zynga is a fascinating Internet success story: Using a $219 million venture capital stake, piggy backing on Facebook, and adopting an unusual cloud strategy that stands conventional wisdom on its head, Zynga has grown faster than any of its predecessors.
Its cloud strategy goes like this. Most companies think of using the cloud as an extension of their data centers, shifting work once it becomes too much for in-house systems into the cloud. Zynga does the opposite, launching its games in the cloud and bringing them in-house only when it knows what resources they'll consume as they settle in to a regular demand pattern.
Zynga is also a leading example of the power of extending free computing cycles to consumers from massive, quasi-public x86 data centers. In Zynga's case, anyone may access for free its games; they're hosted either in Amazon's or its own Z Cloud data centers. This democratization of compute cycles reaches millions of customers quickly. Also, Zynga understood better than most that the Facebook platform brought millions of a new type of game player within reach. It takes time, attention, and competitive intensity to do well in World of Warcraft or Combat Flight Simulator. But nearly anyone can compete for power in Mafia Wars or grow crops in FarmVille, using existing social skills.
In talking to Allan Leinwand, CTO of infrastructure engineering at Zynga, I learned about the uncertainties of what happens with a game launch. In some cases, not much happens, and the game fails to reach a self-sustaining mass of players. But other times, as the authors hold their breath, a game starts to gain more adherents each day, then suddenly some critical mass is reached and an explosion of players occurs. With each game, the trajectory is slightly different. The crucial thing, said Leinwand, is to avoid limiting the explosion after it's gestated. You can avoid that by being prepared to scale up and serve a multitude of new users.
The problem for most companies is that being prepared all the time would mean investing up front in data center capacity. If the pay off fails to materialize, the capital expense of having prepared for it would be crippling.
Leinwand developed a strategy of launching games in Amazon Web Services infrastructure as a service, even if there was little initial demand. Sometimes it takes several months for the reviews to appear in the right places and for word of mouth to attract interest. If a game is catching on, slowly but surely, it's key to capture early interest. If a prospect can't get access to the game when it's fresh, the spark of interest may die out and players drift away before that interest can be fanned into a flame. In other words, scaling to meet demand can't be a hit or miss proposition. The ability to scale has to be there soon after launch. Trying to figure out how to scale as a game starts to succeed is potentially fatal to a game's prospects.
In November, Leinwand launched CityVille using Amazon's EC2's pay-per-use infrastructure. Progress was slow at first, then CityVille took off. After four months, it had 20 million monthly players. Leinwand read the numbers of how much capacity he would need after several months experience and brought it back inside Zynga.
It's running on his public cloud-like, x86 infrastructure known as the Z Cloud, now hosting about 88 million active users a month. He doesn't spend two years planning and building Z Cloud infrastructure; he more typically buys it pre-built from wholesale data center space builders. He can cope with CityVille's continued growth because he can add up to 1,000 physical servers a day to his infrastructure, if necessary.
On May 31, Zynga launched Empires & Allies in Amazon's EC2, which had an even faster ramp up. By the end of June, it added 42 million monthly users, but Amazon's elastic infrastructure had no trouble providing Zynga with the capacity it needs. On May 17, he had launched GagaVille, in conjunction with the celebrity Lady Gaga, in the same manner.
This is a hybrid cloud strategy that leverages the capital investment of the public cloud until Zynga is dealing with a known quantity. Leinwand said he wants to be able to "plot the curve of a game's trajectory," then he is willing to invest IT resources to host another asset.
This stands the usual view of hybrid cloud computing on its head. For the most part, the hybrid cloud has been viewed as a place for cloud bursting, a destination for offloading enterprise workloads when their need for compute cycles exceeds what's available in the enterprise data center. Zynga uses the cloud upfront, when the workload is most unpredictable; only later does it bring into its operations--once it has reached either a steady state of growth or steady state of use.
Multicloud Infrastructure & Application ManagementEnterprise cloud adoption has evolved to the point where hybrid public/private cloud designs and use of multiple providers is common. Who among us has mastered provisioning resources in different clouds; allocating the right resources to each application; assigning applications to the "best" cloud provider based on performance or reliability requirements.