3 Comments

Summary:

It looks like Zynga’s infrastructure strategy is rubbing off on gaming startups. Digital Chocolate, purveyor of social games such as Millionaire City and Pro MMA Fighter, is following in Zynga’s footsteps of launching games in the cloud, then bringing them back in house when demand levels.

storm clouds

It looks like Zynga’s infrastructure strategy is rubbing off on social gaming startups. Digital Chocolate, purveyor of social games such as Millionaire City and Pro MMA Fighter, is following in Zynga’s footsteps of launching games in the cloud, then bringing them back in house when demand levels off. It doesn’t have Zynga’s much ballyhooed hybrid cloud infrastructure quite yet, but Digital Chocolate is working on its own flavor of the ideal gaming cloud.

Alfred Tsai, Digital Chocolate’s director of global IT and network operations, told me the company was operating entirely in Amazon Web Services, but decided to bring some games back in-house when performance issues got to be too much. AWS’s database servers were plenty robust, he explained, but the backend network was just too slow. Digital Chocolate couldn’t utilize the database servers to their maximum performance because the network couldn’t keep up with the transaction volume.

Big Games Come Home

Now, Digital Chocolate still launches new games on AWS, but it runs its most-popular games in-house. It relies on a 10 GbE network backbone, as well as storage appliances from a startup company called Tintri. Tintri’s VMstore appliance mixes solid-state and hard disk drives to balance performance and cost, and is designed specifically for virtualized environments. Tsai said Digital Chocolate is still experiencing performance levels around 50,000 IOPS with its Tintri boxes, despite the fact that it has yet to upgrade from the beta version of the product.

One of the big benefits, he said, is that Tintri lets Digital Chocolate manage its entire set of gaming data as one big data store rather than having to break it up into multiple volumes. Each Tintri box provides 8.5TB of capacity that act as a single datastore, which Tsai said is far more than what VMware supports using typical virtualized SAN storage. Additionally, because Tintri monitors storage performance per VM, Tsai said it lets Digital Chocolate better architect games to run on its internal infrastructure.

For those curious about Tintri, it’s backed by NEA and Lightspeed Venture Partners, and was founded by Kieran Harty, who led all desktop and server research and product development at VMware from 1999 through 2006.

Long way from a Z-Cloud

Meanwhile, Tsai acknowledges that Digital Chocolate has a way to go before it has a hybrid cloud system as apparently seamless as Zynga’s Z Cloud, but it’s working toward such a setup.

He also noted that Digital Chocolate is experimenting heavily with other cloud providers in the hopes of achieving more resiliency. AWS is a great short-term replacement for physical hardware, he said, but it’s prone to outages, particularly its Elastic Block Storage service, which Tintri Digital Chocolate uses. EBS was the cause of AWS’s four-day outage in April that made national headlines.

One of the cloud providers Tsai is particularly impressed with is Virtustream. AWS launched its Direct Connect feature yesterday that provides direct connections between Equinix and AWS data centers, but, Tsai said, Digital Chocolate worked with Virtustream to run a 20-Gigabit pipe from the cloud provider’s East Coast data center to Digital Chocolate’s colocation cage in San Francisco. He said the support and costs are big improvements over AWS, too.

And Virtustream is working with RightScale and developing its own API, Tsai explained, which will make the easier to manage. If you’ll recall, RightScale is the middle man for Zynga’s cloud infrastructure, providing a uniform management interface for its AWS- and Cloud.com-based resources. APIs let customers write applications that can make calls directly to the cloud infrastructure to perform certain tasks.

Image courtesy of Flickr user Bill Ward’s brickpile.

  1. Digital Chocolate is misspelled in the headline FYI.

    Share
  2. Jeremy L. Gaddis Sunday, August 7, 2011

    “EBS was the cause of AWS’s four-day outage in April that made national headlines.”

    Sorta. EBS failures are what affected everyone, but the root cause was network issues, caused by human error:

    http://evilrouters.net/2011/04/29/the-ec2-ebs-outage-what-amazon-didnt-tell-you/

    Share
  3. Claudio Gentile Friday, August 12, 2011

    It has become rather apparent that companies need to make sure they arent just moving over to a cloud provider out of haste, as there can be some important decisions to make prior that may impair your solution (or turn out to be overkill). Based on a particular solution you may have different factors such as data payload, network bandwidth, I/O performance, etc…

    Ultimately we can say that each cloud solution will perform based on what hardware its been built with. In a public cloud scenario, we are still dealing with commodity hardware that isnt “top of the line” but still gives acceptable performance for most environments.

    The movement of resources from a public to a private cloud is simply progress from a business standpoint — you do your development on a server (cheap) in your garage and move to big horsepower when its time to shine (expensive). Private cloud can truly be a blessing in this case. Having a common cloud management platform enhances the simplicity of accomplishing these tasks without a significant amount of rework if any.

    Plenty more to cover, but this is a good start.

    Share

Comments have been disabled for this post