How Zynga Survived FarmVille | Hacker News Reader

How Zynga Survived FarmVille

gigaom.com

60 pointsmattjung12 comments

He mentions that FarmVille maintains both their own servers and EC2 servers. That's an interesting approach to maintain some baseline capacity locally and use the cloud for scaling. I suppose its obvious, but usually with the cloud I hear about all or nothing (well or just some specific task in the cloud).

The article is somewhat confusing, for FV we didn't split our servers across ec2 and a data center. games are either fully on ec2 or fully in the data center.

The crux of the problem is data and latency, right? I imagine it's infeasible to have a hybrid architecture the way cloud providers structure their networks and pricing today.

exactly. its the same reason you want to have your databases and application servers in the same data center.

Best of both worlds I guess, I'd imagine they could run it on their own servers a lot cheaper but would get destroyed when a new game peaked. Using the EC2 as overflow they can make the best use out of their own servers rather than buying up big computing power that is only ever used in user spikes.

Yes. An suitable analogy can be made to a hybrid car. Local servers (electric motor & batteries) for constant/idle-demand and EC2 (internal combustion engine) for high bandwidth & storage overflow during version deployments and peak loads.

We actually recommend this solution our clients who already have significant sunk costs and long-term SLA contracts in hardware and datacenter services.

The story is interesting, but I still can't figure out why having both their own servers and EC2 splitting 50/50 is a good idea. If EC2 dies the private server won't be able to handle all the traffic and FarmVille will stop working. If the private servers die, they will have a little hiccup until they manage to rescale EC2 and everything will be back to normal. With this system, in practice they already fully depend on EC2 and have an extra point of potential failure. What am I missing?

The might just be amortizing the cost of the servers they already bought in their data center prior to moving to EC2.

Probably that the cost per player is higher on EC2, so they want to be able to migrate games that have stopped growing unpredictably to their own infrastructure.

They did say they didn't want to test Amazon's capacity. Also, presumably, running on their own servers is cheaper.

EC2 is actually broken down into a few completely different clouds. If they only used one EC2 location they would be silly.

I agree, they really shouldn't use/need their private servers, but I suspect its a data control issue.

I doubt it's a question of data control.

EC2 is significantly more expensive than running your own servers so it's a good idea to keep close to your base load on your own servers and then just scale up/down as necessary with EC2. This way you're optimizing your cost per player.

Reply on news.ycombinator.com