How Zynga Survived FarmVille • Hacker News Board

Re: How Zynga Survived FarmVille

#11

Post by smokinn » Wed, Jun 09, 2010, 5:59 PM UTC

dalore wrote: post #7

fserb wrote: post #5

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…

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.

Re: How Zynga Survived FarmVille

#12

Post by aristus » Wed, Jun 09, 2010, 8:11 PM UTC

amitt wrote: post #10

pwim wrote: post #2

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.

Re: How Zynga Survived FarmVille

#13

Post by amitt » Wed, Jun 09, 2010, 8:45 PM UTC

aristus wrote: post #12

amitt wrote: post #10

Earlier quoted context omitted.

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.