hey everyone,

I've been researching hosting providers for a multiplayer card game I'm working on.

Something that came up in my research was Amazon GameLift.

It looks VERY bad ass for session based multiplayer games. In a nutshell, you use their SDK in your game client and server, and then upload your server binary to GameLift. You then set parameters for how much load a game server Instance should handle (e.g., say 50 players). It will magically spin up new servers to handle peak time loads and then spin them down after peak hours. This should not only keep costs lower than traditionally pre-paid/static hosting, but should in theory make the game "infinitely scalable" (well, at least the game/session server part).

Anyone here using this already and have feedback? Any thoughts in general?

Lots of games already did this using Amazon EC2, or the equivalent services from other hosting companies, but yeah, that requires you to program a lot of stuff yourself. If you use GameLift, a lot of that work is done for you... The major downside is that you're locked in to use Amazon EC2 and can't ever switch your game to another hosting provider...

Amazon is very cheap for computing resources (CPU/RAM), but they're extremely expensive for bandwidth costs.

hi Hodgman, thanks for the quick reply.

Yes, I agree, the downside is that you're locked into AWS. *However*, even using EC2 OnDemand and creating my own 'intelligent provisioning engine' would still keep me locked into the AWS API to spin up and shut down instances, right? Since there's really a minimum amount of GameLift SDK code needed to get this up and running, I think the 'commitment' level is reasonably low. If AWS proved to be too expensive even with the dynamic provisioning, I could move to something like Digital Ocean and create my own smart provisioning (which was the other alternative I am considering)...

Also, I agree, the downstream data cost sucks with AWS. Since I'm making a card game and not a shooter or action game, I think that it should be fine though. Although benchmarking tests would probably prove that better.

Also - the other thing that drew me to GameLift is Amazon Aurora. Amazon Aurora is super bad ass, scalable SQL. Since as you pointed out, getting data out of Amazon is expensive, using something like Digital Ocean for the app servers with Aurora for DB servers would be silly since I'd be paying Amazon every time I ran a query. So with AWS you get super scalable SQL too...

PS, I'm open to alternatives to GameLift if anyone has any. The other one already mentioned is Digital Ocean + home brew 'auto provisioning'. I also found GameSparks and PlayFab, but don't love either of them as GameSparks seems too locked down into their own proprietary server engine and PlayFab doesn't have great documentation on how to setup and use their Authoritative Server hosting. It also comes with a whole bunch of bloat I don't need personally (e.g. inventory tracking, player rewards, friends list)

Amazon servies work. They are not particularly cheap, compared to doing all the legwork yourself and finding the cheapest option for each instance, but sometimes, it's worth signing away a little bit of possible future profits, to get a complete working solution that's already been tested.