question

1. Photon Realtime - There is actually an option to have your logic be server-authoritative in this model, using their plugins model. It's more expensive than their core Realtime service, so it may not be an option, but it's worth talking to them about that, to get a feel for the costs. Your idea of using a calculation to determine the "best host" is a good one in some ways - I actually worked with a team in Xbox on a similar system, though it was specifically to find the best host purely from a bandwidth/quality perspective. The potential issues you'll have with it are when people identify the elements used in your logic and start finding ways to "game" it (which, for any popular title, is inevitable). Also, there is the issue that the host will have a far better ping time than anyone else. So if there's any realtime component to the game at all, you'll be giving an unfair advantage to a subset of players frequently. And because your algorithm would necessarily not favor new players, that could make for a bad new user experience. But it really depends upon the game specifics, so that's something you should playtest, if you have any doubts there - just things to consider.

2. Unity Server Library - I really can't offer any advice on this, as they haven't given us any information outside what you've already seen. Personally, I would never take a dependency on anything that you can't get a solid date on for release, and solid docs to help make sure it meets your needs well ahead of that. That's one of the reasons we don't provide "guesstimate" dates for anything we're not actively working on. If someone were to take a dependency on it and then we missed that date, it would be bad for the developer.

3. Photon Server - For info on Photon Server pricing, I'd have to direct you to the Exit Games team. It sounds like you already understand this, but please bear in mind that the Photon Server license is a license to run their software on servers you host (or have hosted for you). Our custom game server service runs in EC2 and automatically scales to support your players, so we spin up additional servers (using the values you supply) as needed, and spin them down when they're not. There's a complete write-up of this here: https://api.playfab.com/docs/custom-game-servers. The costs are specifically the AWS EC2 pricing, plus 10% to cover our costs. You can find the EC2 base pricing here: https://aws.amazon.com/ec2/pricing/. The prices to look at are Windows On-Demand servers. We'll also be providing a way to run your own servers in any server host later this quarter. This will allow you to run servers that integrate with our matchmaker and game wrangler (which is what pulls the log and output files when your server instances are done), but we won't be able to control spin-up/spin-down, as we won't have that level of access, so that would be your responsibility in that model. To calculate costs for custom game servers, you need to work out what you believe the average number of servers you'll need to run in each region where you want to have active servers - and note that server hosting costs vary by region, so you'll want to make note of those prices for your calculations. We provide a single small server instance for free while you're in development, which you can use to make some calculations on resource usage by your game per player, so that you can determine what server model you want to use when you go live. We recommend using the New Relic add-on from our marketplace to get that data. There's no way for us to predict how many users you could host on any particular server model, as games vary wildly in this respect. A game with low usage per player might be able to host thousands of players on a server type that, for another game type with much higher usage, might only be able to hosts dozens, so it's important to do thorough testing on your title. It's also worth noting that we provide a server framework you can use in our GitHub repo, as well: https://github.com/PlayFab/PlayFabGameServer.