Aside from working on a side project to my side project, which I recently finished, I’ve gotten back to my original side project (my next game) and found myself researching on some tech for a large feature. Talk about “writing in code”! Spit it out! That tech I’ve been researching on is multiplayer networking solutions for Unity. There. Like the title didn’t give it away.
For the uninitiated, multiplayer games are games that can be played with other human beings across the Internet. Yes, this includes social games like those popular Facebook games of old*, and games like Pokemon GO, all the way to games like Call of Duty and World of Warcraft. Some of these games have competitive play and others have cooperative play, and sometimes there’s a combination of both. This, of course then, is different from single player games, like Crossy Road or Flappy Bird, which don’t involve multiple people playing together. Some games actually have both singleplayer and multiplayer modes, like the aforementioned Call of Duty. Nonetheless, I’ve already established that I’m writing about multiplayer here.
Game developers know full well that one of the most difficult systems to build is the multiplayer networking system of a game. I’m not even going to go into how difficult it is to build a Massively-Multiplayer Online (MMO) game, as that’s way out of scope of this article. If you’re still thinking about it, I’d highly recommend reconsidering your life goals.
But enough of the intro.
Unity Networking, Photon SDK, and PUN
For the past several days, I’ve been looking at potential network solutions for Unity, namely Unity Networking (Unity’s first-party solution) and Photon by Exit Games. From what I gathered, Unity Networking still needs some time to mature, whereas Photon is a more established networking package, with a long running track record. I decided to go with Photon for this reason, as well as for its healthy amount of documentation, so I’ll go into more detail about it here.
As of this writing, Photon has a sort of mixed personality. For Unity, it comes in two flavors. Photon SDK for Unity, and Photon Unity Networking. Whazzawha? Right. Photon SDK for Unity is available on the Exit Games website. Whereas Photon Unity Networking (PUN) is available as a downloadable asset package from the Unity asset store. The difference is that PUN is a wrapper around the underlying Photon SDK, to help you save time with the lower level groundwork code already implemented for you. PUN is also supposed to emulate the API of Unity Networking, so that it is more familiar to the Unity network application programmer.
So, I chose to use the Photon SDK for Unity. Why would that be if PUN already provides a lot of the underlying work for me? Well, there’s one feature that is currently not supported by PUN that I need for my next game: Persistence of asynchronous game sessions. I’ll go into more detail about what that is in a moment.
To make things more complicated, doing any Google searches for Photon Networking will inevitably yield results on “Realtime” vs “Turnbased” Photon. The hope for Exit Games was that Photon can be marketed for developing games that are fast-paced, and played simultaneously, or synchronously, between gamers (Realtime), and for games that are slower-paced, and played by one player at a time, or asynchronously, like Chess (Turn-Based). As of May 2016, Exit Games discontinued Photon Turnbased, which by itself sounds as if Photon would no longer support turn-based games. In reality, the Turnbased features have been merged into Photon Realtime. So that means, I’m using Photon RealTime for a turn-based game. How confusing! At the same time that Exit Games was attempting to simplify the delivery of their product(s) to its customers, it managed to confuse even newer customers like me.
Asynchronous Gameplay and Persistence
To support my next game, which is a turn-based game, I need to implement asynchronous gameplay and persistence via something called Webhooks/WebRPCs.
“Asynchronous gameplay” means that the players do not play simultaneously, together on the same turn. From a network programming perspective, this is far less difficult to implement than synchronous gameplay because with synchronous, realtime gameplay, there is no concept of “turns”. With synchronous games, developers have to implement the system such that every player connected to the game can give input and receive a result immediately, so as to produce the illusion that all players “exist” in the same virtual world. In reality, it takes fractions of a second, or even whole seconds to transmit that data across to all other connected players. Network programmers have to implement numerous tricks and techniques like dead reckoning and client prediction to generate the illusion that a player has moved, without any lag in time. Otherwise, if the game goes out of sync, the illusion is completely broken.
For each turn that a player takes in my turn-based game, the current state of the game has to be saved so that the next player can continue from that state. This is where persistence comes in.
So for persistence, or storage, of game sessions in progress, the Photon SDK provides Webhooks and WebRPCs to load and save game data; something that PUN does not provide.
Webhooks and WebRPCs go hand-in-hand for game clients to communicate with web servers. WebRPCs (Remote Procedure Calls) are used by clients to send messages across the network, so that web hooks on a web server can receive those messages and process them in a secure, authenticated environment. Note that web servers are your typical type of server for hosting web pages. So while in this case, a web server is being used as a game server, it could also be the host for web pages to convey that data, say, for players to check their progress, or the states of their game sessions. Also note that web servers are not used as game servers for real-time games. Servers for real-time games require support for higher loads and bandwidth, because they generally process a lot more network traffic than for turn-based games, to maintain synchronicity across all connected clients, as mentioned above. This is why web servers are more suitable for turn-based games.
And with all of that knowledge, I discovered that Photon does not provide web servers, only the hooks! Bah! Before learning about it, I thought Photon was going to be my one-stop shop for all my network and multiplayer needs for my game. At least, that’s what their marketing made me feel like what it was. In reality, a more seasoned network gameplay engineer probably would have been able to see past the bullet points and tell me that I still needed another component to get Photon working the way I wanted for my turn-based game.
Luckily, through Photon documentation (or somewhere on the Photon or Unity forums), I found out that PlayFab is the recommended solution that provides that web server that I’m looking for. Not only that, it provides some other good stuff that I’d need that Photon doesn’t provide, like user accounts, leaderboards, etc. Also, GameSparks is a competitor which provides pretty much the same thing, but I admit that I didn’t do much research into which would be better for me. Since PlayFab has a partnership with Photon, I decided to go with that.
Once I found out that I had to use yet another system besides Photon, my heart sunk. I was already feeling pretty overwhelmed with having to learn the new Photon API. I was reluctant to learn even more than I was planning to. But we’re game developers, right? We keep learning and work through it! That’s what we do!
Turns out PlayFab, after learning what they offer and how to navigate their browser-based Game Manager, provides quite a bit for the entry price of free. I like how it provides user accounts, tracks when and where they’ve logged on, and other analytics, such as segmentation, which is a business term that is beyond the scope of this post. They have this system called PlayStream which is intended for logging, but I haven’t figured this one out yet. What I was most interested in was, of course, how to implement web hooks for the Photon service.
Keep On Keepin’ On
At the time I am writing this, I’m still in the midst of working through Photon and PlayFab so that I can get a working game flow for the player. While there’s a wealth of information on both services, it’s very fragmented and difficult to piece everything together. In future posts, I’d like to go more in-depth with my Photon and PlayFab setup, because I’m finding it a pain to get through right now. Perhaps others planning to take this route will benefit from my findings. In the mean time…
Make it fun!
* Of course Facebook is not that old. But in “technology-and-video-game” years, 4-5 years is like 4-5 decades…. Is Farmville even still a thing?
Are you still working with PUN and PlayFab?
If so, what are your thoughts about them, would you recomend them to somone starting now? If not, what would you recomend?
Hi! For a time I considered GameSparks, as the feature set seemed a lot nicer than PlayFab. However, since GameSparks was acquired by, I think, Amazon?, their pricing model was outside the reach of any budget indie game developer. So, I’m still considering PlayFab. I’m still in the middle of working on my game, putting multiplayer on the back burner for now, but keeping the architecture adaptable to a future multiplayer solution.
As far as Photon is concerned, after invetigating it, I really had trouble seeing why I needed it altogether. I think for my game, PlayFab will be suitable enough, since I just need a way of loading/storing turn-based data, but for real time games, perhaps Photon is still valid.
Another concern I have is matchmaking, but it’s such a low priority issue for me right now. I’m hoping over the next several months, PlayFab’s matchmaking system is more established. I’m not even sure where they are with that at the moment.
I will also considering ChilliConnect (haven’t looked into it yet) when the time gets closer for me to work on multiplayer, and a more recent article I’ve read (may have to dig for it) suggested that even Firebase is a solution. I’ve used Firebase for a smaller project, for simple content-authoring, and I really liked the workflow. I haven’t evaluated it for a multiplayer game yet though, but it looks promising.
I’ve been meaning to write up a more formal update, ever since the GameSparks news, but haven’t had time.