They can solve the problem of their server setup by using the Invisible Sharding method, and they are possibly one of the only games that can do such. Sharding is when a game server is split into multiple servers, which are separated from each other. This is why there are many different “realms” on World of Warcraft, and many of the other MMORPGs. Servers can only handle a certain amount of load at once, and splitting all the users across multiple server lets you still have a Massively Multiplayer experience while being infinitely scalable to more users. It works out pretty well, but it comes with some problems as well. If I want to join World of Warcraft and play with my friends, I need to know which server to start my character on in order to join with them. This can cause problems due to the way that social networks are very tightly wound. If you take you and three levels of friends, you have a very large amount of people. As a game grows, more people want to play with their friends and that means more people are on the same server. I heard that recently Blizzard had planned to migrate people from one server to another, which was a major event and then they improved their servers so they cancelled that migration. In a perfect world, no migration would be needed – everyone would be on the same server.

MTGO is the perfect world, but they suffer from the problem that the MMORPGs would have if everyone was logged onto the same server now – they have a server overload. So how do you solve this problem? The solution hinges on the good parts of the current architecture: games are separated from the main server. The main server can now be separated from the games, and we have a situation which is similar to a distributed consensus problem. The solution is to reduce the size of the data which needs to be consensus. There are two ways to do this. The first is to reduce the overall size of the set by limiting interaction between players – this is the sharding solution. The second way to do it is to be more intelligent with the data that you choose to pass between your servers. MTGO is unique in this respect because there is really a small amount of data which needs to be replicated in order for the system to run smoothly.

Version 3 of the program is launching in less than 2 weeks according to the timer – here’s hoping that they got the new architecture right, and we can see many thousands of players on at the same time.

Soon after I started playing the game again, I discovered that there was a new client and server being developed. You see, the Magic: The Gathering Online servers have a problem – they don’t hold up to pressure very well. From what I have gleaned from the forums (which the developers actually read and respond sometimes) the architecture currently in use looks like the diagram here.

Anyone notice a scalability problem? Would it help to point out that every single client is connected to the “main” server, and absolutely everything that doesn’t involve logging in or gameplay is done through that server? This includes trading, chatting, and just navigating to other games. Apparently the previous company who created this server structure didn’t think it would grow too large, or they were using it as some type of insurance. Because of this horribleness, the servers crash a lot – sometimes close to daily. The developers at Wizards of the Coast have been working on stability lately and the uptimes have improved. Still, lag is horrendous when using the main server — at least it doesn’t affect games. I’m glad that they’re replacing this bad server design with something that is better. I’ve been playing in the beta the last few weeks, and the new servers held up to a stress test with a large number of actual users — around 700-800, depending on who you ask — and a bunch of bots which were simulating a much larger load than a normal user. There wasn’t even a blip on my radar. It gives me some hope for the future of the game.