Persistent Worlds and Their Storage
Over the past few months I have been putting together an MMO-style bit of software. Since it is more of an experiment than anything else I did not start with a design plan. That is not to say that most things are not planned before hand but I have no idea what will work best so I am trying a number of things off the hip first.
Right now I am working on the basis of what will make it multiplayer. The decision I have to make now is how will the data be stored and how will the clients access it?
- I could store everything in an SQL database. This is attractive for its persistence and accessibility across multiple platforms and languages. The down side is I can not control what is cached and what is on disk as much as I would like. Every now and again I may take a huge hit in performance as it was not designed for this task. I may hit a bottle neck much sooner in a high concurrency situation than I otherwise would.
- I could use memcached. This is attractive for the obvious reason: blinding speed. The down side is I would have to do so much more work in code since it does not guarantee stored data would exist when I need it. This increased work could place my bottleneck on my CPU when it is already pretty high from other tasks. I would not know the full effects of this until after the project is mostly complete leaving me in a chicken or egg situation.
I am sure there are many other options. These are the two that seem the best suited for my task right now that I am aware of.
No matter what I do I will build a very light-weight abstraction layer as to switch between different designs quickly. This will save a lot of time later on so I do not have to reinvent the wheel over and over again with each test.