What are some best practices for making sure your .NET code will scale well?

Question

Last week I interviewed for a position at a TripleA MMORPG game company here in NE. I didn't get the job but one of the areas that came up during the interview was the about the scalability of the code that you write and how it should be considered early on in the design of your architecture and classes.

Sadly to say I've never thought very much about the scalability of the .NET code that I've written (I work with single user desktop and mobile applications and our major concerns are usually with device memory and rates of data transmission). I'm interested in learning more about writing code that scales up well so it can handle a wide range of remote users in a client server environment, specifically MMORPGs.

Are there any books, web sites, best practices, etc. that could get me started researching this topic?

Answer 1

Here are some places to start:

http://highscalability.com/blog/2010/2/8/how-farmville-scales-to-harvest-75-million-players-a-month.html

http://www.cs.cornell.edu/people/~wmwhite/papers/2009-ICDE-Virtual-Worlds.pdf

In particular, http://highscalability.com is full or articles about huge websites that scale and how they do it (Digg, flickr, facebook, YouTube, ...)

Answer 2

If you are looking for physical validation, what I usually find that helps is doing some prototyping. This gives you a good idea usually of any unforeseen problems that might be in your design and just how easy it is to add onto it. I would try to apply any design patterns possible to allow future scalability. Elements of Reusable Object-Oriented Software is a great reference for that. Here are some good examples that show before and after code using design patterns. This can help you visualize how design patterns could make your code more scalable as well. Here is an SO post about specific design patterns for software scalability.

Answer 3

Just one point I'd like to highlight here. Just cache your reads. Work out a proper caching policy where you determine which objects can be cached and for what periods. Having a distributed caching farm will take load off your DB servers, which will greatly benefit performance.

Even just caching some pieces of data for a few seconds - in a very high load multi-user scenario - will provide you with substantial benefit.