Summary

In 2001, the research community is still trying to study the behaviour of multi-player online games. Therefore, the authors setup a simulation environment to find out the load on a multi-player game server.

This paper claims that memory requirement of the server is not an issue, but the processing power is. Doubling the speed of the CPU results in a less than double increase in the number of players. It also recommends using server response rate and average response time as benchmark for server evaluation.

Discussions

There was not much time for discussions as there papers were presented, but a few points were discussed:

1. A student suggested using multiple servers to relieve the load, but as the author points out in one of his later papers, keeping the states consistent among multiple processors is the bottleneck. Thus multiple server should yeild worse outcome as the communication among servers would be a lot slower than among processors within one server.
2. A student suggested dividing the map into cells, so each cell can be processed independently. This is a topic of a another paper that would be presented later as Buck pointed out, thus there was no further discussion on that.
3. A student misunderstood the drop of server response rate when there were to many users as the server being more efficient. In fact, the server was so overloaded, it started to skip some replies and responses.