Investigate methods to make clusters more fluid.

[mike]

Currently, once data is moved onto a machine, it is more or less stuck there. There are ways to manually move it off, but that would be hard for large amounts of data.

It would be nice if a cluster of servers could be handled as a blob of computing power, and if Overseer could just spin up a new process on the least loaded server.

One of the thoughts here is the possibility of having a central database instead of a distributed one.