Increasing data throughput can be achieved relatively easy by using specialized database management tools for storing and retrieving data.
Optimizing network usage is less simple. A client/server architecture is definitely the fastest implementation for any query based system. This model relies on dedicated servers that a distributed architecture lacks, however. Which model gives the best trade off between overall responsiveness and resource utilization spreading is currently unknown. To my knowledge no tests have been done to identify how virtual structures, such as rings, hypercubes and freeform networks layered on top of a heterogeneous physical structure alter the performance of the environment. Clearly, more research need to be done to identify the different strongpoints and weaknesses of possibly useful topologies. For now, we can only look at existing implementations.