Performance

Although performance was not a very important issue during testing we cannot deny its significance in the real world. The implementation I created was absolutely not fitted for widespread use. Especially the parsing of the XML document proved, even with specialized parsers, to be a very time-consuming task. Large scale networking tests could not be run due to the lack of resources. From other research mentioned in the first section we can conclude that our implementation cannot scale reasonably either. These two deficiencies represent by no means absolute truths, however. It only indicates that a lot of work will have to be put into optimizing network code and improving data throughput.

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.