Generally an MMOG can be described as a distributed event-based system. The requirements for event-processing in this domain are high due to the soft real-time requirements which ensure a fluent gaming experience. For example, a low latency or a certain level of consistency has to be guaranteed. Especially the scalability has to be kept in mind when designing such systems, because the fluent gaming experience must be sustained, even when hundreds of thousands of players compete.

In the i6 M2EtIS Project, we examine the exploitation of event semantics in MMOGs to improve scalability and performance. Different characteristic types of requirements are identified, in order to allow for an abstract semantic classification of events. The visionary goal is the development of generic event-processing systems, which are optimized by the specification of abstract event-semantics.

In order to achieve this goal, a description language is developed that allows the classification of events based on their semantic properties. Based on such a classification, machine-learning algorithms can be applied that support the optimal configuration of event-processing systems. This automated configuration step is performed at design-time in order to minimize the performance impact at run-time. Measurements support this claim, as well as the fact that automated configuration based on machine learning is possible with a high precision in reasonable time.