Open Access

David Alessandro Bauer, Juho Mäkiö

• Common actor implementations often use standardized thread pools without special optimization for the message passing. For that, a high-performance solution was worked out. The actor-oriented software framework Akka uses internally a ForkJoinPool that is intended for a MapReduce approach. However, the MapReduce approach is not relevant for message passing, as it may lead to significant performance losses. One solution is to develop a thread pool that focuses on the message passing. In the implementation of the Actor4j framework, the message queue of the actors is placed in threads to enable an efficient message exchange. The actors are operated directly from this queue (injecting the message), without further ado. It was further enhanced by the use of multiple task-specific queues. Fairness and the responsiveness of the system have been raised. In particular, the performance measurement results show that an intra-thread communication towards an inter-thread communication is much better and has very good scaling properties.