Abstract

Stream processing systems (SPSs) are created by the need to process a continuous flow of information with variable magnitude in order to detect and process events as close as possible to real time. These systems are deployed in clusters of processors to get benefits from the distributed processing in order to scale up applications. The present work addresses the load balancing problem in SPSs that appears when working on distributed systems. A hybrid solution is presented which is assessed in the S4 system. The results show that it is possible to generate a solution that is decentralized, information lossless and transparent to application developers, that is capable of evenly distributing the load on the processors.