چكيده به لاتين
One of the major challenges of distributed complex event processing systems is the asymmetry and unpredictability of the event rate in real systems. Due to the rate of some explosive events, the computational power of machines to process incoming events is not enough. Detecting a complex event is very sensitive to applications, and lacking processing power can cause a loss of time or even a failure to identify a complex event. Therefore, in order to obtain suitable computational power and escape from the run-time bottlenecks, we use dynamic adaptive methods used in High Performance Computing systems. Therefore, in this thesis, a mechanism for adaptive scaling of rules and their distribution between machines is presented. This mechanism is based on the dynamic combination of rules that if the load passes through a certain threshold, it is activated by the runtime system to balance machine load. Current solutions are mostly static, and parallelism is explained during the initial setup, or the migration from a critical machine to another machine is considered. None of the proposed solutions have paid attention to the cost of running and scaling of the run time. The proposed mechanism and its evaluation, compared with similar solutions, show up to 5% improvement in lost events.
Keywords: Adaptive Scaling of Rules, Rules Distribution, Rules Migration, Distributed Complex Event Processing.
