Towards Highly Available Complex Event Processing Deployments in the Cloud

Recent advances in distributed computing have made it possible to achieve high availability on traditional systems and thus serve them as reliable services. For several offline computational applications, such as fine grained batch processing, their parallel nature in addition to weak consistency requirements allowed a more trivial transition. On the other hand, on-line processing systems such as Complex Event Processing (CEP) still maintain a monolithic architecture, being able to offer high expressiveness and vertical scalability at the expense of low distribution. Despite attempts to design dedicated distributed CEP systems there is potential for existing systems to benefit from a sustainable cloud deployment. In this work we address the main challenges of providing such a CEP service with a focus on reliability, since it is the most crucial aspect of that transition. Our approach targets low average detection latency and sustain-ability by leveraging event delegation mechanisms present on existing stream execution platforms. It also introduces redundancy and transactional logging to provide improved fault tolerance and partial recovery. Our performance analysis illustrates the benefits of our approach and shows acceptable performance costs for on-line CEP exhibited by the fault tolerance mechanisms we introduced.