High performance staged event-driven middleware

In this paper, we investigate the design of highly efficient and scalable staged event-driven middleware for shared memory multi- processors. Various scheduler designs are considered and evaluated, including shared run queue and multiple run queue arrangements. Techniques to maximise cache locality while improving load balancing are studied. Moreover, we consider a variety of access control mechanisms applied to shared data structures such as the run queue, including coarse grained locking, fine grained locking and non-blocking algorithms. User- level memory management techniques are applied to enhance memory allocation performance, particularly in situations where non-blocking algorithms are used. The paper concludes with a comparative analysis of the various configurations of our middleware, in an effort to identify their performance characteristics under a variety of conditions.peer-reviewe