Overview : Load balancing in publish/subscribe broker networks is challenging as the workload is multi-dimensional and content-dependent. In this work, we develope the framework design for a middleware, called Shuffle, that achieves optimal load balancing in a publish/subscribe broker networks. Shuffle features a suite of active workload management schemes that handle the four different types of workloads in a in a publish/subscribe service---message parsing, matching, delivery, and forwarding. Shuffle leverages its traffic randomization scheme and Chord, a DHT substrate, to build overlay trees for active workload aggregation and distribution. We prove the optimality of the load balancing scheme upon any input traffic distribution on individual Shuffle aggregation trees. We also show that the workload management problem for multiple correlated aggregation trees is NP-hard.
Paper:
Acknowledgement: I worked on this project during my internships at NEC Labs, America .