A Scalable Stream-Oriented Framework for Cluster Applications

Abstract

This paper presents a stream-oriented architecture for structuring cluster applications. Clusters that run applications based on this architecture can scale to tenths of thousands of nodes with significantly less performance loss or reliability problems. Our architecture exploits the stream nature of the data flow and reduces congestion through load balancing, hides latency behind data pushes and transparently handles node failures. In our ongoing work, we are developing an implementation for this architecture and we are able to run simple data mining applications on a cluster simulator.