SOLARE: Self-Organizing Latency-Aware Resource Ensemble

• Main Authors: Eom, H., Wolinsky, D. I., Figueiredo, R. J.
• Format: Conference Proceeding
• Language: English
• Subjects: Clustering; Clustering algorithms; Computer architecture; Heuristic algorithms; latency-aware; Monitoring; Peer to peer computing; Performance evaluation; Servers; structured P2P networks; utility functions
• DOI: 10.1109/HPCC.2011.38

This paper proposes and evaluates Self-Organizing Latency-Aware Resource Ensemble (SOLARE), a peer-to-peer self-organizing and self-managing cluster system based upon network coordinates and utility functions. In contrast to previous works, SOLARE is a fully decentralized clustering algorithm without any central units such as servers, super peers, cluster heads or landmarks. Furthermore, SOLARE allows for adaptability to dynamic network changes by monitoring the utility of a cluster and migrating nodes to other higher-utility clusters when the utility of an existing cluster is low. Quantitative, simulation-driven evaluations show that SOLARE is able to satisfy user demands expressed by utility functions that integrate system parameters in terms of intra cluster latencies and the number of cluster members. Also, we verify the ability of SOLARE to adapt to dynamic network changes through simulation based experiments that consider the number of nodes which migrate into another cluster and average utility value as nodes join SOLARE.