Cache for workflows

Abstract

This paper discusses the motivation for the design of a decentralised data-caching scheme for Internet scale distributed computing applications. We provide three target applications in distributed cycle sharing applications, music information retrieval workflows and for workflow-based audio visual searching across distributed information servers. We argue that extensions are needed in existing toolkits to advance super peer topologies to support the caching of application data as well as adverts for specific applications. We discuss the Alchemist system that defines super peers style overlays for caching application-data by employing the use of dynamic groups, which are flexible in defining replicating, forwarding and security policies. This framework is built on top of existing technologies, such as P2PS and WSPeer and exploits standardised Web Services technologies as well as SOAP and more recently defined Web Services specifications. We present results to indicate the usefulness of such a data caching approach for distributing workflows and application data, and discuss future simulations that aim to sustain the evolution of the system.