Introduction

Peer-based desktop grid computing, the harnessing of idle computer cycles throughout the Internet, offers exciting new research challenges for peer-to-peer networks beyond current file sharing applications. Exprience has shown that not only are vast quatities of idle cycles widely available throughout the Internet, but in addition, many users are willing to share cycles. This creates a compelling opportunity for research in this exciting new juncture between the fields of networking and distributing computing.

The Cluster Computing on the Fly (CCOF) architecture provides a lightweight mechanism for forming a community-based cycle-sharing overlay networks in which any node can be a client seeking idle cycles or a donor volunteering its cycles. CCOF also includes a reputation system to support fairness, quality of service, and security.

Research Issues:

• Incentives: What kinds of incentives are needed to motivate hosts to donate cycles?
• Resource Discovery and Application-based Scheduling: How can a client efficiently discover hosts that offer idle cycles in a open and dynamic environment? How does a client decide which hosts best suit the specific needs of its application?
• Trust and Reputation: which hosts can a client trust to run its job and return the correct results? Which host will return results most quickly?
• Fairness: How does one ensure fairness in a large scale and dynamic cycle-sharing community?
• Quality of Service and Performance Monitoring: How can a host provide Quality of Service in an open environment? How can clients evaluate the service they are receiving?
• Security: How can a host defend itself against denial of service attacks, such as when a malicious node occupies large numbers of hosts with useless code? Or worse yet uses hosts to launch a distributed denial of service attack or a worm?