Y. Wang and P. Dasgupta (USA)
Distributed software systems and applications, Operating systems, Distributed computing, Fault tolerance
In recent years, there are more and more critical services being developed on the Internet with high requirements on availability and reliability. Server farms are one of the key approaches to improving the availability and reliability. However, due to the high maintenance cost and the “site disaster” problem of server farms, we are looking for cheaper infrastructure that is logically and geographical distributed to provide highly reliable services in spite of attacks and server failures. In this paper we provide highly reliable, attack resilient servers by using P2P farms of volunteers. The volunteer group can be dynamically constructed, is scalable and is resistant to service failures resulting from the loss of a few peers. In our approach, the system will first form a self-organizing group of peers who are interested in a service. Then statistical and operational data, service response delay, client request pattern, and so on are collected. Finally, the system applies this information to select new servers, and integrate into the server farm till the target level of service is met.
Important Links:
Go Back
