The India protocol - Project report

Abstract

The goal of this project is to explore the potential of the IDA (Information Dispersal Algorithm) as the basis for an efficient, highly available, fault tolerant, and secure distributed data storage system. To this end, we have developed and implemented protocols that are significantly less complex than traditional replicated services, yet provide many of their benefits. The distributed systems aspect of our project is the design of protocols with which clients and servers can maintain apparently consistent information without server-to-server communication. The protocols we have developed for our system have the following features: Simplicity. The basic India protocols are easy to understand and implement. The enhanced protocols (see chapter 3 introduce some added complexity, but do not change the underlying nature of the protocol.) Scalability. The computational complexity of the system (amount of computation, and the number of messages required for each operation) scales linearly with the number of clients and servers. In addition, nearly all of the computation is performed by the clients, so each server can support a large number of clients. (We expect that servers will be bandwidth-limited rather than CPU-limited, and our benchmarks support this belief.) Our prototype system, implemented in Java, demonstrates the feasibility of this approach.