Cooperative Computing in Dynamic Environments MIT9904-12
Nancy Ann Lynch, Idit Keidar · 2002
Cooperative Computing group at NTT on developing models and analysis methods for distributed systems, with a focus on cooperative group activities in networks. Such group activities range from human social activities in cyber communities to powerful distributed applications involving data sharing and cooperative work. These activities are often supported by agent communication services, which provide distributed intelligence, or by group communication services (GCSs), which manage group membership and guarantee coherent communication. The environments in which such activities take place are highly dynamic: participants come and go (and change location), network topology changes, and components fail and recover. Coping with such difficult environments leads to complex implementations, which are difficult to build, understand, and analyze. This project addresses these problems using formal modeling and verification techniques, in particular, a combination of Input/Output automaton methods used at MIT and process algebraic and knowledge-based methods used at NTT. This involves extensions to the existing techniques, for example, extending I/O automata to allow dynamic process creation and destruction. As the basic framework is developed, it is being applied to a collection of typical examples from cooperative computing applications, including computer-supported cooperative work, e-commerce, and distributed databases. Other issues being studied include analysis of performance and fault-tolerance properties, and connecting the formal models with actual runnable code.