Change programming in distributed systems

Distributed programs consist of multiple processes that cooperate by message passing to fulfil some global task. An interesting extension is the provision of dynamic changes for distributed programs as a means to establish evolutionary systems. The author models the development of distributable programs using the configuration programming approach (J. Kramer, 1990), that is developed further in the Esprit II project REX on reconfigurable and extensible parallel and distributed systems. A key principle of this approach is the separation of the functional description of individual process behaviour from the description of system structure viewed as a set of processes and their interconnections. The author presents fundamentals of the distributed system model followed by an explanation of programming notation, as used for the distinct levels of configuration and process programming. He describes two distinct views of reconfiguration states. Then all important notions according to a change programming approach are introduced. He shows a detailed example and briefly discusses some correctness aspects. The author also evaluates the proposed model against that of J. Kramer, J. Magee (1990) and an outline of future research work is given.<
>