Distributed Process Creation Within a Shared Data Space Framework

This paper describes the design and implementation of a remote process instantiation mechanism which is consistent with the Linda paradigm and semantics of the EVAL operation, and which uses shared data space as the medium for passing process and environment parameters. The motivation fo such an implementation stems from our effort to rehost a uniprocessor version of the Linda computational system to a network of workstations. The baseline version relies on the semantics of the UNIX fork() system call to create processes and to pass the proper execution parameters to them. In creating a distributed version of the Linda environment, two major issues are addressed: (1) how to instantiate a remote process that knows where, among several possibilities, to begin its execution, and (2) how to communicate the proper runtime values of relevant variables to each new remote Linda process. Guiding our implementation was a desire to employ existing interprocess communication facilities, e.g., shared data space, to pass process creation and execution parameters.