Sequencing computational events in heterogeneous distributed systems

Distributed systems are growing in number, size, and complexity. Some technological advances have been made to program these systems, most notably the remote procedure call. However, the nature of heterogeneous distributed systems allows for much more complex interactions and new programming technologies must be developed to properly and effectively program these systems. This report develops the use of high-level algorithms that run on the virtual machine made available by the interconnection of the individual machines. These algorithms are specified using traditional programming technologies. Small stubs are then generated that contain an instance of an emulator for the virtual machine. By ensuring a consistent run-time environment for the algorithms regardless of architectural and linguistic differences, and by maintaining the state of the running algorithm in a persistent entity which travels the system, complex algorithms can be specified and executed hierarchically above the applications that form the system. Additionally, by keeping the state of the running algorithm out of the applications which provide the resources for the virtual machine, the application functions that are referred to in these algorithms can often be idempotent. A version has been in place in the Utah Retrieval System Architecture project since 1985.