Distributed Programming with Objects and Threads in the Clouds System

The CLOUDS operating system supports a distributed environment consisting of compute servers, data servers and user workstations. The resulting environment logically simulates an integrated, centralized computing system. In addition, CLOUDS supports a programming paradigm that makes distributed programming simpler. Distributed programs can be written in a centralized fashion and yet they can exploit parallelism and distribution at runtime. The system paradigm is based on an object/thread model. The basic building blocks for applications are persistent memory (called objects) and computation (called threads). Unlike most systems, CLOUDS separates the notion of memory from computation. Programming environments based on these abstractions, though unconventional, provide powerful tools for composing applications that exploit concurrency and distribution. * This work was supported in part by NSF contract CCR-8619886. @ Computing Systems, Vol. 4 . No. 3 . Summer 1991 243 This paper discusses programming techniques that use persistence and distribution of memory. The examples show how separation of computation from memory can be used to the programmer's advantage. We also present a distributed programming technique called implicit distributed programming. The implementation details of the programming support subsystems are presented. The system performance measurements demonstrate the usability of CLOUDS as a distributed programming platform.