In this paper, we describe ASPEN, a concurrent stream processing environment. ASPEN is novel in that it provides a programming model in which programmers use simple annotations to exploit varying degrees and types of concurrency. The degree of concurrency to be exploited is not fixed by the program specification or by the underlying system. Increasing or decreasing the degree of concurrency to be exploited during execution does not require rewriting the entire program, but rather, simply re-annotating it.
Examples are given to illustrate the varying types of concurrency inherent in programs written within the stream processing paradigm. We show how programs may be annotated to exploit these varying degrees of concurrency. We briefly describe our implementation of ASPEN.
This work was done within the Tangram project, supported by DARPA contract F29601-87-C-0072.
Brian Livezey is now at the Lockheed Artificial Intelligence Center, 2710 Sand Hill Road, Menlo Park, CA 94025.