Search

A loosely coupled multiprocessor system contains multiple processors which have their own local memories. To balance the load among multiple processors is of fundamental importance in enhancing the performance of such a multiple processor system. Probabilistic load balancing in a heterogeneous multiple processor system with many job classes is considered in this study. The load balancing scheme is formulated as a nonlinear programming problem with linear constraints. An optimal probabilistic load balancing algorithm is proposed to solve this nonlinear programming problem. The proposed load balancing method is proven globally optimum in the sense that it results in a minimum overall average job response time on a probabilistic basis.

A new language construct, called molecule, is described for the efficient implementation of algorithms on parallel computers. A molecule can be considered a procedure associated with a molecule type. Each molecule type characterizes a particular computation mode (sequential, pipelining, array processing, dataflow, multiprocessing, etc.). Basic concepts of molecule are introduced with a procedural language, called PAL. A concrete example is presented to illustrate layered software development using PAL on a multicomputer (the iPSC). It is concluded that high-level languages, augmented with the molecule construct, offer application flexibility, user friendliness, and efficiency in implementing parallel programs. >

This paper presents a Markov chain model to analyze the performance of shared-resource array processors for multiple vector processing. Such a parallel processor contains multiple control units sharing a resource pool of processing elements and operating with multiple single-instruction multiple-data streams (MSIMD). In the steady state, the Markov model corresponds to a two-dimensional Markov chain, which can be expressed by a set of equilibrium equations. An iterative method is developed to solve the Markov chain after projecting the equilibrium equations onto a one-dimensional state space. The convergence rate of the iterative method can be greatly enhanced by choosing starting values corresponding to the approximated analytical results obtained earlier by the authors.

In the above paper1, an error was made in the Load Balancing Algorithm. A more clear recursive way to present this algorithm is to modify steps S3 to S5 as follows.