Barry Clifford Goldstein

Directions in cooperative processing between workstations and hosts

Advancements in technology have provided us with the availability of high-peformance processors from the high end of computing to the personal computer. In addition, technology growth has enabled us to envision sixteen megabytes of real storage for a personal computer. As a result, we have witnessed not only a tremendous growth at the high end of the computing spectrum, but also the development of sophisticated personal computers (e.g., the IBM PC XT/370) with real storage capacities approaching those of high-end computers of a decade ago. This growth at both ends of the computing spectrum has given us a choice. We can either allow a clean separation to grow between personal computer and host or provide a means by which they cooperate in providing quality service to the user without the complexity normally associated with high-end systems. This paper explores what such a cooperation could mean.

Data communications: the implications of communication systems for protocol design

The construction of a communication network architecture, specifying protocols by which systems communicate, is a complex art. Much has been written about the optimal protocols for theoretical models of systems. This paper points out that protocol design must depend on the “nuts and bolts” of the systems which implement the protocols. Numerous examples are provided to support this thesis. The paper also briefly discusses other issues that influence protocol design and draws lessons for standards activities.

A machine-oriented resource management architecture

An architecture for resource management in a computer system is presented. The architecture is a subset of an APL-like higher level language machine architecture. The approach provides for a natural division of function between software and microcode/hardware. Any object in the system can be managed as a resource, and the fact that an object is managed can be transparent to the program using the object. Finally the resource management concepts are related to the current resource management problems of effective third-party control of resources, intelligent work scheduling, and deadlock resolutions.