Comparison of the Performance of Two Service Disciplines for a Shared Bus Multiprocessor with Private Caches

TL;DR

This paper compares two analytical models for cache coherence overhead in shared bus multiprocessors with private caches, highlighting the effectiveness of priority discipline as a lower bound through numerical analysis.

Contribution

It introduces and compares two analytical models based on different service disciplines for evaluating cache coherence overhead in shared bus multiprocessors.

Findings

Priority discipline serves as a lower bound for performance.

Numerical results illustrate differences between the models.

Analytical models aid in understanding cache coherence impacts.

Abstract

In this paper, we compare two analytical models for evaluation of cache coherence overhead of a shared bus multiprocessor with private caches. The models are based on a closed queuing network with different service disciplines. We find that the priority discipline can be used as a lower-level bound. Some numerical results are shown graphically.