Fault Tolerance via Endocrinologic Based Communication for Multiprocessor Systems

Abstract

The communication mechanism used by the biological cells of higher animals is an integral part of an organisms ability to tolerate cell deficiency or loss. The massive redundancy found at the cellular level is fully taken advantage of by the biological endocrinologic processes. Endocrinology, the study of intercellular communication, involves the mediation of chemical messengers called hormones to stimulate or inhibit intracellular processes.

This paper presents a software model of a multiprocessor system design that uses an interprocessor communication system similar to the endocrine system. The feedback mechanisms that govern the concentration of hormones are mimicked to control data and control packets between processors. The system is able to perform arbitrary dataflow processing. Each processing stage within the system is undertaken by a separate group of microprocessors. The flow of data, and the activation of the next stage within the process is undertaken using the bio-inspired communication technique. The desired result is a system capable of maintained operation despite processor loss. The feasibility of the multiprocessor system is demonstrated by using the model to perform a simple mathematical calculation on a stream of input data.