A survey of fault tolerance and reliability techniques in hypercubes

الملخص EN

Massively parallel computers, using thousands of processors, will be the future trend for producing tremendous computational power.

However, in the current technology, if one processor fails, the entire system may fail.

A major drawback of hypercubes is that a single processor failure may destroy the whole network.

The existence of a large number of components in such systems makes them subject to failures.

As the probability of any one or more processors failing in such a complex system is large, building some fault-tolerance feature into them becomes extremely important.

Fault tolerance in highly parallel computers is important for achieving reliable high-performance computing.

The problem of tolerating faulty processors or links in hypercubes has been studied by many researchers, either by using spares or by reconfiguration.

This manuscript is mainly a chronological survey of fault tolerant issues and reliability evaluation of hypercubes.