Fault Tolerance on Symmetric Multiprocessor Machines (93597)



The information in this article applies to:

  • Microsoft Windows NT Server 3.1
  • Microsoft Windows NT Workstation 3.1

This article was previously published under Q93597

SUMMARY

A symmetric multiprocessor (SMP) machine is designed to provide fault tolerance. If one processor fails, the SMP machine continues processing threads and keeps the system going using operative processors. Windows NT supports SMP platforms to increase processor power, not for fault tolerance. If a processor fails, Windows NT stops responding ("hangs"). This is not due to a flaw in Windows NT but in the hardware support on 80x86-based SMP machines. At this time, no 80x86-based SMP machine provides the hardware support to recover from a failed processor.

The reliability of a multiprocessor (MP) machine compared to a uniprocessor (UP) machine is difficult to determine. Although a single processor failure can bring the system down, processor failures are rare. The benefit of several times the processing power most likely outweighs the cost of possible system failure.

Modification Type:MajorLast Reviewed:10/30/2003
Keywords:kbother KB93597