Fault tolerance cost effectiveness

Abstract

In switching applications, the implementation of fault tolerance has traditionally resulted in high hardware costs, by exclusively relying on proprietary hardware and by using monolithic recovery techniques external to the applications to achieve high quality service. The main advantages of this strategy were no unscheduled downtime, deferred maintenance, easy restart when needed, and easy growth and degrowth. While these objectives are still attractive, they can be achieved in a more cost-effective way by increased reliance on standard fault recovery components distributed closer to and inside the applications, and by using hardware recovery at the system level. A recent trend toward service customization and system distribution allows limiting the use of traditional recovery techniques where absolutely necessary to satisfy performance requirements.