Toward Automatic, Run-Time Fault Management for Component-Based Applications

Although components are well tested and intrinsically more reliable than custom-made software, applications built of components generally lack global verification (especially if they are independently extensible) and are subject to distributed failure scenarios. We discuss a simple component framework based on a data flow architecture that addresses these problems and allows for building failure-resilient applications. We illustrate the characteristics that make this architecture particularly suitable for automatic fault management and explain the mechanisms we use for detecting, diagnosing and correcting faults at run-time.