Analyzing the effectiveness of fault-management architectures in layered distributed systems

Fault management infrastructure in distributed systems includes manager processes and agents with various kinds of interactions for monitoring and surveillance of the status of the application software and hardware. The system architecture now includes these additional components and interactions, and they affect the system availability. This paper describes an architecture model called MAMA (Model for Availability Management Architecture) with an architecture definition language MAMA-dl for the combination of the application and management parts, and its analysis. The analysis extends the Fault Tolerant Layered Queueing Model to account for propagation of knowledge of the system state in the management sub-architecture. The model is demonstrated on a problem of placement of manager tasks in a system.