Performance analysis of a controlled database unit subject to decision errors and control delays

Abstract

This paper extends the performance analysis of a controlled database unit to include the cases where errors and delays can occur in state-based control actions as a result of uncertainty in the knowledge of the system state. The paper details the way such errors and delays are captured through augmenting the state space in the Markov model of the database unit. State variable feedback is used to activate the process of restoration upon the failure of one of the database servers in the unit. The performance of the database is evaluated in terms of the resulting mean time to unit failure, the steady state availability, the expected response time, and the service overhead of the database unit. All performance measures are examined with respect to the likelihood of decision error and the amount of control action delay