Analysis of preventive maintenance in transactions based software systems

Abstract

Preventive maintenance of operational software systems, a novel technique for software fault tolerance, is used specifically to counteract the phenomenon of software "aging." However, it incurs some overhead. The necessity to do preventive maintenance, not only in general purpose software systems of mass use, but also in safety-critical and highly available systems, clearly indicates the need to follow an analysis based approach to determine the optimal times to perform preventive maintenance.In this paper, we present an analytical model of a software system which serves transactions. Due to aging, not only the service rate of the software decreases with time, but also the software itself experiences crash/hang failures which result in its unavailability. Two policies for preventive maintenance are modeled and expressions for resulting steady state availability, probability that an arriving transaction is lost and an upper bound on the expected response time of a transition are derived. Numerical examples are presented to illustrate the applicability of the models.

Optimal Trigger Time of Software Rejuvenation under Probabilistic Opportunities

“…Under these opportunity processes, we find the existence condition on the optimal trigger time of rejuvenation in an existing software rejuvenation model proposed by [5]. Note that the opportunity-based software rejuvenation policy is a generalization of the rejuvenation policies discussed in [9] and [10].…”

“…The OTSRJ policy is similar to opportunity-based rejuvenation policies discussed in [9] and [10], where the rejuvenation is performed when all the buffered tasks are completed in a transaction-based system. Under the OTSRJ policy, the state transition diagram can be depicted as Fig.…”

“…Garg et al [9] dealt with such problems by considering two different rejuvenation policies in a stochastic model for transaction-based system. They proposed the two rejuvenation policies; the rejuvenation is immediately performed at a scheduled time interval (Policy I) and the rejuvenation is started at which a queue of arriving jobs is empty (Policy II).…”

An Adaptive Cost-Based Software Rejuvenation Scheme With Nonparametric Predictive Inference Approach

“…Pfening et al [32] consider a server-type software system with degradation as a queueing system, and formulate as the Markov decision process. Garg et al [21] take account of the presence of system failure caused by software aging and analyze the time-based optimal software rejuvenation schedule. This model is extended latter in Okamura et al [27][28][29][30] by introducing the workload-based rejuvenation schedule and/or a more general arrival process of transactions.…”

Prediction-Based Software Availability Enhancement

“…These models operate on a very high level of abstraction and are generated in an analytical manner. An analytical model for transaction processing systems has been presented by Garg et al in [2]. Due to aging, the service rate of the system in question decreases over time and the software itself experiences hangs and crashes which result in unavailability.…”