A Novel Family of Queuing Network Models for Self-adaptive Systems

A Self-adaptive System (SaS) consists of an autonomic manager which is able to adapt the system’s behavior by operating on a managed sub-system that perceives and affects the environment through its sensors and actuators, respectively. Self-adaptation may occur at different levels, devising a number of knobs that the autonomic manager can properly regulate in order to produce actuation in response to environment sensing. This paper is an extension of our previous work introducing a generalized QN model that allows performance modeling and assessment of SaSs. We here extend previous work by defining modeling patterns and controller selection policies to conform to during the instantiation of the generalized model, resulting into a novel family of QN models aimed at representing the different parts of the system and the dynamics occurring over and among them. A controlled experiment addressing a realistic SaS for emergency handling shows that, by adhering to the defined patterns and controller selection policies, QN models behave as expected, and that the latter can be immersed into a performance optimization context that opens to the development of automated solutions to support the identification of efficient system configurations.