Automated multi-objective control for self-adaptive software design

Abstract

While software is becoming more complex everyday, the requirements on its behavior are not getting any easier to satisfy. An application should offer a certain quality of service, adapt to the current environmental conditions and withstand runtime variations that were simply unpredictable during the design phase. To tackle this complexity, control theory has been proposed as a technique for managing software's dynamic behavior, obviating the need for human intervention. Control-theoretical solutions, however, are either tailored for the specific application or do not handle the complexity of multiple interacting components and multiple goals. In this paper, we develop an automated control synthesis methodology that takes, as input, the configurable software components (or knobs) and the goals to be achieved. Our approach automatically constructs a control system that manages the specified knobs and guarantees the goals are met. These claims are backed up by experimental studies on three different software applications, where we show how the proposed automated approach handles the complexity of multiple knobs and objectives.

Automated control of multiple software goals using multiple actuators

“…Recent work automatically synthesizes MIMO controllers for discrete systems by chaining multiple SISO controllers (one for each goal) together in a hierarchy [16]. The hierarchy reflects goal prioritization, and each controller produces a continuous reference signal that is converted into a mixture of the discrete input configurations using Pulse Width Modulation [36].…”

“…Actuators that are used to reach a higher priority goal cannot be changed to meet the lower priority ones, limiting the controller's ability to achieve all goals optimally at the same time (see also the experimental comparison in Section 5.3). The need for finite domains for the actuators requires the discretization of continuous control inputs; while this can be done automatically, as in [16], the complexity of the control law may grow exponentially, limiting the practical applicability of the approach when timely decisions are required or the controller has to run on low-power devices, like embedded systems. The approach in [49] extends [16], formulating the conversion of the continuous references into discrete settings as a linear optimization problem.…”

“…Very little research has addressed automating the design of selfadaptive software that meets multiple goals. A first step proposes a hierarchy of single-goal controllers [16,24]. In this approach, goals are ordered.…”

“…Thus, there is a need for a formally verifiable methodology to automate the design of self-adaptive software that meets multiple goals using multiple actuators. Finally, the approach in [16,24] requires actuators to assume only a finite number of values. While continuous actuators can be automatically discretized [16], the complexity of control may grow exponentially, limiting its applicability when timely decisions are required or when the available computational capacity is limited (e.g., IoT).…”

An Iterative Decision-Making Scheme for Markov Decision Processes and Its Application to Self-adaptive Systems

“…These two approaches rely on the Bayesian (point) estimation method while IDMS exploits both point and interval estimates from the frequentist statistics theory. Finally, Filieri et al [17] constructed approximate dynamic models of a self-adaptive system and for synthesizing, from those models, a suitable controller that guarantees prescribed multiple non-functional system requirements. The method they used is from control theory, which is quite different from the theory of MDPs.…”