Self-Tuning of Software Systems Through Goal-based Feedback Loop Control

Abstract

Quality requirements of a software system cannot be optimally met, especially when it is running in an uncertain and changing environment. In principle, a controller at runtime can monitor the change impact on quality requirements of the system, update the expectations and priorities from the environment, and take reasonable actions to improve the overall satisfaction. In practice, however, existing controllers are mostly designed for tuning low-level performance indicators rather than high-level requirements. By linking the overall satisfaction to a business value indicator as feedback, we propose a controltheoretic self-tuning method that can dynamically adjust the tradeoff decisions among different quality requirements. A preference-based reasoning algorithm is involved to configure hard goals accordingly to guide the following architecture reconfiguration.

Modern software cybernetics: New trends

“…In practice, however, existing controllers are mostly designed for tuning low-level performance indicators rather than high-level requirements. Peng et al (2010) combined goal models with feedback loop controllers to make dynamic trade-offs among conflicting soft goals (i.e., the goals with no binary satisfaction criteria). Reflecting the business value of customers, the controller adjusted the preference ranks of soft goals on the basis of runtime feedback.…”

Decisions, Models, and Monitoring -- A Lifecycle Model for the Evolution of Service-Based Systems

“…for supporting dynamic configuration of distributed systems. At the requirements level, Peng et al [20] address the selfconfiguration of software systems by introducing a formal reasoning procedure at runtime for supporting dynamic quality trade-off among alternative OR-decomposed goals.…”

“…Then, the algorithm creates an associative array mapping available actions to a utility value incorporating the action's execution costs as well as the importance of currently unsatisfied goals (lines 6, 11-18). The most valuable action a max is added to the set of actions A to be executed, and goals G ac amax that are achieved by a max are removed from the set of unfulfilled goals G u (lines [20][21][22][23]. If no actions are found to achieve a goal g, a system-provided escalation action a e g is added to the set of actions A to be executed to indicate the need for operator intervention to satisfy goal g (lines 7-10).…”

Requirements-Driven Qualitative Adaptation

“…Several approaches in the literature propose adaptation through reconfiguration, i.e., switching the system's behavior by finding a new configuration for its parameters. Wang & Mylopoulos [26] propose algorithms that suggest a new configuration without components that have been diagnosed as responsible for a failure; Nakagawa et al [19] developed a compiler that generates architectural configurations by performing conflict analysis on goal models; Fu et al [11] use reconfiguration to repair systems based on an elaborate state-machine diagram that represents the life-cycle of goal instances at runtime; Peng et al [20] assign preference rankings to softgoals and determine the best configuration using a SAT solver; Khan et al [16] apply Case-Based Reasoning to find the best configuration; Dalpiaz et al [5] propose an algorithm that finds all valid variants to satisfy a goal and compares them based on their compensation/cancelation cost and benefit (e.g., contribution to softgoals).…”