Self-Adaptation of Software Using Automatically Generated Control-Theoretical Solutions

Abstract

Control theory has contributed a set of foundational techniques to handle "change" at runtime in software applications. These techniques however have fundamental limitations as well: (i) they require the development and understanding of mathematical models; (ii) synthesizing solutions is often done on a per-problem basis, discouraging flexibility and generality. Software engineering, as a discipline, has always aimed at finding reusable and modular solutions. The combination of the desire to apply formally grounded control-theoretical principles and reuse existing solutions has motivated research on the topic of automatically generated control solutions. This research aims at designing control strategies in an automated way from data that qualifies the given problem at hand. This chapter provides an overview of the research topic of automatically generated control-theoretical solutions, explaining the key research contributions and paving the way for future research.

K8-scalar

“…The most well-known suite for evaluating the performance of databases is YCSB [11]. Although this benchmark suite supports different workload types, it is not designed to test the performance of advanced self-adaptive approaches such as those based on control theory [17,42] and machine learning [34] that can take into account the history of seasonal or oscillating workloads. Many tools and approaches have been proposed to automate scalability testing [2,10,23,25,43,45].…”