Comparison And Improvement Of Proactive Software Self-adaption Methods

Self-adaptation is a promising approach to enabling software systems to address the challenge of uncertainty.Different from traditional reactive adaptation mechanisms that focus on the system's current environment state only,proactive adaptation mechanisms predict the potential environmental changes and make better adaptation plan accordingly.Proactive Latency-aware Adaptation(PLA for shot)and Controlbased Requirements-oriented Adaptation(Cob RA for short)are two representative approaches to build proactive self-adaptation mechanisms.Despite their different design and implementation details,PLA and Cob RA are reported to have a very similar performance in supporting self-adaptation.In this paper,we first perform an in-depth investigation of the methodology of proactive self-adaptation approaches.We divide a proactive software self-adaption mechanism into three modules,namely system modelling,environment predicting,and uncertainty filtering.Starting from these three modules,we make a detailed comparison of two proactive self-adaption methods: PLA and Cob RA and take the Rice University Bidding System(RUBi S)system as an example to analyze in detail the similarities and differences between the two approaches' implementations.Secondly,we conduct an empirical study on the role of the different modules of PLA and Cob RA in supporting the self-adaption effect.We perform an ablation study on the three modules of PLA and compared their performance with Cob RA in different scenarios.The experimental results in this paper show that inaccurate system modelling can be compensated by the uncertainty filtering module to a certain extent,and the environment predicting module's support for adaption effectiveness is not as good as system modelling and uncertainty filtering.Thirdly,based on this observation,we propose a new self-adaptation mechanism,MAPE-U(monitoring,analyzing,planning,executing with uncertainty).MAPE-U mechanism,which is designed based on “separation of focus”,promote uncertainty handling as a first-class element in a self-adaptation mechanism.Specifically,MAPEU defines compensation terms for describing the impact of uncertainty and thus optimizing its adaptation strategies under different situations.MAPE-U also simplifies its environmental prediction module for reducing its time and space cost,considering that environmental prediction is less effective according to our empirical study.Our experimental evaluation shows MAPE-U's effectiveness in supporting self-adaptation for RUBi S and SAVE,comparing with PLA and Cob RA.