Study On Environment State Oriented ASBS Adaptation Strategy Evaluation Methods

In order to keep system operating effectively and stably in dynamic open evironment, enhance Service Based Sofware system (SBS) with adaptive ability and construct Adaptive Service Based Software system (ASBS), has been put forward as an important paradigm for distributed software system. As policy driven adaptation is widely adopted in ASBS, environment state oriented ASBS adaptation strategy evaluation, which is a primarily basis in ASBS adaptation strategy design, has received lots of attention from the field of ASBS research.Model based evaluation method has been commonly recognized as a practical method for design time software evaluation. However, existing model based evaluation method face many problems in ASBS strategy evaluation. At first, the effection on Web service caused by the environment is usually neglected, which to the absence of Web service environment in system description model. Secondly, the policy driven adaptation of ASBS which adjust the system itself during execution brings great challenge to system behavior modelling and strategy evaluation. Finally, there are very few researches focus on using strategy evaluation to improve strategy adaptation design process. As a result, the actual availability of exiting evaluation methods in ASBS adaptation strategy evaluation is limited.To handle the defect of existing model based evaluation methods and tackle the problem in ASBS adaptation strategy evaluation, environment state oriented ASBS adaptation strategy evaluation method is studied in this thesis. Environment state oriented Web Service description model is proposed to describe Web service with consideration of the environment it deployed, based on which, a environment state oriented ASBS description is provided. By separating the adaptation behavior and business behavior, a reflective Petri net based evaluation model is put forward with modelling process provided and ASBS adaptation strategy evaluation method based on which as well. With the evaluation method, an artificial bee colony algorithm based adaptation strategy recommendation method is proposed to facilitate the ASBS adaptation strategy design process. General achievements of this dissertation are presented as follows,(1) Environment state oriented Web service non-functional property description model EWnFM. Considering the difference of Web Service non functional properties under different environment states, EWnFM builds a connection between Web service non-funcitonal properties and the environment it deployed, lays the foundation of environment state oriented ASBS. Model definition and parameter estimation method are introduced, based on which the EWnFM of Web service could automatically build using ASBS monitor data. Experiment results indicate that EWnFM has a better description accuracy comparing to other existing models.(2) Environment state oriented ASBS description. By introducing the definition of ASBS and Environment state oriented ASBS description, and extending existing Web service description language and adaptation policy language, environment state oriented Web service description language and ASBS adaptation policy description language are proposed to consiste Environment state oriented ASBS description model.(3) Reflective Petri net based ASBS evaluation model A-REM. In order to separate the different behaviors in ASBS behavior modelling, a reflective Petri net based ASBS evaluation model, A-REM, is proposed. The business behavior and adaptation behavior are modelled in different levels in A-REM independently, which brings the flexibility of modelling method and surportion of adaptation strategy evaluation. Model deninition and model built process from ASBS description are provided.(4) Environment state oriented ASBS adaptation strategy evaluation method. Based on the evaluation model provided above, an adaptation stratege evaluation method is proposed which evaluate the effect of the strategy both quantitatively and qualitatively. The definition of soundness of adaptation strategy is proposed and the validation method is provided to evaluate the qualitative evaluation of adaptation strategy. System mean response time and mean cost calculation and comparison between with and without strategies are provide to illustrate the evaluation non-functional properties of ASBS adaptation strategy. Experiment shows the validity of the method and lays the basis for ASBS adaptation strategy design.(5) ASBS adaptation strategy recommendation oriented artificial bee colony algorithm. Considering the difficulty in ASBS adaptation strategy design, an adaptation strategy recommendation method is proposed to alleviate strategy design problem. The recommendation method is realized by an Aritificial Bee Colony algorithm, which use food sources to model strategy recommendations, and mimic the behavior of bees to solve the optimization problem of finding recommendations. By introducing Configuration-Strategy Graph according to ASBS adaptation strategy feature, basic ABC algorithm is modified to escalate convergence speed and promote recommendation quality. Experiments show that the modification to ABC algorithm could apparently improve the basic recommendation method.To sum up, well-evaluated solutions for some critical issues in strategy evaluation of ASBS are presented in this dissertation. The contributions make a nice groundwork for future research and engineering on ASBS both in theory and in practice.