A framework for Performance Prediction of Service-Oriented Architecture

Abstract

Service-Oriented Architecture (SOA) considered as one of the important architectural styles to build future applications. This architecture consists of a group of homogenous and autonomous components that interact with each other to accomplish a task. However, performance prediction of SOA based applications is regarded one of the complex tasks that face software developers and engineers. This paper presents a novel approach for SOA performance prediction at early stages of SDLC by using Machine Learning technique. Firstly, annotated UML diagrams are presented. Secondly, translate the UML diagrams into (QNM) model in order to extract performance indices such as Response Time, Throughput, and Utilization. Finally, machine learning technique used to predict the application model performance. The prediction result "Risk" means design does not meet customer requirement and "No Risk" means the design satisfies the customer requirements. Machine learning technique predicts the performance based on the training set against the extracted test set of the application model performance indices. The new method has many advantages, such as reducing time, scale with large system size, and avoiding problems before the service put into the production environment. To illustrate our approach, we present the results of a simple practical example.

Mobile Agent Based Distributed Network Architecture with Map Reduce Programming Model

“…Our simulation model was meant to test performance improvement after introducing agents into the existing SAF and OSD models and also sorting metadata using map reduce approach. The performance measures have been described by [25,90,50] and [85] which were adopted as a performance measure for our SABSA engine to provide the metrics that were used for this research.…”