iNetLab: A Model-Driven Development and Performance Engineering Environment for Autonomic Network Applications

Abstract

A key software engineering challenge in autonomic computing is the complexity of administrating operational policies of applications. In order to address this challenge, this chapter proposes and evaluates a new development environment, called iNetLab, which is designed to improve the productivity of designing, maintaining and tuning operational policies in autonomic network applications. iNetLab consists of (1) a set of visual modeling languages specialized to define operational policies in network applications, (2) a set of supporting facilities for those modeling languages, and (3) tools estimates the performance of a network application with its operational policy under development. The proposed visual modeling languages and their supporting facilities can simplify and semi-automate the process to design and maintain operational policies by allowing application administrators (i.e., non programmers) to graphically deal with operational policies in an intuitive manner. The proposed performance estimation tools leverage the performance history of each network application (i.e., pairs of an operational policy and a performance result obtained in the past) and approximate the application's performance without deploying and running it actually. This simplifies the process to tune operational policies against desirable performance requirements and contributes to shorten the time to develop autonomic applications.