Source of Acquisition NASA zyxw Goddard Space Flight Center Joaquin Peiia Michael G. Hinchey Roy Sterritt NASA Goddard Space Flight Center University of Seville University of Ulster j oaquinp@us.es MichaeI.G.Hinchey@nasa.gov r.sterritt@ulster.ac.uk Spain USA Northern Ireland Antonio Ruiz-CortCs University of Seville aruiz@us.es Abstract zyxwvuts Autonomic Computing (AC), self-managementbased on high level guidance fiom humans, zyxwvut is increasingly gain- ing momentum as the way forward in designing reliable systems that hide complexiv and conquer IT management costs. Effectively, AC zyxwvuts may be viewed as Policy-Based Self- Management. The Model Driven Architecture (MDA) ap- proachfocuses on building models that can be transformed into code in an automatic manner In this papel; we look at ways to implement Policy-Based Self-Management by means zyxwvutsr of models that can be converted to code using trans- formations that follow the zyxwvutsrq MDA philosophy. We propose a set of UML-based models to spec& autonomic and au- tonomous features along with the necessary procedures, based on mod$cation and composition of models, to deploy a policy zyxwvutsr as an executing system. 1 Introduction and Motivation Autonomic Systems (encompassing both Autonomic Computing and Autonomic Communications) is an emerg- ing field for the development of large-scale, self-managing, complex distributed computer-based systems zyxwvuts [ 11. In intro- ducing the concept of Autonomic Computing, IBM’s Paul Horn likened the needs of large scale systems management to that of the human Autonomic Nervous System (ANS). The zyxwvutsrqp ANS, through self-regulation, is able to effectively monitor, control and regulate the human body without the need for conscious thought [5]. As in all emerging fields, there are many fruitful areas for concern, that are worthwhile targets for research and development. Many issues are yet to be addressed, such Manuel Resinas University of Seville resman@tdg.lsi.us.es as, for example, how autonomic managers, which together with the component being managed make up an autonomic element, should be designed in order to exist in a collabo- rative autonomic environment, and ultimately provide self- management of the system to the highest degree possible. The long-term strategic vision of AC highlights an over- arching self-managing vision where the system would have such a level of “self’ capability that a senior (human) man- ager in an organization could specify business policies- such as profit margin on a specific product range, or sys- tem quality-of-service for a band of customers-and the computing and communications systems would do the rest themselves. The main idea behind a Model-Driven Architecture is separation of the specification of the operation of a system from the details of the way that system uses the capabili- ties of its platform. With the purpose of abstracting away platform details, MDA involves two main types of models. The platform-independent model (PIM) which provides a model of the system without platform details, and the plat- form specific-model (PSM), which is obtained by means of transformation of the PIM model. We propose an MDA approach for applying policies to autonomic systems. This avoids platform-dependent details unnecessary at the level of abstraction of a policy, and em- ploys transformations of models to bring the policy through the necessary levels to be transformed into an implementa- tion [8]. This is based on our previous work applying an Agent-Oriented methodology called MaCMAS (Methodol- ogy Fragment for Analyzing Complex Multi-Agent Sys- tems) to model, specify and deploy policies at runtime [ 151. In this paper, we extend this work to, using an MDA ap- proach, automate the process of applying a new policy. Es- sentially, we propose UML-based PIMs for specifying au- tonomous and autonomic properties of the system and an 1 https://ntrs.nasa.gov/search.jsp?R=20070005050 2019-05-01T01:53:17+00:00Z