Abstraction, decomposition/composition, modularity, and hierarchy are important principles of effective and efficient system design, but their application to new, more complex, and distributed applications is challenging. Using these principles, this paper describes a proposed design method for a class of control systems, called autonomic management systems (AMS)--defined as software that monitors the dynamics of physical or virtual entities and makes decisions to ensure required performance. Typically, an AMS is designed as a composition of individual manager/director modules, each addressing a particular aspect, for example, node failure, resource allocation, or mode change. This paper addresses the challenge of coordination and consistency of decisions and actions of multiple composed managers.

The proposed design method employs synchronous programming for reactive systems design for the managers, and discrete controller synthesis for coordination and consistency of the composed managers. The authors illustrate the proposed design method using a replication-based multi-tier hierarchical application--an abstraction of a four-tier web server application for dynamic content and database access. In general, these applications consist of a hierarchy of composed service tiers, each of which in turn consists of replicated servers. The design and implementation is supported by (academic) synchronous programming languages and discrete event synthesizer tools that enable static verification of coordination and consistency goals.

The paper builds on a base of very technical control system literature, but it is well organized and provides motivation and background material, explanations at several levels of detail, and discusses limitations and related and future work. It includes a comprehensive list of technical references.

The paper is of interest to the control system research community and to those interested in various control applications, for example, detection and avoidance in operating or network systems, coordination of behavior of independent devices, resource management, and, more generally, qualitative requirements such as reliability, availability, and scalability.

The research described in the paper is promising, but much work remains, including additional experiments, techniques for hierarchical behavior modeling, and abstraction to less specific types of systems and requirements. In addition, from the perspective of important system development principles, it gives the reader exciting expectations for future systematic methodologies for the development of new, more complex, and distributed applications, consisting of reusable, composable AMS modules, utilizing multiple paradigms (for example, control, synchronous, object based) and design patterns that ensure coordination, consistency, and coherence.