Component-ware for Autonomic Supervision Services The CASCADAS Approach

This paper describes two complementary mechanisms for the supervision of large scale and highly distr ibuted systems structured as a cloud of autonomic computing components. The fi rst one is based on the creation of supervision pervasions, for the supervision of clusters of components (i.e., aggregates structurally organized through one or several contracts) implementing specific services in accordance to service-specific management policies. This mechanism is designed as a supplementary service that can be requested by operational components and is structured as an ensemble of self-contained objects that implement an autonomic control loop, which does not require any a priori knowledge on the structure of the supervised system. The second mechanism promotes supervision logic embedded in the autonomic components, which exploit autonomic features and cooperate through dedicated protocols over self-organized overlay networks; this mechanism is suitable for supervising infrastructural (service-independent) functions of autonomic components, and thei r aggregates. The main contribution of the paper is to define those two mechanisms and to shown that they are complementary, and can be combined to achieve cross-layer supervision.