Autonomic Adaptation of Publish/Subscribe Middleware in Dynamic Environments

Quality-of-service (QoS)-enabled publish/subscribe (pub/sub) middleware provides powerful support for scalable data dissemination. It is hard, however, to maintain key QoS properties (such as reliability and latency) in dynamic environments for distributed real-time and embedded (DRE) systems (such as disaster relief operations or power grids). Managing QoS manually is often not feasible in dynamic environments due to slow response times, the complexity of managing multiple interrelated QoS settings, and the scale of the systems being managed. For certain domains, DRE systems must be able to reflect on the conditions of their environment and adapt accordingly in a bounded amount of time. This paper describes our architecture of QoS-enabled middleware and corresponding algorithms to support specified QoS in dynamic environments. The results of our work show that the middleware can monitor and adjust to a changing environment while addressing timeliness concerns of DRE systems. In particular, the middleware can autonomically determine and implement appropriate adaptations in microseconds.