When Distributed Hash Tables Meet Chemical Programming for Autonomic Computing

With the rise of Service Computing, applications are more and more built as temporal compositions of autonomous services, in which services are combined dynamically to satisfy constantly arriving users´ requests. These applications run on top of web-based, large, unreliable, and heterogeneous platforms, in which there is a high demand for autonomic behaviours, such as self-optimisation, self-adaptation, or self-healing. Chemistry-inspired computing consists in envisioning a computation as a succession of implicitly parallel, distributed and autonomous reactions, each reaction consuming molecules of data to produce new ones, until the state of inertia, where no more reactions are possible. This vision of a computation makes it a promising candidate to inject autonomic behaviours in service computing. However, while the benefits of such a model are manifold, its deployment over large scale platforms remains a widely open issue. In this paper, after identifying the main obstacles towards such a deployment, we propose a peer-to-peer framework able to execute chemical specifications at large scale. It combines distributed hash tables and algorithms for the atomic capture of molecules, and proposes an efficient method for inertia detection, which is a critical problem, in particular when addressed in a large scale environment. The sustainability of the framework is established through a complete complexity analysis.