Framework for ensuring runtime stability of control loops in multi-agent networked environments

The idea of autonomic computing, and accordingly autonomic networking, has drawn the attention of industry and academia during the past years. An autonomic behavior is widely understood as a control loop which is realized by an autonomic entity/agent that manages some resources, in order to improve the performance and regulate diverse operational aspects of the managed network or IT infrastructure. Self-management, realized through autonomic behaviors, is an appealing and dangerous vision at the same time. On one hand, it promises to reduce the need for human involvement in the network and system management processes. On the other hand, it bears a number of potential pitfalls that could be even dangerous to the network, the IT infrastructure, and the corresponding services. One of these pitfalls is constituted by the stability of the control loops, and correspondingly by the interference among multiple autonomic agents operating in parallel. In this paper, a novel approach to ensuring runtime synchronization and stability of multiple parallel autonomic control loops is presented. We formally model the problem of runtime action synchronization, propose different possible solutions, and provide a case study, as well as different performance measurements based on a prototype that implements our approach.