A Tradeoff Approach for the Scalability Issue in Large Self-Managed Networks

Autonomic systems for network management are available and present significant benefits such as the reduction of human intervention, performance improvement and optimization of available resources. However, large networks composed by hundreds of nodes (routers) present a challenging scalability issue in relation to the development and implementation of self-management solutions. In effect, computational complexity required for autonomic solutions increases with network size resulting, typically, in unacceptable response time for an effective near real-time set of management action implementations. Alternatives that may be explored to deal with the scalability problem include the increase of computational resources supporting the autonomic management solution and the reduction of managed network size and, consequently, computational complexity. The first approach is realizable but implies, typically, in a significant cost increase for the autonomic solution. The reduction of managed network size, whenever realized preserving the quality of the overall autonomic solution, may represent an important tradeoff for managers in the direction of realizable and affordable autonomic management solutions. This paper proposes the systematic use of two parameters, robustness and performability, to deal with the scalability issue in large self-managed networks. These parameters will support managers in implementing self-management solutions by considering the tradeoff between solution's quality and execution time. Robustness and performability are, in turn, parameters resulted from a network partitioning method based on network density.