A New Approximation Model for Guaranteed QoS Information Aggregation

The expansion of the networks in the last decade, in size and number of services, accentuated several problems. An important one is the management and the control of available resources, in particular, to provide QoS. To deal with this problem, networks are often structured hierarchically and each domain keeps its own detailed state and an aggregated view about the other domains. After aggregation, the amount of information is outstandingly reduced but we can not speak about relevance only if the distortion from the real network state is minimal. This distortion can result from a strict algorithm which leads to reduce unfeasible connections or a generous algorithm allowing to enlarge the amount of feasible connections even if the real network state does not allow it. Although it under-estimates the network capacities, the telecom operators are in favour of the first aggregation algorithm, because it guarantees better the quality of service. In this paper, we propose a new model for network state aggregation bringing a trade off between strict and generous algorithms. Contrary to traditional approaches where the probability of crank back is large, our solution, based on the concept of "supported services guarantee", does not over-estimate the network capacities and keep the admissibility amount reasonable. Our simulations show that the new approximation model achieves the best performance concerning the crank back reduction compared to the other aggregation methods.