Distributed channel allocation in ATM networks

Abstract

This paper proposes a distributed channel allocation mechanism for ATM networks. In this mechanism, agents are assigned to traffic sources and allocate channels to incoming calls in a distributed manner. Channel allocation is based only on the sampled channel utilization values locally available to agents. There is no direct exchange of information assumed between agents.Due to the distributed nature of the proposed mechanism, this mechanism causes oscillations during channel utilization when there is a delay associated with the sampling of channel utilizations and the processing of sampled values at agents. This oscillation increases the cell loss in a channel. According we propose methods for estimating channel utilization in order to reduce the oscillations and show their effectiveness through simulations.The proposed mechanism requires little computational processing, is simple t o implement, and achieves efficient channel allocation even if highly bursty calls. It is therefore suitable for ATM networks.

Clustered volatility in multiagent dynamics

“…From the right hand side of the last equation in the triplet above we can state this mathematically as s = f s on E r . Since this means that df str s dt = 0 (16) on that set, the agent population is macroscopically unbiased towards any particular population of strategies on such a set. In this way, if there exists a strategy distribution that stabilizes resource contention, the system can automatically adapt to that distribution and stabilize the entire system without any "preprogramming".…”

“…Studies of distributed systems have shown that when there is resource contention in the presence of delays, their dynamics is very complex, giving rise to nonlinear oscillations and chaos which drive the system far from optimality [13][14][15][16]. In some cases one can still observe a stable equilibrium .…”