No abstract

In this paper we investigate the problem of scalable admission control for real-time traffic in sink-tree networks employing peraggregate resource management policies, like MPLS or DiffServ. Every traffic flow entering the network at an ingress node, and flowing towards a given egress node, specifies its leaky-bucket parameters and a required delay bound for traversing the network. We propose an algorithm that admits a new flow if a guarantee can be given that the required delay bound, besides those of other already established flows, are not exceeded. We identify properties of sink-tree networks based on which we considerably reduce the complexity of the proposed algorithm, and we show that the latter approaches the theoretical lower bound on the worst case complexity of any algorithm working under the same hypotheses. Finally, we show that the algorithm lends itself to a distributed implementation, thus allowing for better scalability.

No abstract

In a DiffServ architecture, packets with the same marking are treated as an aggregate at core routers, independently of the flow they belong to. Nevertheless, for the purpose of QoS provisioning, derivation of upper bounds on the delay of individual flows is required. In this paper, we consider the derivation of per-flow end-to-end delay bounds in DiffServ domains where peregress (or sink-tree) FIFO aggregation is in place. We expose a general methodology to derive delay bounds, and we instantiate it on a case study. We show that the methodology yields a tighter bound than those available from the literature, and we express a worst-case scenario for the case study network, in which the bound is actually achieved