Prioritized traffic restoration in connection oriented QoS based networks

Abstract

Connection oriented quality-of-service (QoS) based networks offer several service classes each designed to handle connections supporting various types of applications. A network failure will typically result in a large number of disrupted connections in each service class, all of which need to be restored simultaneously. In this paper we propose a distributed fault recovery priority scheme for dynamic restoration of connections. The proposed scheme exploits available network resources immediately after a failure together with properties of typical applications carried by the connections in order to give the applications sufficient network performance. The scheme is formulated within the context of switched virtual circuit routing in ATM networks, but it is applicable to ATM virtual path restoration as well and to other class based QoS networks like DiffServ MPLS based networks. The priority based restoration scheme is formulated as a centralized path flow integer programming optimization problem serving as a comparison benchmark for suboptimal distributed schemes. We propose a distributed heuristic scheme and give numerical results illustrating the quality of the proposed distributed scheme in comparison to the centralized optimal approach. Our numerical results show that the proposed distributed scheme can significantly improve utilization of network resources and enhance network performance after a failure. q

Lagrangean heuristic for primary routes assignment in survivable connection-oriented networks

“…We use in our simulations 4 networks having 10 nodes: Net34, Net38, Net42 and Net46 having 34, 38, 42 and 46 arcs, respectively. The topology of Net42 used also in [18] is shown in Fig. 1.…”

“…Since there are no coupling constraints between variables f a and variables x k p , problem (17)(18)(19)(20) can be divided into two independent subproblems: Subproblem 1…”

“…Given multiplier's vector λ we can solve the Lagrangean relaxation problem given by (17)(18)(19)(20) by solving two subproblems as presented above. Let x k p (λ) and f a (λ) be an optimal solution of the Lagrangean relaxation for a fixed vector λ of multipliers.…”

“…Set i := i + 1, ρ iter := ρ iter + 1. Given λ i , solve the Lagrangean relaxation problem (17)(18)(19)(20) as decoupled subproblems to obtain φ(λ i ), x i , f i . Step 2a.…”

“…the flow of connections that are not restored after a failure due to limited resources of capacity. A similar approach was applied in [18,23,24,[30][31][32].…”