Preplanned restoration of multicast demands in optical networks

Abstract

The evolution of multicast applications has speeded up. providers have to offer cheap and reliable services for their customers. One way to achieve this is to use a restoration method that provides relatively fast restoration calculated upon a failure, where nearly the minimal amount of resources is allocated. In this paper we propose a new, faster way of restoring multicast demands after link failure that is based on preplaning. We prove its optimality if a few preconditions hold. To do so a new formalization of the wavelength graph transformation (WGT) is presented. We show, based on simulations, that the beneficial properties of our method hold even when one or more preconditions are not kept.

Introduction

Demand for multicast HD-TV and even higher resolutions grows rapidly, requiring increasingly more bandwidth. After a link failure the total of available network resources drops. Remaining ‘healthy’ links have to transmit also the traffic that was routed over the link that failed. In such cases, it is a key question how to save as many resources as possible. However, users would not accept high outage time – called mean down time (MDT) – of any service Ref. [1]. In case of multicast traffic the problem is even more important as not only one user, but a set of users may be affected by even a single link failure. Different kinds of protection methods (e.g.: 1 + 1 protection) offer very low MDT. However, the resources used by the back-up paths are reserved even if no link is down Ref. [2]. Heuristic restoration techniques – i.e. calculation of the new path ‘on the fly’ – may provide solution within affordable but not zero time. Nevertheless, these methods tend to produce solutions that may not be optimal in the sense of resource utilization, especially for multicast. As the MDT is a very important, exact method, such as ILP (integer linear programming), cannot be applied, due to their calculation complexity.

Preplanned restoration of unicast demands has been studied deeply. In Ref. [3] a stochastic approach is presented, in Ref. [4] a distributed, probabilistic solution is proposed. Methods for single – and multilayer restoration, and their comparison were given in Ref. [5].

Preplanned restoration of multicast demands has been studied for different networks and architectures – including application level multicasting (ALM) Ref. [6] – in different phases of the life of a demand – i.e. at creation, when it is modified, etc. – Refs. [7], [8], etc. These approaches are well known for their good properties – i.e. cost effectiveness, acceptable blocking ratio, etc. However, according to our knowledge, the preplanned restoration of a static multicast demand in multi-layer optical network has not been studied. In the following, we will investigate this question.

In core networks most of the multicast demands are static – e.g. virtual private networks, TV multicasts, etc. Their paths do not alter in time. Thus, it is possible to calculate restoration paths for all the plausible failure scenarios. However, it is not affordable to calculate the global optimal restoration paths for all demands at once. This restriction is the result of three circumstances:

• There is dynamic background traffic in the network. This traffic cannot be incorporated into the calculation of restoration plan.

• It is possible that the global optimum can only be achieved if existing and intact demands are rerouted. By interrupting and rerouting these demands the Quality of Service (QoS) drops. This is not permissible.

• Calculation of optimal paths for multiple unicast demands or a single multicast demand is NP-hard, Ref. [9]. This calculation is affordable only in special cases.

Our aim in this paper is to show how the problem of preplanned restoration of multicast demands can be simplified and solved in a way that it may give a good and valid solution in most cases, meanwhile its off-line and on-line calculation time is moderated.

The paper is organized as follows: Section 2 gives a basic overview of the topic. Formalization of the used environment is given in Section 3. Known solutions are summarized in Section 4. The details of the proposed method are given in Section 5, while our simulation results in Section 6. We conclude our work in Section 7.