ECOC Technical Digest 2011 OSA

A Novel Technique for Survivable Multicast Routing in Optical WDM Mesh Networks
Costas K. Constantinou and Georgios Ellinas
Department of Electrical and Computer Engineering, University of Cyprus, Nicosia, 1678, Cyprus {constantinou.k.costas,gellinas}@ucy.ac.cy

Abstract: This paper presents a new multicast protection technique for optical WDM mesh networks based on a novel Steiner tree algorithm. The new technique has lower blocking probability and cost, compared to conventional protection methods.

1.

Introduction

Optical ber has prevailed as the medium of choice for high-speed communications. The reasons are the huge amount of bandwidth available on each ber, as well as its low bit error rates compared to copper-wire transmission systems [1]. Bandwidth-intensive multicasting applications such as high-denition television, video conferencing, interactive distance learning, live auctions, distributed games, and video-on-demand are applications that are becoming widely popular and are now becoming more feasible because of the high-capacity and increased capabilities provided by intelligent Wavelength Division Multiplexing (WDM) optical networks [2]. These multicasting applications are based on the calculation of light-trees, utilizing optical splitters in the network nodes [3]. On the other hand, the vast amount of information that a ber carries, as well as the amount of information loss in case of a failure on a light-tree that can affect the trafc to multiple destinations have led to the necessity for the development of efcient multicast protection techniques. The proposed protection technique attempts to provision protected multicast services, and at the same time reduce the blocking probability of the multicast demands that arrive in the network, compared with other existing protection techniques, and also reduce the corresponding cost for the working and protection trees (calculated in terms of link weights). The proposed protection technique is designed to protect the network against any single link failure. The method that is described in this paper achieves the above by utilizing a novel Steiner tree routing algorithm called the Steiner Node Heuristic (SNH). 2. Multicast Protection Techniques

A large number of the multicast protection techniques in the literature are directly related to the Steiner tree problem in graph theory, namely the problem of nding the minimum cost tree connecting a node (source) with a set of other nodes (destinations) in the network (the set of destinations being a subset of the set of network nodes). This is an NP-complete problem, hence a number of heuristic algorithms has been developed for solving it. The two widely used in the literature are the Pruned Prim Heuristic (PPH) [4] and the Minimum Path Heuristic (MPH) heuristic algorithms [5]. If PPH is used, a minimum spanning tree (MST) is created with Prims algorithm and the unnecessary arcs and nodes are deleted to create the desired multicast tree. In MPH, the tree initially consists only of the source. Then, in every step, the closest destination to the tree is added to it (along with the corresponding path), thus creating a new tree, until all the destinations are connected on the tree. The purpose of a large number of the multicast protection schemes is the calculation of two disjoint paths from the source to each destination of the specic multicast demand. If this is realizable, the multicast demand will be serviced regardless of any single link failure in the network. One way to achieve this is by calculating two link-disjoint (LDT) or arc-disjoint (ADT) trees for every multicast demand utilizing the aforementioned PPH and MPH routing techniques (resulting in PPH-ADT, PPH-LDT, MPH-ADT, and MPH-LDT multicast protection techniques) [6]. In this paper we present a more efcient multicast protection method based on a novel Steiner tree algorithm called the Steiner Node Heuristic (SNH). Our novel routing algorithm, that uses MPH as its basis, has enhanced cost results compared to the MPH and PPH techniques. It consists of the following steps: 1. The Steiner tree (T REE ) is calculated using MPH and its cost is denoted as C.

978-1-55752-932-9/11/$26.00 Optical Society of America

We.10.P1.88.pdf 1 7/27/2011 4:33:05 PM

ECOC Technical Digest 2011 OSA

2. For every node i that is not on the tree, the node is added in the destinations set, a new tree (treei ) of cost ci is calculated and node i is removed from the destinations set. The tree with the minimum cost among them, named Tree, is found. 3. If (cost of Tree) (cost of T REE ), the algorithm stops. 4. If (cost of Tree) < (cost of T REE ), the corresponding node is added permanently in the destinations set, T REE is replaced with Tree and, if not all nodes are added to the destinations set, the algorithm returns to step 2, otherwise it stops. 5. T REE is the output of the algorithm. The SNH heuristic algorithm has a polynomial time complexity, namely O(m2 n3 ) for a network with m edges and n vertices. The main idea behind SNH is that it is able to nd the nodes that, if are added in the set of destinations, will give a tree with less cost. The following simple example shown in Figure 1 makes the algorithm more understandable and shows its enhanced cost performance:

100 s 100

d1 60 d2

60 n 60

Fig. 1. Example of the SNH Algorithm While the MPH algorithm nds tree {S d1 , S d2 } with cost 200, the SNH technique, with the addition of node n to the destinations set, is able to nd a tree with less cost, i.e., tree {S n, n d1 , n d2 }, with cost 180. 3. Performance Evaluation

The performance of the SNH-ADT multicast protection heuristic algorithm was evaluated via simulation, using the sample network shown in Figure 2, consisting of 24 nodes and 43 bidirectional links. The cost of each link is denoted in the gure. This cost may be the actual length of the link or any other cost assigned for it (e.g., latency, or monetary/port cost, etc).

Fig. 2. Test Network Used for Performance Evaluation Let D = k 1 be the number of destinations of the multicast group that must be established. The experiment is executed for all possible multicast groups (from D = 1 (unicast call) to D = 23 (broadcast call)) using the PPH-ADT, MPH-ADT and SNH-ADT techniques to calculate the pair (working and protection) of the arc-disjoint trees. (The case of link-disjoint trees is not considered here, as two arc-disjoint trees are capable of providing two link-disjoint paths for each destination as well, with less use of arcs compared to the link-disjoint case). For each multicast group case, the simulation is repeated 5, 000 times while the source and destinations of each connection are randomly generated

We.10.P1.88.pdf 2

7/27/2011 4:33:05 PM

ECOC Technical Digest 2011 OSA

and are distributed uniformly across the network. A multicast request is established, if a pair of working and protection arc-disjoint trees can be found for that request, otherwise it is blocked. The blocking probability and the average cost (in terms of link weights utilized for the working and protection trees) for the three arc-disjoint multicast protection techniques are presented in Figures 3(a) and 3(b) respectively for different multicast group sizes (a multicast group as shown in the plots does not include the source node). For the calculation of the average cost, only the non-blocked commands are considered. As it can be seen from the plots, the new proposed SNH-ADT multicast protection method gives lower blocking probability compared to the MPH-ADT technique and lower blocking compared to the PPH-ADT approach for larger multicast group sizes. Also, it is always much better than the PPH-ADT technique and always equal or better than the MPH-ADT approach in terms of the average cost. Thus, the SNH-ADT technique will be preferred over both existing techniques when both the blocking probability and average cost are taken into consideration. This is a direct consequence of the proposed SNH Steiner tree routing heuristic algorithm that has increased cost performance compared to other Steiner tree routing techniques.
0.14

MPH.ADT
0.12

45000 40000 35000 30000

PPH.ADT SNH.ADT

0.1

Blocking pro obability

Average Cost

0.08

25000

0.06

MPH.ADT
20000

0.04 15000 0.02 10000 5000 1 3 5 7 9 11 13 15 17 19 21 23 1 3 5 7 9 11 13 15

PPH.ADT SNH.ADT

17

19

21

23

Multicast Group Size

Multicast Group size

Fig. 3. (a) Blocking Probability and (b) Average Cost, for the MPH-ADT, PPH-ADT, and SNH-ADT Multicast Protection Techniques 4. Conclusions

A new multicast protection technique based on the SNH multicast routing algorithm is presented in this work. It has been proved through simulations that the new method has enhanced results in terms of blocking probability for larger multicast group sizes and less average cost, compared with the existing MPH- and PPH-based protection methods. References 1. Rajiv Ramaswami, Multiwavelength Lightwave Networks for Computer Communication, IEEE Communications Magazine, 31(2):78-88, 1993. 2. T. E. Stern, G. Ellinas, and K. Bala, Multiwavelength Optical Networks: Architectures, Design, and Control, 2nd ed. Cambridge University Press, 2008. 3. L. Sahasrabuddhe and B. Mukherjee, Light-Trees: Optical Multicasting for Improved Performance in Wavelength-Routed Networks, IEEE Communications Magazine, 37(2):67-73, 1999. 4. T. H. Corman, C. E. Leiserson, and R. L. Rivest, Introduction to Algorithms, 2nd ed. Cambridge, MA: MIT Press, 2001. 5. H. Takahashi and A. Matsuyama, An Approximate Solution for the Steiner Problem in Graphs, Math. Japonica, 24(6):573-577, 1980. 6. Narendra K. Singhal, Laxman H. Sahasrabuddhe, Biswanath Mukherjee, Provisioning of Survivable Multicast Sessions Against Single Link Failures in Optical WDM Mesh Networks, IEEE/OSA Journal of Lightwave Technology, 21(11):2587-2594, 2003.

We.10.P1.88.pdf 3

7/27/2011 4:33:05 PM