Multicast Routing Algorithms for Sparse Splitting Optical Networks 1

Transcription

1 Muticast Routing Agorithms for Sparse Spitting Optica Networks 1 Presenter: Aniang Cai PhD candidate, Department of Eectronic Engineering, City University of Hong Kong, Hong Kong SAR Emai: caianiang@outook.com Group Meeting, Friday, October 21, C. K. Constantinou, K. Manousakis, G. Einas, Muticast routing agorithms for sparse spitting optica networks, Comput. Commun., vo. 77, pp , Mar

2 Outine 1. Background 2. Probem Statement 3. Existing and proposed heuristics 4. Numerica resuts 5. Concusions 2

3 Background One-to-many muticast appications - data transmission from one source to mutipe destinations Teepresence, onine teaching, utra-highdefinition TV deivery, video conferencing, etc. Optica muticasting is more spectray efficient than IP muticasting^ Current networks may not have fu muticast capabiity sparse spitting ^ L. H. Sahasrabuddhe and B. Mukherjee, Light-trees: optica muticasting for improved performance in waveength routed networks, IEEE Commun. Mag., vo. 37, no. 2, pp ,

4 Node Muticast Capabiities Muticast-capabe (MC) node an input signa going through it can be dropped ocay and/or switched to one, many, or a of its output ports Muticast-incapabe (MI) node Drop-or-continue (DoC) an input signa can be either dropped ocay or switched to an output port Drop-and-continue (DaC) enhanced over DoC with extra capabiity that an input signa can be spit into two copies, one dropped ocay and the other switched to an output port 4

5 MC Node Architecture Fu waveength conversion λ " λ # λ $ [5]. X. Zhang, J. Y. Wei and C. Qiao, Constrained muticast routing in WDM networks with sparse ight spitting, J. Lightw. Techno., vo. 18, no. 12, pp ,

6 DoC Node Architecture λ " λ # 6

7 DaC Node Architecture λ " λ # λ " λ # 7

8 Exampe of Muticast Session Routing Subgraph for a muticast session λ & λ & λ % λ % λ % 8

9 Probem Statement Input Network graph: G=(V +, A + ) Fu waveength conversion in each node c "# : the cost of arc [i, j] in A + Number of waveengths on each network fiber (arc): W Set of MC nodes: MC 567 Muticast session consisting of a source and k destinations: S, S = s, D = {s, d &, d %,, d $ } Output routing subgraph RSG = (V CD+, A CD+ ) with the minimum cost 9

10 Existing Heuristics DaC networks DaC and MC nodes Member-ony (MO) No cyces permitted -> high cost DoC networks DoC and MC nodes On-tree MC node first (OTMCF) [21] Connect MI destinations to the cosest MC nodes Connect source to the MC nodes and MC destinations Nearest MC node first (NMCF) [21] reversed procedure to OTMCF Muticasting using spitters (MUS) [22] Improvement over NMCF [21]. C. Y. Hsieh and W. Liao, A-optica muticast routing in sparse spitting WDM networks, IEEE J. Se. Areas Commun., vo. 25, no. 6, pp , [22]. S. Cho, T. J. Lee, M. Chung, and H.Choo, Minimum cost muticast ro uting based on high utiization MC nodes suited to sparse-spitting optica networks, in Proc. ICCSA,

11 Proposed Heuristics MPH* - based on minimum path heuristic (MPH) Sparse-spitting muticast routing heuristic (SSMRH) 11

12 MPH* 12

13 DoC Network Exampe of MPH* X = {s}, Y = {d 1, d 2 }, c = 0 P KL 5 = d & b a s, c = 12 PQL R P KS 5 = d & s, c = 15 PQS R Seect P 5KL, c = c + c PRQL = 12 X = s, b, Y = d %, c = 12 P KS 5 = d & s, c = 15 PQS R P KS V = d % a b, c = 11 PQS W Seect P VKS, c = c + c PWQS = 23 X = s, b, Y =, c = 23 13

14 Other Exampes MPH as we as the existing agorithms have improved performance if specific MC nodes are added in the destination set 14

15 SSMRH 15

16 Test Conditions Test networks: USNET, NSFNET, and 18 randomy created networks Link cost varies from 1 to 1000 Different k, the number of destinations and different z, the number of MC nodes which are paced at the nodes that have the argest degree^ 500 muticast sessions MO and MUS were modified to support both networks whie OTMCF and NMCF were appied without any changes ^ S. W. Wang, Aocation of ight spitters in a-optica WDM networks with sparse ight spitting capabiities, Teecommun. Syst., vo. 52, no. 1, pp ,

17 Performance Metrics I \ is the extra average cost of heuristic H compared to the optimum SO \ is the percentage of the cases where heuristic H fais to find the optimasoution 17

18 Numerica Resuts SSMRH performs the best for a cases and cose to optimum SSMRH performs the best in terms of the percentage of the derived optima soutions 18

19 Numerica Resuts (Cont.) 19

20 Performance Metrics (Cont.) gives the average vaue of the % reative increase of the average cost compared to SSMRH for heuristic H and network j J \ is the average vaue of randomy created networks over a 20

21 Numerica Resuts (Cont.) SSMRH performs best For the base heuristics of DoC case, OTMCF and MUS have the best performance For the base heuristics of DaC case, MPH* gives resuts coser to the ones obtained by SSMRH 21

22 Concusions Investigated the probemof muticast routing for DaC and DoC networks with sparsespitting capabiities Proposed ILP formuation and heuristics The proposed agorithms achieve an important decrease of the average cost of the derived soutions, compared to existing agorithms The proposed agorithms obtain the optima soution for the majority of the investigated cases 22

23 Thank you. Questions or comments? Presenter: Aniang Cai PhD candidate, Department of Eectronic Engineering, City University of Hong Kong, Hong Kong SAR Emai: Group Meeting, Friday, October 21,