Protection Schemes on MG-OXC Networks

In this section, we describe our heuristics to handle the protection problem in MG-OXC networks. We first present an intuitive heuristic, Tunnel Based Path Protection (TPP) that deals with the two sub-problems independently. Then we introduce the improved algorithm, Tunnel Based Segment Protection (TSP) that takes the protection requirement in mind while allocating tunnels.

A. Tunnel Based Path Protection (TPP)

TPP allocates tunnels in the same way as CLWTA. After allocating tunnels, we can start to serve the incoming requests. For each request, both working path and protection path should be found or the request should be blocked. Fig. 13 illustrates TPP. Two tunnels, E-F-G and A-B-C (i.e., the thick lines), are allocated on the network. For the request from A to D, the working path A-B-C-D (i.e., the bottom dashed line) is found where sub-path A-B-C is in the tunnel layer and sub-path C-D in the wavelength-switching layer. The protection path A-E-F-G-D (i.e., the top dashed line) is found where sub-paths A-E and G-D are in wavelength-switching layer and sub-path E-F-G in the tunnel layer.

Protection path

E F G

A B C D

Working path

Fig. 13 Example of path protection with MG-OXC

Note that two tunnels for different node pairs on the logical topology may actually traverse the same link on the physical topology, which may cause both tunnels disconnected simultaneously if fiber link failure occurs on that common link.

In Fig. 14 (a), the two tunnels, A-E and B-F may be used for the working and protection path of a request. But in Fig. 14 (b), these two tunnels traverse the same link C-D and may fail simultaneously if a fiber cut occurs on link C-D. Thus, in hierarchical cross-connect network, we must make sure that working and protection paths for a request is physically link-disjoint.

(a)

(b)

A E

C D

B F

A E

C D

B F

Fig. 14 (a) Two tunnels, A-E and B-F in logical topology (b) Physical route of the two tunnels

B. Tunnel Based Segment Protection (TSP)

Although TPP provides a simple protection solution, it complicates the finding of link-disjoint lightpaths since two different tunnels on the logical topology may actually traverse the same physical link. Thus, we have pay additional attention to the overlapping of tunnels when finding link-disjoint path pair.

The main difference of TSP and TPP is that TSP takes protection into consideration when allocating tunnels. TSP divides the working path into segments according its switching types along the route and each segment is protected in its corresponding switching layer. Fig. 15 illustrates the idea of our approach. In Fig. 15 (a), the working path goes though the path A-B-C-D (i.e., the top dashed line) where sub-path A-B-C is in tunnel layer and C-D is in wavelength-switching layer. A-B-C-D is then divided into two segments, A-B-C and C-D. Segment A-B-C is protected by A-E-F-C in the tunnel layer and segment C-D by C-F-D (i.e., the bottom dashed line) in the wavelength-switching layer. Fig. 15 (b) depicts the layered concept of the

protection in each layer.

(a)

C B

A D

E F

(b)

Fig. 15 (a) Division of the working path according to the switching type (b) Finding a protection path for each working segment in each switching layer

Therefore, each time we try to allocate a tunnel between a node-pair, the working and protection tunnels for this node pair are allocated simultaneously. Note that if only the working tunnel can be found, we would abandon this working tunnel since it con not be protected.

C_T BT

Tunnel layer D_T

AT

E_T F_T

C_W B_W

Wavelength-switching

A_W D_W

layer

E_W F_W

Tunnel Based Segment Protection (TSP)

Step1. Form the auxiliary graph by adding all possible tunnels to the physical network.

Step2. Compute weight for each possible tunnel by routing the traffic matrix on the auxiliary graph

Step3. If the weight for each auxiliary link is smaller or equal to 0, go to Step 6.

Step4. Try to allocate fiber tunnels (working & protection) for the auxiliary link with maximum weight. If successful, decrease the weight of this auxiliary link by δF and go to Step 3. Otherwise, go to Step 5.

Step5. Try to allocate waveband tunnels (working & protection) for this auxiliary link. Decrease the weight of this auxiliary link by δB. Go to Step 3.

Step6. Wait for the lightpath request. When it comes, go to Step 7.

Step7. Find a working path for the request. If successful, find the protection path for each sub-path in wavelength switching layer. If the working path or any of the protection paths for the sub-path in wavelength switching layer can not be found, block the request.

Step8. Go to Step 6.

In Step 1, we construct the auxiliary graph. In Step 2, the importance of each auxiliary link is computed. Step 1 and Step 2 are the same as the method mentioned in CLWTA. Step 3 to Step 5 allocate tunnels on the network. We pick the auxiliary link with the maximum weight and try to find a link-disjoint path pair in the fiber-switching layer. The working tunnel is found prior to the corresponding protection tunnel. If the fiber tunnel pair (working and protection) are allocated successfully, the weight of the corresponding auxiliary link is decreased

by L F D

δ , where Wi,j is the weight of the auxiliary link connecting node i and

j, L the number of directional links in the original network topology, FT the number of fibers dedicated for tunnel allocation in each directional link and D the length constraint. Otherwise, we find a link-disjoint path pair in the waveband-switching

layer for this auxiliary link. Note that whether the waveband tunnel pair can be

allocated or not, weight of this auxiliary link is decreased by

D

B is the number of wavebands in a fiber. The process repeats until the weight of all auxiliary links are smaller or equal to 0. After all tunnels are allocated, we start to serve each coming lightpath request.

Dijkstra’s shortest path algorithm is applied to find routes for each request. The cost of each channel-link used to find a working path for the request is as follows.

⎪⎪ tunnels more easily than wavelength-switching channels.

After finding working path, it is divided according to the switching type. We do not have to consider the protection path in a tunnel since a protection tunnel has been allocated during tunnel allocation stage. Thus we only have to find a protection path for each segment in wavelength-switching layer. The cost of each channel-link in finding protection path can be classified into three categories. If a channel-link is occupied and cannot be shared, its cost is set to infinite. If a channel-link has been used by other protection path and can be shared, the cost is set to zero to increase the sharing efficiency. Otherwise, the channel-link cost is assigned the same way as the cost of a working channel-link is assigned. Following is the cost function for channel-links used to find protection path.

⎪⎩

A request is satisfied only if the working path and the corresponding protection paths are found. Either the working path or a protection path for a working segment can not be found, the request is blocked.

The performance of the protection schemes, TSP and TPP, in hierarchical cross-connect WDM networks are evaluated in the next section.

3.4 Numerical Results

In this section, we evaluate the performance of the Tunnel Based Segment Protection (TSP) on the 16-node topology in Fig. 16. The intuitive protection scheme, Tunnel Based Path Protection (TPP) is also implemented to compare with TSP. We assume that each directional link has five fibers. Each fiber contains forty wavelengths which are divided into four wavebands. That is, each waveband has 10 wavelengths and the first to the tenth wavelengths are in the first waveband, the eleventh to the twenty-first wavelengths are in the second waveband, …, and the thirty-first to the fortieth wavelengths are in the forth waveband. Each node is assumed to have enough wavelength conversion capability.

(F1)F(F2)B(F3)L stands for the experiment with F1 fibers for fiber-switching, F2 fibers for waveband-switching, and F3 fibers for wavelength-switching. The traffic is uniformly distributed in this simulation and each request is for a lightpath.

1

Fig. 16 The 16-node topology for this simulation

(a)

200 250 300 350 400

Number of requests

Blocking Prob.

TSP TPP TSP-PTLC

(b)

600 700 800 900 1,000

Number of requests

900 1,000 1,100 1,200 1,300 Number of requests

Blocking Prob.

TSP TPP TSP-PTLC

Fig. 17 Simulation results of TSP and path protection with different number of lightpath requests

Fig. 17 shows the simulation results in terms of blocking probability. TSP-PTLC stands for the TSP case where puts the length constraint on the protection tunnels. The results show that TPP is outperformed by TSP in all switching type combinations.

There are two reasons for the better performance of TSP. First, the resources of channels in fibers are used more efficiently in TSP. The channels dedicated for protection paths can be shared more easily if the working paths are divided into segments since two working paths passing the same physical link are not allowed to

share the same protection resources. Secondly, the protection tunnels in TSP can use the same wavelength-switching ports with working tunnels since the ingress and egress nodes of working and protection tunnels are the same. Once a link failure occurs and affect the traffic in a tunnel, we only have to reconfigure the fiber- or waveband-switching box to switch the affected traffic to the protection tunnel while using the original switching ports on the two ends of the affected tunnel.

Wavelength-switching ports are critical resources in MG-OXC networks, so we can get better performance and accommodate more lightpath requests in TSP. The defect in TPP is that wavelength-switching ports are required for each tunnel. Fig. 18 illustrates this concept of port saving.

Ports dedicate for a tunnel

Fig. 18 MG-OXC only reconfigures the fiber-switching box to switch the traffic in working tunnel to protection tunnel

Besides, the results also show that TSP-PTLC has higher blocking probability than TSP and TPP since it is difficult to find two link-disjoint paths that both follow the length constraint for a tunnel. But when there are more link resources dedicated for tunnel allocation (Fig. 17 (a)), TSP-PTLC has better performance than TPP since more tunnels can be allocated successfully.

To verify the significance of the wavelength-switching ports, we conduct the

WB-switching box

WB-switching box Fiber-switching box

Fiber-switching box

Working tunnel Protection tunnel

Wavelength-switching box

simulation under different switching combinations in TSP. The results are shown in Fig. 19. It shows that the more the wavelength-switching ports are, the less the traffic is blocked. The wavelength-switching ports influence the performance of the networks critically, thus it make sense to derive an algorithm that can save ports. TSP can save ports by letting a pair of working and protection tunnel use the same ports thus it will have better performance.

0 0.1 0.2 0.3 0.4 0.5

600 700 800 900 1,000

Number of requests

Blocking Prob.

2F2B1L 1F2B2L 1F1B3L

Fig. 19 Traffic load vs. blocking probability in different switching combinations in TSP

Chapter 4 Conclusion

This thesis investigates the problems related to MG-OXC networks. Although applying MG-OXC can save network costs, some problems are also raised. Works done in this thesis are to give solutions to these problems in MG-OXC networks, including the tunnel allocation problem and the protection problem.

For the tunnel allocation problem, we investigate the Capacity-Balanced Static Tunnel Allocation (CB-STA) in the hierarchical cross-connect network which employs three-stage multiplexing MG-OXCs, and find that CB-STA has some drawbacks. Since CB-STA does not consider the tunnel length constraint during the I-E pair selection stage, it resulted in few tunnels being allocated during the tunnel allocation stage. We propose a heuristic, Constant Length Weighted Tunnel Allocation (CLWTA) and Port-Constraint Constant Length Weighted Tunnel Allocation (PC-CLWTA) for allocating tunnels efficiently. In CLWTA and PC-CLWTA, allocation is only attempted for potential tunnels that complied with the tunnel length constraint. The simulation results show that that CLWTA outperforms CB-STA, since CB-STA allocates most tunnels in the makeup stage which tries to fill the fiber and waveband layers with tunnels to maximize the network resource utilization. The results also show that in situations involving low wavelength-switching ports, considering the wavelength-switching ports when allocating tunnels improves the performance. PC-STA thus takes effect in this situation.

For the protection problem, we investigate the protection scheme for the single-link failure in the MG-OXC networks. Since the protection problem has not studied intensively in MG-OXC networks, the mass MG-OXC deployment is at a risk

of huge data losses once a link failure occurs. This work thus aims to provide an efficient protection scheme for MG-OXC networks.

An intuitive solution is to allocate tunnels off-line by CLWTA and then find two link-disjoint lightpaths from source to destination for each incoming request.

Although the intuitive heuristic provides a protection solution for the MG-OXC networks, it does not consider the protection while allocates tunnels. The lack of protection consideration while allocating tunnels complexes the finding of link-disjoint lightpaths since two different tunnels that have a common link can not be utilized by working and protection paths at the same time.

The heuristic, Tunnel Based Segment Protection (TSP) that considers tunnel allocation with protection requirement in mind, is then proposed. In the tunnel allocation process of TSP, a protection tunnel is always allocated simultaneously with a working tunnel. Once a lightpath traverses the fiber or waveband tunnels, the segments in those tunnels are protected by the channels in the corresponding protection tunnels. The remaining segments of the lightpath in wavelength-switching layer are protected by the channels in the same layer. The channels dedicated for protection thus can be shared more easily than in the TPP. Besides, the performance of the network is improved since the working and protection tunnels use the same wavelength-switching ports in the network with MG-OXC, in which wavelength-switching ports are rare resources.

This thesis solves the static tunnel allocation and protection problems, but there are still some improvements left for the future works. In this thesis, tunnels are allocated off-line and will not be removed. However, when the traffic has changed, dynamic tunnel reconfiguration should also be considered. Additionally, various node architectures with slightly differences that support multi-granularity traffic have been proposed. We will extend our tunnel allocation and protection algorithms to be

adopted in networks with these node architectures.

Reference

[1] P. –H. Ho and H. Mouftah, “Routing and Wavelength Assignment with Multigranularity Traffic in Optical Networks,” IEEE Journal of Lightwave Technology, Vol. 20, No. 8, Aug. 2002, pp. 1992-1303.

[2] K. Harada, K. Shimizu, T. Kudou, and T. Ozeki, “Hierarchical optical path cross-connect systems for large scale WDM networks,” in Proc. Optical Fiber Communications (OFC’99), Vol. 2, Feb. 1999, pp. 356-358.

[3] L. Noirie, M. Vigoureus, and E. Dotaro, “Impact of intermediate traffic grouping on the dimensioning of multi-granularity optical networks,” in Proc. Optical Fiber Communications (OFC’01), Vol. 2, March 2001, pp. TuG3.1-TuG3.3.

[4] M. Lee, J. Yu, Y. Kim, C.-H. Kang, and J. Park, “Design of hierarchical crossconnect WDM networks employing a two-stage multiplexing scheme of waveband and wavelength,” IEEE Journal on Selected Areas in Communications, Vol. 20, Jan. 2002, pp. 166-171.

[5] X. Cao, V. Anand, Y. Xiong and C. Qiao, “Performance Evaluation of Wavelength Band Switching in Multi-fiber All-Optical Networks,” in Proc. IEEE Infocom, Vol. 3, Mar. 2003, pp. 2251-2261.

[6] O. Gerstel, R. Ramaswami, and W.-K. Wang, “Making use of a two stage multiplexing scheme in a WDM network,” in Proc. Optical Fiber Communications (OFC’00), Vol. 3, Mar. 2000, pp. 44-46.

[7] R. K. Ahuja, T. L. Magnanti, J. B. Orlin, “Network Flows: Theory, Algorithms, and Applications,” Prentice Hall, 1993.

[8] T. E. Stern, K. Bala, “Multiwavelength Optical Networks: A Layered Approach,”

Addison-Wesley, 1999.

[9] Z. Zhang and A. S. Acampora, “A Heuristic Wavelength Assignment Algorithm for Multihop WDM Networks with Wavelength Routing and Wavelength Re-Use,” IEEE/ACM Transactions on networking. Vol. 3. Issue. 3. June 1995, pp.

281-288.

[10] C. Chen and S. Banerjee, “A New Model for Optimal Routing in All-Optical Networks with Scalable Number of Wavelength Converters," in Proc. IEEE Globecom ‘95, Vol. 2, Nov. 1995, pp. 993-997.

[11] C. Xin, Y. Ye, S. Dixit, and C. Qiao, “A Joint Lightpath Routing Approach in Survivable Optical Networks,” Optical Network Magazines, May/June, 2002, pp.

23-32.

[12] P. –H. Ho and H. T. Mouftah, “Issues on Diverse Routing for WDM Mesh Networks with Survivability,” In Proc. IEEE International Conference on Computer and Communication Networks, Scottsdale, AZ, Oct. 2001, pp. 60-65.

[13] E. Bouillet, J.Labourdette, G. Ellina, R. Ramamurthy, and S. Chaudhuri,

“Stochastic Approaches to Compute Shared Mesh Restored Lightpaths in Optical Network Architectures,” In Proc. IEEE Infocom‘02.

[14] P. –H. Ho and H. T. Mouftah, “A framework for service-guaranteed shared protection in WDM mesh networks,” IEEE Communications Magazine, Vol.

40, No. 2, Feb. 2002, pp 97-103.

[15] D. Xu, Y. Xiong and C. Qiao, “A new PROMISE algorithm in networks with shared risk link groups,” In Proc. IEEE Globecom’03, pp. 2536-2540