Wavelength Assignment Problem Optimization using Neural Network Deepinder Kaur M.tech (Scholar), ECE BBSBEC, Fatehgarh Sahib deepinderb28@gmail.com Harpreet Kaur Mavi Asstt. Prof, ECE BBSBEC, Fatehgarh Sahib harpreet.mavi@gmail.com Abstract-To support WDM multicasting, MC-OXCs must be embedded into WDM network. Butkk they are very costly. This research investigates the problem of traffic due to multicast requests with the objective of minimizing the blocking probability, delay and load. Instead of using traditional methods, here, new heuristic method has been with session blocking probability has been utilized in proposed work. In addition to this, seqr-p and SeqR results have also been evaluated in comparison with SBP. Keywords- Wavelength division multiplexing (WDM), Traffic Engineering, Multicast Routing 1. Introduction In optical communication, wavelength division multiplexing (WDM) is a technology which carries a number of optical carrier signals on a single fibre by using different wavelengths of laser light [1].Main motivation of fiber optics is to meet demands of increase in the telecommunication data transmission. It totally works on the total internal reflection (critical angle, using Snell s law). An optical fiber is flexible, transparent fiber made of high quality extruded glass (silica) or plastic, slightly thicker than a human hair. It includes a core surrounded by transparent cladding material. From last year s, point to point WDM transmission technology in internet has been witnessed. The requirement of high bandwidth in the network has highlighted the need of faster switching. In addition to this, there is also another important need of enhancement in Internet Protocol to support traffic engineering. MPLS makes the internet architecture to behave in a connection-oriented fashion to support QoS and traffic engineering [2]. The proposed work studies request scheduling on wavelength division multiplexed (WDM) all-optical networks. The objective is to minimize the overall scheduling path by choosing the dynamic path routing which is discussed below in the flow chart. 5
To achieve high bandwidth utilization, guarantee low delay, and ensure short-term fairness, we try to construct a schedule with the path finding solution. The WDM (wave length division multiplexing) offers large network capacity, so high speed data transfer is possible. The optical paths guarantee network availability for job-execution assurances, so data transfer is reliable. Motivated to achieve a smaller latest job completion time, we then investigates the pre-emptive scheduling problem. Pre-emption allows jobs to be scheduled more flexibly, and thus may yield a smaller make span. λ λ λ n Individual fiber lines Optical multiplexer Figure 1: WDM Multiplexing Model 2. Literature Survey λ 1, λ 2, λ 3,.., M.Marsan et al [3] proposed a scheduling algorithm for multi-hop transmission improvement. At that point, the author has proposed single multi-hop calculation that performs better than already proposed algorithm. Andrew et al [4] presented two algorithms for scheduling variable-length packet transmissions in an optical inactive star system using λ n. Single fiber lines wavelength division multiplexing (WDM). Mahmoud Elhaddad et al [5] shown Proactive Reservationbased Switching (PRS) as an exchanging construction modelling for IP/WDM systems in view of Labelled Optical Burst Switching. A channel scheduling algorithms forces limitations on burst take off times to guarantee proficient use of wavelength channels and to keep up the separation between continuous blasts through the system. G.Mohan et al [6] proposed the execution of a multi-channel scheduling algorithm in view of the deficit roundrobin (DRR), which we call multichannel DRR (MCDRR). The authors have extended the first DRR to the instance of different channels with tuneable transmitters and altered collectors to give proficient reasonable queueing in half and half time division multiplexing (TDM)/wavelength division multiplexing (WDM) optical systems. Cheng Lai et al [7] proposed a scheduling algorithm for timeopened WDM show and-select optical systems. Z. Zhang et al [8] formalized the issue of boosting system throughput and minimizing aggregate delay in such a system. In the end, a basic and quick algorithm called the Scan and Swap Algorithm is given. Hang Jhang [9] proposed multi-jump scheduling algorithms for the All-to-All Broadcast (AAB) issue in Wavelength Division Multiplexed (WDM) optical star systems. The multi-jump AAB issue can be part into two sub problems: Logical Topologies Construction (LTC) issue, and Transmission Scheduling (TS) issue. For enhancing the productivity of multi-jump scheduling, the authors have concentrated on another multi-bounce transmission model and exchange the LTC issue to a unique instance of the Round Robin Tournament (RRT) issue. Chun Lin et al [10] proposed a WDM star coupler system comprises of various system hubs 6
associated by means of optical strands to an inactive star coupler. The research concentrates on the issue of scheduling multicast data packets in single bounce WDM systems. The issue is defined as an advancement issue and indicated to be NP-complete. Another heuristic calculation is proposed for this issue. 3. Workflow Model Formally, the enhancement of QoS parameters problem can be defined as follows: Input: A WDM network having set of various wavelengths ad set of multicast requests. Output: A set of light-structures satisfying the three constraints { multicast requests are rooted from source, MC-OXC has must one arc input, MC-OXC must be able to split the signal} Objective: Minimize delay, blocking probability and fewer wavelengths. Step: 1 Initialize No. of channels Step: 2 Initialize No. of links Step: 3 Total no. of loads Step: 4 Total no. of generations Step: 5 Wavelength Step: 6 Get MX-OXC To support WDM multicasting efficiently, the networks should be equipped with multicast capable cross-connects (MC-OXCs) that are capable of splitting an incoming signal to multiple outgoing ports. However, they are costly in both fabrication and power consumption. That is why, it is common in recent optical networks that all nodes are not capable of multicast. A small percentage of MC-OCXs mixed with multicast incapable cross-connects (MI-OXCs) may suffice to support optical multicasting. Traditionally, the light-tree was assumed to be a costeffective route with the support of MC-OXCs at the branching nodes. Recently, a new multicast structure called light-hierarchy has been proposed in literature survey. A light hierarchy allows an MI-OXC to be crossed several times by using different input output links, and hence relaxes the constraint of tree structure required by the routing. Step: 7 No. of requests For each request ri to be considered (selected according to the indexed order r1; r2 rn ), on the layers (wavelength graphs) λ1; λ2; λ n, one by one (also according to the indexed order). Whenever a light-hierarchy is computed in a certain layer, it is directly assigned the corresponding wavelength. Then the affected layers are updated in such a way that the ar updating takes place in such a way that wavelength assignment is removed from the wavelength graphs. Repeat the operations until all the requests are attempted. The pseudo-code of the SeqR algorithm is shown in Algorithm below. A request is considered to be accepted if all of its destinations are accepted. If not, the request is blocked, and the algorithm frees the wavelengths (restores the status for the used arcs) that have been assigned for the light-hierarchies computed for it. In contrast, if the request is not totally accepted, the accepted destinations are still served. This method is natural and relatively straightforward. It does not require global information of all the requests. Thus it can be applied for the dynamic traffic case. Step: 8 Compute blocking Step: 9 Plot blocking Step: 10 Show: legend('seq- Priority','SeqR'); xlabel('number of Wavelength(W)'); 7
ylabel('session Blocking probability [%]') Step: 11 Show legend('seq-priority','seqr'); xlabel('channel Request'); ylabel('session Blocking probability [%]'); Step: 12 Show legend('seq-priority','seqr'); xlabel('number of Request'); ylabel('session Blocking probability [%]'); Step: 13 Plot Channel Usage Step: 14 Plot destination block Step: 15 10 to 12. Plot blocking probability as in step Algorithm: Sequential Request Algorithm Input: set of wavelengths (ww, w2,..wn) and requests (r1, r2..rn). Output: Set of light hierarchies For all wn and rn. Do Compute light hierarchy. Assign different wavelengths. Start Initialise= load, delay, channels, link, wavelength Break End Equipment of network with MC-OXC Note no. of request as SeqR Compute Blocking Calculate Blocking Probability 4. Result Analysis The whole simulation is being done in MATLAB environment to give result analysis for blocking probability in WDM Model. Table 1: Variables Used Notations Meanings Optimize using neural network. C L No. of channels Loads w Wavelength Figure 2: Proposed Architecture Show session blocking probability with load, delay. R S SBP Multicast requests Total MC-OXC Session Blocking Probability 8
Figure 3: SPB versus Wavelength To show the difference between the light-hierarchy solution and the light tree solution, we compare the performance of seqr, neural network with that of seqr-p. As we can see from the figures, seqr outperforms seqr-p in session blocking probability used in all the studied contexts, but the result of neural network is better among other two methods. This is because it allow us to exploit all the available wavelengths in the current network state by taking advantage of using all possible directions (arcs) on every link and cross connect. Mainly strategies try to route as many requests as possible by making use of the available wavelengths on the layered graph, resulting in lower blocking probability. ON the other hand some strategies just compute the physical topology neglecting the current state of the network, giving high blocking probability. The fixed alternate strategy has more choices compared with the fixed strategy; it still suffers from high blocking probability due to the inherent shortcoming of static routing. Above figure shows that SBP for NN is better than SeqR-P and SeqR. Figure 5: SBP versus No. of Requests Indeed, since it is easier for small requests to be totally accepted, giving high priority to them probably results in lower SBP. However, when most of the small requests have been adopted, the availability of wavelengths becomes exhausted and the incoming (larger) requests will be blocked, causing high DBP. The above figure shows that NN is better SeqR and seqpr. 5. Conclusion and Future Scope Figure 4: SBP versus Channel Request This thesis investigated the problem of providing multiple static requests in MPLS networks in WDM. 9
The main aim is to minimize the blocking probability, delay and load considering Session blocking Probability. We employed neural network algorithm to solve this problem. The contributions of this work include the following: Providing IP Problem formulation over internet using exact solution for the Qos optimization problem in WDM network. Providing sequential request adaptive heuristic algorithm for dynamic traffic management. Showing that neural network is better than SeqR and SeqR-P in terms of delay, loads and blocking probability. Future scope lies in the investigation of more topics in WDM like traffic grooming, QoS-aware RWA, and power-aware RWA, are not investigated in this work. They will be the main concerns in our next study. References [1] ParthaGoswami, S.K. Ghosh2 and DebasishDatta, Dimensioning and Resource Provisioning for IP/MPLS-over-WDM Network, IEEE Department of Electronics & Electrical Communication Engineering, Indian Institute of Technology, Kharagpur, 2011. [2] Fenglin Li, Jianxun Chen, MPLS Traffic Engineering Load Balance Algorithm Using Deviation Path, International Conference on Computer Science and Service System, 2012. [3] M.AjmoneMarshan et al, Multichip scheduling in WDM /TDM, IEEE, pp.1105-1111, 1998. [4] Andrew Muir J.J. Garcia-Luna-Aceves, Distributed Queue Packet Scheduling Algorithms for WDM-Based Networks, UCSC, 1996. [5] Mahmoud Elhaddad, Rami Melhem, TaiebZnati, DebashisBasak, Traffic Shaping and Scheduling for OBS-based IP/WDM Backbones Citeseer, 2003. [6] G. Mohan, M. Ashish, and K. Akash, Burst Scheduling Based on Time-slotting and Fragmentation inwdm Optical Burst Switched Networks, Citeseer, 2002. [7] Chenglaicheah*, borhanuddinmohdali,, s. selvakennedy, a scheduling algorithm for wdmoptical networks, malaysian journal of computer science, vol. 14 no. 1, pp. 46-57, 2001. [8] Zhenghao Zhang and YuanyuanYang, Scheduling in BufferedWDMPacket Switching Networks with ArbitraryWavelength Conversion Capability, IEEE, 2004. [9] Hann-JangHo, Efficient multihop scheduling algorithms for packet transmissions in WDM optical star networks. Int. J.Electron.Commun.(AEU ), Vol. 6, pp.1186 1191, 2010. [10] Hwa-Chun Lin and Chun-Hsin Wang, Scheduling Multicast Packets in Single-Hop WDM Networks National Science Council, Taiwan,R.O.C., under grant NSC 87-2213-E- 007-004. 10