[lecture notes in computer science] distributed computing and networking volume 7730 || poster: a...
TRANSCRIPT
POSTER: A New Approachto Impairment-Aware Static RWA in Optical
WDM Networks
Sebastian Zawada, Shrestharth Ghosh, Fangyun Luo, Sriharsha Varanasi,Arunita Jaekel, and Subir Bandyopadhyay
School of Computer Science, University of Windsor, Canada N9B 3P4
In order to set up a number of lightpaths to satisfy user requirements for datacommunication, the static (also called offline) Route and Wavelength Assign-ment (RWA) problem must be solved. The quality of transmission (QoT) of anoptical signal propagating through an optical network degrades, due to physi-cal layer considerations such as optical noise, chromatic and polarization modedispersion, four wave mixing, cross-phase modulation and cross-talk [4]. Thisleads to an increase in the Bit Error Rate (BER) of the optical signal and thecorresponding lightpath becomes infeasible for communication if the BER valuecrosses a certain threshold limit. We have used an analytical model proposed byPereira et al [3] to estimate the BER. The interdependence between the physicaland the network layers makes the RWA problem in the presence of impairmentsa cross-layer optimization problem [1]. To address this problem, a number ofapproaches are emerging, usually referred to as impairment-aware-RWA (or IA-RWA) algorithms that take into account the interaction between the networkand the physical layers. Our objective is to design a transparent network, wherethe IA-RWA algorithm must provision lightpaths so that the BER value of eachlightpath never exceeds a given threshold.
The objective of this heuristic is to carry out static RWA for a set R ofrequests for data communication, taking into consideration both class 1 andclass 2 physical layer impairments [1]. Here each request is denoted by a pairof nodes (s, d), meaning that a source node s wishes to communicate with adestination node d. If RWA is successful for a pair of nodes (s, d) ∈ R, it meansthat a transparent lightpath may be deployed from node s to d, using some pathΨsd from node s to node d, and a channel c that is not used by any lightpath
using any edge on the path Ψsd . The heuristic takes an iterative approach to RWA
where, in a given iteration, lightpaths are assigned to as many source, destinationpairs in setR as possible. In a given iteration, the heuristic considers each requestin R that has not yet been assigned a lightpath successfully and i) determines,if possible, an appropriate path Ψs
d on the physical layer, for the request beingconsidered, ii) determines, if possible, the best channel c for the request, iii)determines whether the impairments on all the lightpaths, including this newone, meet the QoT requirements. These impairments we consider are due to a)class 1 impairments on this new lightpath, b) class 2 impairments on existinglightpaths, if this new proposed lightpath is set up, iii) class 2 impairments onthis new proposed lightpath due to existing lightpaths. If all these steps aresuccessful, the request is deemed to be “handled” in this iteration. We includethe pair (s, d), the path Ψs
d and the channel number c in the list of established
D. Frey et al. (Eds.): ICDCN 2013, LNCS 7730, pp. 456–457, 2013.© Springer-Verlag Berlin Heidelberg 2013
A New Approach to Impairment-Aware Static RWA 457
lightpaths and remove the pair (s, d) from the set of requests R. Otherwise,the request is retained in R for subsequent iterations. The process terminateseither when all requests have been handled or when no other paths remain forthe requests currently under consideration. The heuristic consists of 3 phases asfollows. Using some relatively large value for k, phase 1 generates a set P s
d of kor fewer shortest paths from s to d, for each request (s, d) in set R. For eachremaining request (s, d) in set R, phase 2 selects, if possible, a path Ψs
d ∈ P sd
from s to d. For each route Ψsd found in phase 2, phase 3 assigns a valid channel
to set up a transparent lightpath from s to d, whenever possible.In order to evaluate the proposed heuristic, we have compared it to the follow-
ing approaches i) Classical RWA, ii) Shortest Path First (SPF) [2], iii) LongestPath First (LPF) [2]. Classical RWA assumes an ideal physical layer, with no im-pairments and the corresponding results provide an upper bound on the numberof successful connections. Fig. 1 shows the percentage of demands that could besuccessfully handled for different network sizes, when presented with a demandset of 50 demands. We see that our proposed heuristic consistently outperformsboth SPF and LPF. The amount of improvement varies with the network sizeand ranges from about 10.33% (13.42%) for a 10-node network to 4.41% (5.56%)for a 30-node network compared to SPF (LPF) and is also quite close to cRWA(which provides the upper bound).
Fig. 1. Comparison of successfully routed demands for different network sizes
References
1. Christodoulopoulos, K., Manousakis, K., Varvarigos, E.: Offline routing and wave-length assignment in transparent wdm networks. IEEE/ACM Transactions on Net-working 18(5), 1557–1570 (2010)
2. Ezzahdi, M.A., Al Zahr, S., Koubaa, M., Puech, N., Gagnaire, M.: Lerp: a quality oftransmission dependent heuristic for routing and wavelength assignment in hybridwdm networks. In: Proceedings of the15th International Conference on ComputerCommunications and Networks, ICCCN 2006, pp. 125–136 (October 2006)
3. Pereira, H.A., Chaves, D.A.R., Bastos-Filho, C.J.A., Martins-Filho, J.F.: Osnrmodel to consider physical layer impairments in transparent optical networks. Jour-nal of Photonic Network Communications 18, 137–149 (2009)
4. Shen, G., Tucker, R.S.: Translucent optical networks: the way forward. IEEE Com-munications Magazine 45, 48–54 (2007)