1. The Research of Applying Random Early Blocking strategy to Dynamic Lightpath Routing National Yunlin University of Science & Technology

2. Outline Background Background Related works Related works –SCP, WSCP, EWSCP –Layered-Graph model Simulation of routing strategies Simulation of routing strategies Random Early Blocking (REB) Random Early Blocking (REB) –Simulation results Conclusion Conclusion

3. Background Communication service Communication service –More bandwidth is needed by customs –The poplar services »Internet, cell-phone……

4. Background WDM (Wavelength Division Multiplex) WDM (Wavelength Division Multiplex) –Wavelengths are carried on a fiber –Bandwidth grows much

5. Background WDM network routing strategies WDM network routing strategies –Fixed routing –Alternative routing –Adaptive routing Routing & Wavelength Assignment (RWA) Routing & Wavelength Assignment (RWA) –Wavelength continuous constrain –Wavelength conversion –Converter placement

6. The adaptive routing strategies To search lightpath as calls coming in To search lightpath as calls coming in Routing information Routing information –Distance vector (RIP) –Link state (OSPF)

7. Adaptive routing – Related works Shortest Cost Path (SCP) Shortest Cost Path (SCP) –The path cost function –Equal-weighted link cost –Disadvantage »Some links may be heavy traffic load traffic load »Higher blocking probability

8. Adaptive routing – Related works Weighted Shortest Cost Path (WSCP) Weighted Shortest Cost Path (WSCP) –Distributing traffic load –Better network utilization –Lower blocking probability –The path cost function –Liner link cost function assignment

9. Adaptive routing – Related works Exponential Weighted Shortest Cost Path (EWSCP) Exponential Weighted Shortest Cost Path (EWSCP) –Light load  shortest path –Medium load  load balance –Heavy load  avoid exhaust all wavelength on a link all wavelength on a link –Link cost function

10. Layered-graph model Physical network topology Physical network topology –N(R,L,W) –R is the set of routing nodes –L is the set of links –W is the set of wavelengths per link

11. Layered-graph model – logical network

12. N ’ (V,E,X) N ’ (V,E,X) –V is the set of routing nodes –E is the set of links –X is the set of wavelengths per link –R’ is the set of routing nodes with conversion Number of V is : Number of V is : Number of E Number of E

13. Simulation parameters W=32 per network link W=32 per network link Full converter, sparse converter, non- converter Full converter, sparse converter, non- converter Different network topologies Different network topologies Network traffic : Network traffic : Link congestion index : Link congestion index :

14. Network topology 8-nodes Ring 14-nodes NSFNET

15. Network topology 14-nodes Random Net

16. Ring Sparse converter Full converter

17. NSFNET full-converter

18. NSFNET Sparse converter Non-converter

19. Random Full converter

20. EWSCP results Reducing blocking probability Reducing blocking probability Load balance Load balance Better performance than WSCP and SCP Better performance than WSCP and SCP Rand Sparse converter

21. Random Early Blocking (REB) Early (set active threshold) Early (set active threshold) –Before resource assign out –Resource conservation as heavy load Random Blocking Random Blocking –Block long-lightpath randomly –Random function Result Result –Reduce blocking probability much as heavy load

22. NSFNET Left-top: Non-converter Left-bottom: Sparse converter Right: Full-converter

23. NSFNET (full converter)

24. Random Sparse converter Full converter

25. Random (full converter)

26. Conclusion We proposed EWSCP algorithm to balance traffic load We proposed EWSCP algorithm to balance traffic load –EWSCP enhance performance than WSCP –Use exponential link cost function Applied REB strategy to conserve resource Applied REB strategy to conserve resource –Increase probability of success for shorter lightpath –As heavy load, algorithm with REB has excellent performance