1. ALL-OPTICAL PACKET HEADER PROCESSING SCHEME BASED ON PULSE POSITION MODULATION IN PACKET-SWITCHED NETWORKS Z. Ghassemlooy, H. Le Minh, Wai Pang Ng Optical Communications Research Group Northumbria University, UK http://soe.unn.ac.uk/ocr/

2. Contents Overview of header processing in optical networks Header processing based on pulse-position modulation (PPM) Proposed node architecture Simulation results Summary

3. Optical Communication Network (OCN)  Solution: All-optical processing & switching 1P 100T 10T 1T 100G 10G 1G 100M 1995 2000 2005 2010 Year Demand traffic [bit/s] Voice Data Total NEC-2001 - Future OCNs: faster signal processing and switching to cope with the increase of the demanding network traffic - Existing OCNs: depends on electronic devices for processing the packet address to obtain the routing path. However, the limitation of electronic response will cause the speed bottleneck

4. Future OCNs Optical transparent path - Future OCN will have the processing and switching data packets entirely in optical domain, i.e. generate optical transparent path for routing data packets  Require: compact and scalable processing scheme

5. Current All-optical Processing Schemes All-optical logic gates All-optical correlators Address patterns Decimal value Output ports 0 0 0Port 2 0 0 0 11Port 1 0 0 1 02Port 3 0 0 1 13Port 1 0 1 0 04Port 3 0 1 5Port 2 0 1 1 06Port 2 0 1 1 17Port 1 1 0 0 08Port 3 1 0 0 19Port 2 1 0 Port 2 1 0 1 111Port 3 1 1 0 012Port 1 1 1 0 113Port 1 1 1 1 014Port 2 1 1 15Port 1 Routing table (RT) Example: N = 4, node with M = 3 ? Port 1 Port 2 Port 3 N-bit Problems: Large size routing table  increased processing time Optical device complexity  poor scalability Solution: To reduce the size of the routing table

6. PPM - Operation Address extraction a 0 a 1 a 2 a 3 payload Header (packet address) Clk Data packet PPM (a)(b) (a)(b)

7. PPM Based Routing Table Grouping address patterns having the same output ports Each new pulse-position routing table (PPRT) entry has optical pulses at the positions corresponding to the decimal values of group’s patterns Pulse-position routing table (N = 4, M = 3)

8. Header Correlation Single AND operation is required for matching PPM-address and multiple address patterns (PPRT entry) Processing-time gain:

9. Proposed Node with PPM Processing Clock extraction: synchronize the arrival of data packet and the node processing S-P converter: convert the serial address bits to parallel bits PPM-ACM: (PPM address conversion module): convert binary address to the PPM-converted address PPRT: store M entries (M PPM frames) Switch synchronisation: synchronise SW with data packet All-optical switch: controlled by matching signals to open the correct SW Clock extraction S-P Converter PPM-ACM & M SW1 SW2 SWM Header processing unit 1 2 M All-optical switch... Data H C lk PPRT Entry 1 Entry 2 Entry M... & 1 & 2 Switch Sync. Data H C lk H

10. PPRT with Multimode Transmission Same address pattern can appear at multiple PPRT entries Modes: unicast, multicast, broadcast and deletion Pulse-position routing table (N = 4, M = 3)

11. Node with Multicast Tx Mode Clock extraction S-P Converter PPM-ACM & M SW1 SW2 SWM Header processing unit 1 2 M All-optical switch... Data H C lk PPRT Entry 1 Entry 2 Entry M... & 1 & 2 Switch Sync. Data H C lk H Data H C lk

12. Optical PPM Generation Circuit PPM-format address: y(t) = x(t +  i a i  2 i  T s ) N-bit address-codeword: A = [a i  {0,1}], i = 0, …, N–1

13. PPRT Generation Is self-initialised with the extracted clock pulse. The M entries are filled by: – Single optical pulse + Array of 2 N optical delay lines; Or, – M pattern generators + M optical modulators.

14. Ultrafast Optical AND Gate A/B01 000 101 Implementation: Using optical interferometer configuration + optical nonlinear devices A B A×BA×B SOA1 SOA2 Symmetric Mach-Zehnder Interferometer (SMZI)

15. All-Optical Switch 1 M1 M SMZ-1 SMZ-2 SMZ-M … CP1 CP2 CPM 1 2 M

16. Simulation Results Simulation parameters Values Address length N5 Number of outputs M3 Bit rate50 Gb/s Payload16 bits Packet gap2 ns Pulse width FWHM1 ps Pulse’s power peak2 mW Wavelength1554 nm PPM slot duration T s 5 ps For an all-optical core network up to 2 5 = 32 nodes

17. Simulation Results Demonstrate the PPM processing and Tx modes PPRT with 3 entries:

18. Simulation Results Input Output 1 Output 2 Output 3

19. Simulation Results 0 1 1 1 0 Packet with address 01110 PPM-converted address PPRT entry 1 Synchronized matching pulse

20. Conclusions PPM processing scheme – Reduces the required processing time – Provides the scalability: adding/dropping network nodes and node outputs Applications: – All-optical core/backbone networks (N >> M ~ 3-6) – Optical bypass router (electrical router + optical bypass router) Challenges: – Optical switch with long and variable switching window – Timing jitter and received pulse dispersion

21. Publications H. Le-Minh, Z. Ghassemlooy, and W. P. Ng., “A novel node architecture for all-optical packet switched network”, proceeding of 10 th European Conference on Networks and Optical Communications 2005 (NOC2005), pp. 209-216, London, UK, Jul. 2005 H. Le-Minh, Z. Ghassemlooy, and W. P. Ng., ”Ultrafast header processing in all-optical packet switched-network” proceeding of 7 th International Conference on Transparent Optical Networks 2005 (ICTON2005), Vol. 2, pp. 50-53, Barcelona, Spain, Jul. 2005

22. Acknowledgements Northumbria University for sponsoring the research work

23. Thank you!