Nobel-WP3 Routing in OBS

1. Nobel-WP3Routing in OBS Miroslaw Klinkowski Davide Careglio Josep Solé-Pareta NOBEL meeting, Munich

2. Two different approaches • Put the traffic where the bandwidth is • Routing approach • Need adaptive routing mechanism • Put the bandwidth where the traffic is • Planning / dimensioning approach • Need simple shortest path algorithm or similar mechanism NOBEL meeting, Munich

3. Routing in OBS General assumptions • Signaling: On the fly • Resources allocation: JET • Nodes equipped with FDLs • Resource assignment based on: • Horizon scheme: MINGAP, MINLEN • Void Filling scheme: VF, VF-start, VF-maxmatching NOBEL meeting, Munich

4. Routing in OBS A critical feature • Set of the offset-time in the control packet Offset-time is calculated in the edge node on base of a sum of processing times in all consecutive nodes lying on the path the burst is transmitted on. In order to calculate optimal (minimal) value of offset-time the number of nodes the burst is going through should be know Deflection routing scheme allowed only if offset-time is high enough Note: For small processing times (tens of [ns] ?) these constraints may not be so significant Question: Processing at nodes consists of…? NOBEL meeting, Munich

5. Routing in OBS Contribution to the D23 • Connection-oriented environment with LSP paths and source-base routing • Shortest path (SP) routing • Isolated routing based on bypass based routing (BBR) • Kind of deflecting routing: when an intermediate node the path is found obstructed (all wavelength of the designated output port are congested), that node is bypassed • Connection-less environment with hop-by-hop decisions • Isolated routing based on hot potato mechanism NOBEL meeting, Munich

6. Routing in OBS Contribution to the D23 • Network scenarios • Simple 6-nodes test network • GEANT • NSFNet • Results • Burst Loss Probability • End-to-end delay NOBEL meeting, Munich

7. Routing in OBS Contribution to the D32 • Distributed routing based on periodically distribute LSP/node state information throughout the network • Feasible approach for both connection-oriented and connection-less environments NOBEL meeting, Munich

8. Routing in OBS Ideas (I) • Taken from transporting datagrams over ATM networks (*): • A number of LSPs (e.g. through the k-shortest paths) are pre-established between every source-destination pair of nodes • Bursts are routed through one LSPs according to their destination and a probability assigned to each LSP • The probabilities are dynamically updated (every specific period of time) according to the feedback information that source nodes get from the network about the state of the LSPs/transist nodes (*) J. Sole-Pareta, D. Sarkar, J. Liebeherr, I.F. Akyildiz, “Adaptive multipath routing of connectionless traffic in an ATM network”, ICC 95 Seattle “Gateway to Globalization” 1995 IEEE International Conference , vol. 3, pp.1626 – 1630, 18-22 Jun. 1995. NOBEL meeting, Munich

9. Routing in OBS Ideas (II) • Adaptation of the BBR concept (*) with multipath routing • A number of LSPs allowing to bypass possible obstructed nodes are pre-established • A feedback information about congested links/links on the LSPs is obtained from the network (e.g., periodically) • Since the feedback information is inaccurate, then congested we consider them only OSL (Obstruct Sensitive Links) • Source node send the burst control packets through a LSP including the information of OSL links and how to bypass them (*) X. Masip-Bruin, S. Sanchez-López, J. Solé-Pareta, J. Domingo-Pascual, “QoS routing algorithms under inaccurate routing information for bandwidth constrained applications”, in Proc. IEEE 2003 International Conference on Communications (ICC 2003), Anchorage, AL, May. 2003, pp. 1743-1748. NOBEL meeting, Munich

10. Routing in OBS General assumptions • Assumptions for the routing problem • Source-based: the routing decision at the edge node • Connection-oriented: a MPLS like network is considered • Multipath: a number of LSP paths pre-established • Adaptive • adaptation of weights assigned to LSPs • dynamic reconfiguration of the LPS on base of the network state information • Per-burst routing decision (LSP selection) • Processing at nodes: decide the wavelength to forward the the incoming bursts NOBEL meeting, Munich

11. Routing in OBS Signalling • Feedback information from the network • Periodical – with information about state of the links/nodes (estimated BLPs, resources occupancy, …) • Event driven – e. g., in case of loosing a burst • This information could contain the label of LSP path, the node id. where the burst was lost, the class of the burst lost, … • Such information would be send only towards the edge node that sent the burst lost • Combination of both NOBEL meeting, Munich

12. Edge node Link state information DB LSP’s statere-calculation 2 4 5 6 3 1 Link BLP … Core node 1-2 10-4 Dynamic re-configuration 2-3 10-3 Sending of information towards the edge nodes, either periodically or when an event occurs Maintaining LSPs 2-5 10-4 Update … The selection of pre-established LSP from the table is performed for each new burst A feedback (e.g. periodical with links’ state inf. or incidental like when burst dropped, …) LSPs’ table Dest LSP Prob. … A feedback information from the network 1-2-3-4 0.3 4 1-2-5-6-4 0.7 … Routing in OBS General architecture NOBEL meeting, Munich