ITU Kaleidoscope 2013 Building Sustainable Communities PROPOSAL OF A SUB-λ SWITHING NETWORK AND ITS TIME-SLOT ASSIGNMENT ALGORITHM FOR NETWORK WITH ASYNCHRONOUS TIME-SLOT PHASE Keisuke Okamoto Kyoto University okamoto@cube.kuee.kyoto-u.ac.jp Kyoto, Japan 22-24 April 2013 A label of contents Sub-λ switching network Multi-time-slot bonding Global-time-based scheduling Simulation analysis Kyoto, Japan, 22-24 April 2013 ITU Kaleidoscope 2013 – Building Sustainable Communities 1/11 Sub-λ switching network architecture All nodes are operated synchronously under a certain timecontrol mechanism The scheduler creates a schedule table, which controls the timing of optical time-slot switching Merit ・Low power consumption ・fine granularity ・headerless Kyoto, Japan, 22-24 April 2013 ITU Kaleidoscope 2013 – Building Sustainable Communities 2/11 Time-slot flow in sub-λ switching network Time-slot arrive at the optical switch with a certain propagation delay Kyoto, Japan, 22-24 April 2013 ITU Kaleidoscope 2013 – Building Sustainable Communities 3/11 Operation technique multi time-slot bonding Multiple optical bursts to the same destination are allocated in the successive time slots without guard-time single 3 Multi Time-slot Frame length 3 3 3 3 2 condense 3 3 3 3 3 Kyoto, Japan, 22-24 April 2013 ITU Kaleidoscope 2013 – Building Sustainable Communities 2 2 1 Gurad-time 2 2 2 1 Time 4/11 Problem time-slot assignment We have to add fiber cable for one time-slot duration at worst case to adjust transmission delay (ex:around 2km fiber for a 10 μs time-slot) objective Reducing unused time approach Scheduling algorithm under 1 time-slot length Delay shift packing algorithm 5/11 Approach global-time-based scheduling Reference node is selected in the network We shift the time-slot-search start point by the value of propagation delay Reference node 1 2 3 node 2→3 time Time Propagation delay node 1→2 FirstFit time Time-slot-based optical burst allocation is efficient when empty time-slots are successive 6/11 Shift decision Shift:time-slot position in frame We preselect reference node in the network We calculate shift volume from propagation delay between node i and reference node node 3→7 φ3+φ5 5 φ3 2 6 0 3 1 4 Reference node φ3+φ7 7 8 φ3 Time node 3→5 φ3 Time node 1→3 FirstFit Time 7/11 Shift lag We cannot calculate shift volume specifically in complicated network upstream downstream 3 D32 D23 1 D12 D21 -(D21+D32) 0 D24 2 4 D42 Reference node 2 3 D21 2 1 0 -(D21+D32) or D12-D32 4 D12 D12+D24 Reference node 1 8/11 Simulation setup Performance metric(increase ratio of minimum time-slot)= T:required number of time-slots determined using the timeslot assignment algorithm Tmax:required number of time-slots to accommodate all data on the most heavy link determined by routing Simulation model Toplogy:NSFNET(the number on the link is propagation delay [ms]) Traffic:1~995 Mbps at random routing:shortest path(metric is hop count) Kyoto, Japan, 22-24 April 2013 ITU Kaleidoscope 2013 – Building Sustainable Communities 9/11 Result Global time based scheduling improve the effective time of network capacity (local-time based scheduling: time-slot-search start point=0) Kyoto, Japan, 22-24 April 2013 ITU Kaleidoscope 2013 – Building Sustainable Communities 10/11 Conclusion We proposed a sub-λ switching network, which is all-optical, achieving finer granularity than a λ switching network We also proposed a multi-time-slot bonding technique and delay shift packing We simulated time-slot scheduling in NSFNET and demonstrated the efficiency of these algorithm Kyoto, Japan, 22-24 April 2013 ITU Kaleidoscope 2013 – Building Sustainable Communities 11/11