ANALYSIS AND SYNTHESIS OF OPTICAL BURST SWITCHED NETWORKS - PowerPoint Presentation

1. ANALYSIS AND SYNTHESIS OF OPTICAL BURST SWITCHED NETWORKSLi, ShuoSupervisor: Prof. Moshe ZukermanCo-supervisor: Dr. Eric W. M. WongFurther Credits: Dr. V. Abramov, Dr. Meiqian Wang and Zhang JiananJan. 06, 20141

2. OutlineBackground: Optical burst switching (OBS)Bounds for blocking probability obtained by Overflow Priority Classification Approximation (OPCA) in OBS networks with deflection routingEffective and ineffective utilizations in OBS networksEBSL – a combination of Emulated-OBS (E-OBS), segmentation and least remaining hop-count first (LRHF) Q & A2

3. OutlineBackground: Optical burst switching (OBS)Bounds for blocking probability obtained by Overflow Priority Classification Approximation (OPCA) in OBS networks with deflection routingEffective and ineffective utilizations in OBS networksEBSL – a combination of Emulated-OBS (E-OBS), segmentation and least remaining hop-count first (LRHF) Q & A3

4. Optical networks4Ever-increasing demand for higher bandwidthBandwidth intensive applications – voice over IP, video-on-demandFast increasing number of Internet usersSolution: Optical data communicationUse circuit switching (CS) & packet switching (PS) Drawbacks: CS: low bandwidth efficiency PS: buffer & high energy consuming

5. Optical Burst Switching (OBS)OXC: optical cross-connectPackets with the same destination are aggregated at ingress nodes to form burstsA control packet is sent ahead of a burst to reserve wavelength channels along the transmission path hop by hopBursts may be dumped before reaching their destinationsA trunk Dump5Trunk: A group of fibers connecting two OXCs.

6. OutlineBackground: Optical burst switching (OBS)Bounds for blocking probability obtained by Overflow Priority Classification Approximation (OPCA) in OBS networks with deflection routingEffective and ineffective utilizations in OBS networksEBSL – a combination of Emulated-OBS (E-OBS), segmentation and least remaining hop-count first (LRHF) Q & A6

7. Network model7SourceWACA1CA1CA2TXGADestinationMDILMACDNYMASourceMDILMACDNYMADestinationWACA1CA1CA2TXGAIndependent Poisson process of arrivalsHolding times - independently, exponentially distributed with unit meanFull wavelength conversionThe offered load to each source-destination (SD) pair is identical

8. One Contention Resolution MethodBlocking probability8Performance study of OBS networks with deflection routingDeflection routing

9. Erlang Fixed Point Approximation (EFPA)9Overflow error -- ignore high variance of deflected traffic and dependence Path error-- ignore the effect of traffic smoothing, and the positive correlation of trunk occupancy along the path that increases the probability to admit burstsdecouple a given system into independent trunkstraffic offered to each trunk follows an independent Poisson process

10. 10Overflow Priority classification Approximation (OPCA)Define a surrogate model based on classifying the traffic into different layers (priorities)Layer i for traffic deflected i timesStrict priority regimeJunior bursts – higher prioritySenior bursts –lower priorityThe surrogate is without inter-layer mutual dependence (but may still have intra-layer mutual dependence)Solve the surrogate system by applying EFPA-like algorithm in each layer

11. 11Why OPCA?Calculation taskRunning time of EFPA in secondsRunning time of OPCA in secondsBlocking probability of the whole network and C=500.2710.197Blocking probability of the whole network and C=200064.4512.91Blocking probability of the whole network and C=100003006397Blocking probability of the whole network and C=20000136651232In our example, OPCA needs less time than EFPAComparison of the times used by EFPA and OPCA to calculate the blocking probabilities in the NSFNetD=3C-number of channels per trunkOffered load to each SD pair is 0.5Conly consider 4 significant digits of the fixed-point solutions whenWhen , set j: trunk number k: deflection times

12. Why OPCA?12In our example, OPCA needs less time than EFPAIn practical range, OPCA is more accurate than EFPA and generally it is not worseD=3C=50in the practical loading range, EFPA does not performs better than OPCAEFPA is only more accurate than OPCA when the offered load is within 35–40when the offered load is within 40–50, EFPA cannot converge

13. Objectives13Provide the upper and lower bounds for the blocking probability obtained by OPCA Understand and mathematically prove the conditions under which the bounds draw near each otherFind a way to make the bounds converge faster, use them to find solutions for OPCA

14. Numerical results14C=50Offered load to each SD pair: 30 ErlangsWhen no. of iterations = 7,  

15. Summary15Prove that the upper and lower bounds draw near each otherNumerically demonstrate that the bounds become closer to each other very fast

16. OutlineBackground: Optical burst switching (OBS)Bounds for blocking probability obtained by Overflow Priority Classification Approximation (OPCA) in OBS networks with deflection routingEffective and ineffective utilizations in OBS networksEBSL – a combination of Emulated-OBS (E-OBS), segmentation and least remaining hop-count first (LRHF) Q & A16

17. Performance study of OBS networksBlocking probabilityUtilizationOccupied by bursts that: successfully transmitted or dumped before reaching the destinations 17

18. ObjectiveTo gain insight into the efficiency and performance of OBS networksUtilization (U) [%]Effective Utilization (EU) [%]Ineffective Utilization(IU) [%]Channels used by bursts that eventually reach their destinationsChannels used by bursts that are dumped before reaching their destinationsGoodput[Erlangs]Traffic that successfully reach the destinations18

19. An ExampleTrunk 1:U: 50%EU: 0%IU: 50%Trunk 2:U: 100%EU: 50%IU: 50%Trunk 3:U: 100%EU: 100%IU: 0%Node ANode BNode CNode DBurst ADBurst ADBurst BDBurst CDBurst BDFree channelBusy channel occupied by a burst that can reach its destination EUBusy channel occupied by a burst that dumped before reach its destination  IU19Burst AD is dumpedBurst AD: from A to D through B and CBurst BD: from B to D through CBurst CD: from C to D

20. Network Models2014-node NSFNet6-node ringWith only 3-hop SD pairsWith all possible SD pairs200200200200200200

21. Our Simulation ScenarioIndependent Poisson process of arrivalsTransmission rate: 10 Gb/sFixed packet size: 1250 Bytes/packetBurst size: Exponential distribution (rounded) with mean 250 packets/burstMean burst transmission time: 250 μs1-hop offset time: 10 μs Switching time is ignoredNumber of channels on each trunk – 50Shortest path for each source-destination (SD) pair Full wavelength conversion21

22. Our Simulation ScenarioI-OCS is used as a benchmark for OBSidealized version of optical circuit switchingignore inefficiency associated with reservation and takedownIn I-OCS only EU, no IU22

23. Results in ring networkIn OBS, we observe goodput collapse under overload conditionsIn I-OCS, goodput asymptotically approaches the available capacity23C=50

24. Results in ring network24C=50

25. Results in the NSFNet25C=50

26. Results26In NSFNet: Under heavy traffic, most of the successfully transmitted bursts are 1-hop pairs  guarantee a certain level of effective utilization and goodputC=50

27. SummaryEffective and ineffective utilizations are key factors affecting performance and efficiency of OBS networksThey explain a weakness of OBS under high traffic load conditions leading to goodput degradation way below its I-OCS benchmarkUnderstanding these key effects is important for understanding and improving performance and efficiency of OBS networks27

28. OutlineBackground: Optical burst switching (OBS)Bounds for blocking probability obtained by Overflow Priority Classification Approximation (OPCA) in OBS networks with deflection routingEffective and ineffective utilizations in OBS networksEBSL – a combination of Emulated-OBS (E-OBS), segmentation and least remaining hop-count first (LRHF) ConclusionQ & A28

29. Goodput and effective utilization degradations29

30. Objective30 Building on the concept we have introduced of effective utilization, we aim to increase effective utilization in order to increase goodput & reduce the network blocking probability. Least Remaining Hop-Count First (White et al.) : Let the bursts which have already used more network resources have a higher probability to reach their destinationsOffset-Time-Emulated OBSSolution: EBSLSegmentationLeast Remaining Hop-Count First

31. Least Remaining Hop-Count First (LRHF) (White et al.)31Bursts with fewer remaining hops have higher priority.When all the channels on the output trunk are fully occupied, a new higher priority burst can preempt a lower priority burst on the output trunk.The entire preempted lower priority burst is then dropped.Problems: Preempting the entire burst is not efficient Difficult to control in a distributed system

32. Segmentation32A burst is divided into several segments.One segment contains one packet or several packets.When contention happens, instead of dropping the whole contending burst, only the overlapped segments are dropped.

33. Just-Enough-Time (JET)33The burst control packet carries the information about the burst arrival time, burst length and the wavelength usedThe reservation is made from the time when the first bit of the burst reaches that node until the transmission finish

34. Offset-Time-Emulated OBS (E-OBS)34An additional fiber delay unit (FDU) is inserted in the data path at every core node.∆ is the 1-hop offset time corresponding to the queuing and processing delay of one node. is the switching delay

35. EBSL35A new burst with n-hop path has priority nIts priority increases by one level every time when it accesses to a new hopFirst try to find free channelsIf no free channels, find lower priority bursts transmitted on the output trunk

36. Fair EBSL (F-EBSL)36First try to find free channelsIf no free channels, find if any bursts transmitted on that trunk:have lower priorityoriginally require a path with an equal or lower number of hopsTo protect bursts that require long routes

37. Our Simulation ScenarioIndependent Poisson process of arrivalsTransmission rate: 10 Gb/sFixed packet size: 1250 Bytes/packetBurst size: Poisson distribution with mean 250 packets/burstMean burst transmission time: 0.25 ms1-hop offset time: 10 μs Switching time is ignoredNumber of channels on each trunk – 50Shortest path for each source-destination (SD) pair as primary routeFull wavelength conversion37

38. Network Models386-node ring

39. Results For EBSL39Utilization: almost the sameEffective utilization: significantly increase in EBSL under heavy load conditions!Only 3-hop SD pairsC=50

40. Results For EBSL40With the same offered load:goodput is increased  more bursts are successfully transmitted  the blocking probability is reduced more resources are used effectively  the goodput of the network also increases significantly Only 3-hop SD pairsC=50

41. Results For F-EBSL41EBSL discriminates against traffic that requires more hops in favour of traffic that requires fewer hopsUnder F-EBSL, more 3-hop bursts successfully reach their destinationsAll 3-hop SD pairs and 2-hop SD pairsC=50

42. Results For F-EBSL42The performance of F-EBSL is similar to I-OCSAll 3-hop SD pairs and 2-hop SD pairsC=50

43. Summary43We have introduced the EBSL and F-EBSL strategies to solve the burst contention problem.Numerical results show that EBSL can significantly increase the effective utilization and eliminate the collapse of goodput, and improve QoS. F-EBSL partly sacrifices performance to provide higher probability for the bursts that require more hops to successfully reach their destinations.

44. OutlineBackground: Optical burst switching (OBS)Bounds of the Overflow Priority Classification (OPC) for blocking probability approximation in OBS networks with deflection routingEffective and ineffective utilizations in OBS networksEBSL – a combination of Emulated-OBS (E-OBS), segmentation and least remaining hop-count first (LRHF) Q & A44

45. Q & AThank you for your attention^_^45

46. Poisson arrival and Exponential service time 46Poisson Pareto Burst Process is a good traffic model for real network trafficIn OBS networks, blocking probability is insensitive to the shape of the distribution of the service time J. Chen, R. G. Addie and M. Zukerman, "Performance Evaluation and Service Rate Provisioning for a Queue with Fractional Brownian Input," Performance Evaluation, vol. 70, no. 11, pp. 1028-1045, November 2013J. Zhang, Y. Peng, Eric W. M. Wong and M. Zukerman, "Sensitivity of Blocking Probability in the Generalized Engset Model for OBS," IEEE Communications Letters, vol. 15, no. 11, pp. 1243-1245, November 2011

47. Why OPCA?47In our example, OPCA needs less time than EFPAAlgorithmLayer numberNumber of iterationsTotal running time in secondsEFPAOnly 1 layer783006OPCALayer 06177.9Layer 15119.7Layer 2499.7Layer 310.0024Comparison of the times used by EFPA and OPCA in each layer to calculate the blocking probabilities in the NSFNet with 10000 channels per trunkTotally 16

48. Numerical results48Bounds of OPCA blocking probabilities in the NSFNet with different offered load to each directional SD pairoffered load bounds closer speed

49. 49Numerical resultsBounds of OPCA blocking probabilities in the NSFNet with different number of channels per trunk (C) in which the offered load to each directional SD pair is 0.4Coffered load bounds closer speednumber of channels per trunk bounds closer speed Reason: a larger number of channels per trunk the variance of the number of busy channels is lower smaller (deflected bursts/total bursts) in the networks

50. Numerical results50Bounds of OPCA blocking probabilities in the NSFNet with different maximum allowable number of deflections (D) in which the offered load to each directional SD pair is 20 Erlangsoffered load bounds closer speednumber of channels per trunk bounds closer speedD bounds closer speed

51. Summary51Prove that the upper and lower bounds draw near each other with increasing number of iterationsNumerically demonstrate that the bounds become closer to each other very fastThe speed of the bounds moving closer decreases when the proportion of the deflected traffic increases in the network, due to the growth of the offered load or the maximum allowable number of deflections, as well as the reduction of the number of channels per trunk

52. EBSL with deflection (EBSL-D)52One channel each trunk

53. EBSL with deflection (EBSL-D)53One channel each trunk

54. EBSL with deflection (EBSL-D)54One channel each trunk

55. Bounds for the blocking probabilities of loop based trunks55is a decreasing function ofis an increasing function ofk: number of deflection times a set of SD pairs 

56. Bounds for network blocking probability56

57. d=0Calculate offered load for each trunkCalculate blocking probability for each trunkConverge or not?d+1YESNoSteady state probabilitiesInitial values of trunk blocking probabilityNetwork blocking probability d=D or not?NoYES57D: maximum allowable number of deflection

58. Fair EBSL (F-EBSL)58First try to find free channelsIf no free channel, find if any bursts transmitted on that trunk:have lower priorityoriginally required a path with an equal or lower number of hopsTo protect bursts that require long routes

59. EBSL with deflection (EBSL-D)59Under EBSL-D, segmentation always happens before deflectionOnce a burst or a segmented part of a burst is deflected, its priority will be set to L+1 and its priority will not increase when it completes each one hop transmissionL: total number of trunks in the networkThis guarantees that the deflected bursts always have the same lowest priority in the network  no instability problem under heavy load conditions

60. Results For F-EBSL60EBSL discriminates against traffic that requires more hops in favour of traffic that requires fewer hopsUnder F-EBSL, more 3-hop bursts successfully reach their destinationsAll 3-hop SD pairs and 2-hop SD pairsC=50

61. Results For F-EBSL61The performance of F-EBSL is similar to I-OCSAll 3-hop SD pairs and 2-hop SD pairsC=50

62. Results for EBSL-D under heavy load conditions62C=50

63. Results for EBSL-D under light and medium load conditions63C=50

64. Direct trunks and loop based trunks64nodetrunktrunktraffic flow3 SD pairs:AD DH CBLoop based trunks: A set of trunks is a loop based trunk (LBT) set in layer , if the following hold: has at least one closed loop of trunks in layer If in layer , trunk receives traffic from any trunk in , then Direct trunks: All the trunks that are not LBTs in layer are called direct trunks in layer A trunk where there is no traffic in a trunk in layer A set of trunks that form an isolate tree structure in layer A set of trunks that feed traffic to LBTs but do not receive traffic from any LBT in layer