06117054

Upload: shahabniazi

Post on 04-Jun-2018

222 views

Category:

Documents


0 download

TRANSCRIPT

  • 8/13/2019 06117054

    1/7

    222 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 30, NO. 2, JANUARY 15, 2012

    A Novel Contention Resolution Scheme ofHybrid Shared Wavelength Conversion for

    Optical Packet SwitchingHuan-Lin Liu, Bin Zhang, and Song-Lei Shi

    AbstractWavelength conversion is an efficient way to improvethe optical packet blocking performance for all-optical packet

    switching networks. The synchronous OPS node architecture,with a hybrid shared wavelength conversion (HSWC) schemebased on shared per outputfiber and shared per node wavelengthconversion, is put forward to reduce the packet loss probability inthis paper. A heuristics is developed into assigning the wavelengthchannels for packets in the proposed optical nodes. The limited

    range wavelength converters first available (LFA) algorithm and

    parametric wavelength converters first available (PFA) algorithmare used to schedule contending optical packets to available wave-length channels. Simulation results show that the hybrid shared

    wavelength conversion scheme has significant performance im-provement in terms of packet loss probability. And the utilization

    of wavelength converter is greatly improved for OPS node withhybrid shared wavelength conversion scheme. Furthermore,the PFA scheduling algorithm compared with LFA algorithm isvalidated to improve the wavelength assignment efficiency and

    reduce the required total wavelength converters for OPS nodes.

    Index TermsContention resolution, limited range wavelengthconverter, optical packet switching, packet loss probability, para-

    metric wavelength converter.

    I. INTRODUCTION

    A S the demand for link speed and bandwidth keeps in-creasing, wavelength division multiplexing (WDM)optical network that supports wide bandwidth are believed

    to be future backbone transport networks. Optical switching

    technologies are motivated to rapidly exchange information

    by the enormous bandwidth carried out on any optical fiber.

    Among all of the switching schemes, optical packet switch

    appears to be a strong candidate with the high speed, data

    rate, format transparency and configurability for high-speed

    networks [1][4].

    Manuscript received February 18, 2011; revised July 24, 2011, November 06,2011; accepted December 20, 2011. Date of publication December 28, 2011;date of current version January 26, 2012. This research was funded by the sci-entific research fund of Chongqing municipal commission (KJ110527), and byChongqing Natural Science Foundation (CSTC 2010BB2413), and by nationalnature science foundation of China (NSFC 60972069, 61071117).

    H.-L. Liuand S.-L. Shiare with theKey Laboratory of Optical Fiber Commu-nication Technology, Chongqing University of Posts and Telecommunications,Chongqing 400065, China (e-mail: [email protected]).

    B. Zhang is with School of Economics and Management, Beijing Uni-versity of Posts and Telecommunications, Beijing 100876, China (e-mail:[email protected]).

    Color versions of one or more of the figures in this paper are available onlineat http://ieeexplore.ieee.org.

    Digital Object Identifier 10.1109/JLT.2011.2182036

    In OPS network, contention arises when thereare twoor more

    incoming optical packets at the same wavelength contending for

    the same output wavelength channel. How to resolve the optical

    packets contention, therefore, has become a critical technology

    for OPS nodes. The first candidate for contending resolution

    is to use wavelength conversion that exploits the wavelength

    domain, although other resolution mechanisms also exist, such

    as fiber delay lines and deflection routing [5][8].

    Thus, wavelength converters become the most impor-tant components for contention resolution in wavelength

    domain [8][10]. An ideal full range wavelength con-

    verter (FRWC), however, has to be realized by cascading

    a given number of realistic limited range wavelength con-

    verters (LRWCs). Wavelength converters are costly devices

    and therefore, a particular effort has been devoted to de-

    signing cost-effective optical node architectures by minimizing

    the number of wavelength converters constrained to a pre-

    scribed packet loss probability [11][16]. Several OPS node

    architectures, such as single per link (SPL), share per output

    fiber (SPOF), and share per node (SPN), sharing limited range

    wavelength converters, was proposed in [11]. Also, it wasvalidated by simulation that the SPN architecture employing

    with LRWCs (SPN-LRWC), was the greatest cost-saving of the

    three architectures.

    Recently, the shared per node OPS architecture with para-

    metric wavelength converters (PWC) is proved that it can

    reduce the required number of wavelength converters [15],

    [16]. A PWC allows multiple wavelengths to be simultaneously

    converted from one to another. The PWC is considered to be

    becoming feasible. A number of simultaneous multi-channel

    wavelength conversions of over 30 channels have been reported

    using fiber [17] and a LiNbO waveguide [18]. A high-capacity

    field transmission experiment using multi-channel wavelength

    conversion has also been demonstrated [19]. Moreover, the

    parametric process is fully transparent to various types of

    advanced modulation formats such as differential phase shift

    keying (DPSK) [20] and quadrature phase shift keying (QPSK)

    [21]. These studies show that guard bands can be provided with

    suitable channel spacing.

    In order to further reduce the packet loss probability and

    improve the utilization of wavelength converters, this paper

    presents a hybrid shared wavelength conversion scheme. The

    proposed scheme mixes shared per node with LRWCs and

    shared per output fiber with PWCs. Thus, the affection of guard

    bands in PWCs can be eliminated by the use of LRWCs. There

    are two different scheduling algorithms to assign wavelength

    0733-8724/$26.00 2011 IEEE

  • 8/13/2019 06117054

    2/7

  • 8/13/2019 06117054

    3/7

    224 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 30, NO. 2, JANUARY 15, 2012

    packet loss probability of OPS would be further decreased by

    using the hybrid shared wavelength conversion scheme.

    III. SCHEDULINGALGORITHM

    Since HSWC is a bufferless contention resolution for the

    OPS node, the packet waiting time in the OPS node would be

    smaller compared with the delay of the access network andcan be ignored. A major concern in the switch is packet loss,

    which is owed to be lack of output wavelength channel, be

    lack of wavelength converters or conversion limited. Because

    two different wavelength converters are employed, as shown

    in Fig. 1, there are two different scheduling algorithms on

    which kind of wavelength converters firstly selected: LRWC

    first available (LFA) algorithm and PWC first available (PFA)

    algorithm, respectively. Actually, not all ports of a PWC are oc-

    cupied simultaneously, so that PWCs are preferred to improve

    the utilization of PWC. In this part, we mainly introduce the

    PFA algorithm because the two algorithms are similar except

    the priority of wavelength converters in the last phase.The PFA algorithm assigns the LRWCs and PWCs for the

    scheduled packets to available output wavelength channels,

    and it would be referred to the allocation of LRWCs, PWCs and

    wavelength channels. As shown in Fig. 4, the PFA algorithm

    includes three different phases that are performed sequentially:

    initialization (INI) phase, direct output wavelength channel

    assignment (DOWCA) phase, and LRWCs and PWCs as-

    signment (LPA) phase. Before describing the PFA algorithm

    adopted in the HSWC architecture, the following notations are

    introduced:

    1) , similar to ,

    represents an arrival optical packet carried by wavelength

    and directed to the output fiber . The packet isdenoted as the -th packet arriving on wavelength

    and directed to output fiber .

    2) denotes output wave-

    length channel at output fiber with wavelength .

    3) is the set containing packets that are waiting for being

    scheduled wavelength channels. It is consist of packets

    , and maybe there are two or more same packets

    coming from different input fibers, simultaneously.

    4) is the set containing idle wavelength channels. It is ini-

    tialized as , containing the all output wavelength chan-

    nels . is denoted that

    the number of idle channels currently in is .5) is the set of all possible packets that

    can be scheduled to the output wavelength channels

    by wavelength conversion. The set is obtained

    by using only PWCs, is obtained by using only

    LRWCs, and is obtained by using both PWCs

    and LRWCs.

    In the INI phase, the main variables used by the algorithm

    are initialized. The input wavelength channels of each input in-

    terface are scanned and the incoming packets carried by wave-

    length are collected in the set . The set is initialized as

    , here . Finally, the number of LRWCs is initialized

    as , and the number of PWCs at each outputfiber is .

    In the DOWCA phase, the wavelength channels of the output

    fiber are sequentially scanned. When the wavelength channel

    Fig. 4. PWCfirst available (PFA) scheduling algorithm flow chart.

    is considered, the packets in the set are scanned. If

    there exist such packets as from different input wavelength

    channel in the set , one of the packets randomly selected, is

    scheduled to be on the output wavelength channel . Then

    one packet and the channel are removed from and

    , respectively. If there is no packet in the set , the

    wavelength channel remains available and it may serve

    contending packets in the LPA phase. The computational com-

    plexity of DOWCA phase is for each output, because

    each output wavelength channels have to be scanned.

    In the LPA phase, the remaining packets direct to the outputfiber are gathered in the set . Notice the LPA phase contains

    three parts: PWCs assignment, LRWCs assignment and both

    PWCs and LRWCs combined assignment, which are denoted

    by the variable in Fig. 4, respectively. As

    long as and re not empty and at least one wavelength

    converter is available, the set is scanned from to

    . Firstly, PWCs are selected for converting the packets to

    idle output wavelength channels by canned from to

    . When the channel is scanned, if there exist a set

    belonging to without the intersection of and empty

    and the PWCs on the output fiber are available, a packet

    selected randomly in , is scheduled on the output wavelength

    channel . Then the packet and the channel are re-

    moved from and , respectively. And the state of the PWC

  • 8/13/2019 06117054

    4/7

    LIUet al.: A NOVEL CONTENTION RESOLUTION SCHEME OF HYBRID SHARED WAVELENGTH CONVERSION FOR OPS 225

    Fig. 5. An example of packets schedule for an output fiber carrying 4 wavelengths. (a) Five arriving packets. (b) S cheduled packets at the end of the DOWCAphase. (c) Scheduled packets at the e nd of the LPA phase.

    used is updated. Secondly, LRWCs are selected by scanned

    from to . When the channel is scanned, if

    there exist a set belonging to without the intersection of

    and empty and available LRWC, a packet se-

    lected randomly in , is scheduled to the output wavelength

    channel . Then and are updated. And set .

    Finally, both LRWCs and PWCs are used by canned fromto When the channel is scanned, if there

    exist a set belonging to without the intersection of and

    emptyand there are LRWCs and PWCs available, a

    packet selected randomly in , is scheduled to the output

    wavelength channel . Then and are updated. Both the

    state of the PWC used and the variable are updated, too. In

    the end, the packets still in the set are discarded. Given that

    in the worst case, the computational complexity of LPA phase

    is for each outputfiber because of scanning

    the output wavelength channels for three conversion methods.

    Therefore we can confirm that the PFA algorithm computa-

    tional complexity is obtained by the sum of the complexity in

    the DOWCA phase and LPA phase as

    for each output fiber. Because all of the output fibers

    have to be scanned, the computational complexity of PFA algo-

    rithm in the HSWC architecture is .

    The LFA scheduling algorithm, different from PFA algorithm

    in the LPA phase, tries to find firstly whether available LRWCs

    can solve the contention, then whether available PWCs, and last

    whether combination of a LRWC and a PWC. If one of the three

    methods could solve the contention, the next channel will beconsidered like this until the circle is end.

    Fig. 5 shows an example of packets scheduling in the OPS

    node architecture with HSWC. The architecture includes

    input fibers and output fibers and each fiber supporting

    wavelengths.

    In Fig. 5(a), four packets are arriving on wavelength and

    one packet is arriving on wavelength . Given the available

    number of LRWCs , conversion range and output

    fiber . And there are one PWC shared for each output fiber

    without guard bands. The ratio of coupler before PWCs is 2:1,

    meaning that the parametric is 2. In particular, the conversion

    process of PWC includes and .

    The actions performed in DOWCA phase are described by

    Fig. 5(b). Packets and are scheduled for the channel

  • 8/13/2019 06117054

    5/7

    226 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 30, NO. 2, JANUARY 15, 2012

    and , respectively. The rest idle wavelength channels

    will be scheduled for packets in LPA phase.

    The results of packets scheduled at the end of LPA phase are

    described in Fig. 5(c). For the channel , there is no packet

    can be scheduled just by PWC or just by LRWC. So LRWC and

    PWC have to be combined for a packet. Firstly, the wavelength

    carrying the packet is converted to wavelength by

    a LRWC. Then the PWC can be used to change the wavelength

    from to . Thus the packet has been scheduled for the

    channel . As for , there are no packets can be scheduled

    directly. But wavelength carrying packets can be converted

    to by the PWC. So a packet can be selected randomly from

    the rest packets, suchas . Intheend,both ofthe PWC input

    ports are occupied and one of the LRWCs is used. The packet

    has to be dropped because of no idle output wavelength

    channel.

    IV. SIMULATION ANALYSIS ANDDISCUSSION

    Firstly we discuss optical packet loss probability (PLP) per-formance among HSWC, SPN-PWC and SPN-LRWC architec-

    tures. Provided the OPS nodes are synchronous, meaning that

    all arriving packets have a fixed size and these arrived on each

    wavelength are synchronized in a time-slot. It is assumed that

    thepackets arrived on theNWinput wavelength channels at each

    time slot are independent, and a packet on each input wave-

    length channels occurs with probability that is equal to traffic

    load. Also, each arriving packet has the same probability

    direct to one of any output fibers. In the following simulation,

    we assume that the output or input fiber number equals 4, and

    wavelength channel number equals 32 for the three architec-

    tures. Especially, PWCs number equals 1 for each outputfiber

    in HSWC architecture.In order to contrast the performance of SPN-PWC,

    SPN-LRWC, and HSWC, we keep equal number of wave-

    length converters in the three architectures. The sum of PWCs

    number and the LRWCs number in HSWC architecture is equal

    to LRWC number in SPN-LRWC, and also equal to PWCs

    number in SPN-PWC. The parameters of LRWCs in HSWC

    are same with these in SPN-LRWC, such as conversion range

    . And the parameters of PWCs in HSWC are same with these

    in SPN-PWC, such as coupler ratio . For example, when the

    number of wavelength converters , it is meaning

    that there are 4 PWCs and no LRWCs in HSWC, 4 PWCs in

    SPN-PWC, and 4 LRWC in SPN-LRWC. Actually, we com-pare the performance among the three architectures when their

    implementation costs are different. Therefore, when the same

    number of wavelength converters is used, the cost of HSWC

    is lower than that of SPN-LRWC, and it is higher than that of

    SPN-PWC.

    Figs. 6 and 7 show the differences of packet loss prob-

    ability in SPN-PWC, SPN-LRWC and HSWC when traffic

    load and , respectively. It is obvious that the

    packet loss probability cannot be further cut down in the three

    architectures when reaching certain level. This is because that

    conversion is limited or there are no idle wavelength channels

    for contending optical packets. Also, we can find from the

    comparison among the three architectures that HSWC can ob-

    tain the lowest packet loss probability value when wavelength

    Fig. 6. PLP of SPN-PWC, SPN-LRWC and HSWC when wavelength con-verters number is changed .

    Fig. 7. PLP of SPN-PWC, SPN-LRWC and HSWC when wavelength con-verters number is changed .

    converters are enough. Actually, the HSWC architecture can

    get lower packet loss probability than the SPN-LRWC with

    equivalent wavelength converters number. On the other hand,

    HSWC architecture can greatly reduce the cost for reaching cer-

    tain packet loss probability when compared with SPN-LRWC.

    Although the cost of SPN-PWC is lower than that of HSWCwhen the number of wavelength converters is same, the PLP of

    SPN-PWC is far lower than that of HSWC.

    Fig. 8 shows thepacket loss probability against the traffic load

    for SPN-PWC, SPN-LRWC and HSWC. We assume that there

    are sufficient wavelength converters ( is enough

    here) for gaining the lowest PLP value at the traffic load from

    to . From the results we can see that the PLP of

    three architectures would go up when increasing the traffic load.

    The reason is that the resources of output wavelength channels

    for contending optical packets become scarce when increasing

    the traffic load. Furthermore, we can find that the PLP of the

    HSWC is obviously lowest of the three at any traffic load.

    From Figs. 79, we can conclude that the performance of

    OPS node is obviously improved by the PWCs together with

  • 8/13/2019 06117054

    6/7

    LIUet al.: A NOVEL CONTENTION RESOLUTION SCHEME OF HYBRID SHARED WAVELENGTH CONVERSION FOR OPS 227

    Fig. 8. PLP of SPN-PWC, SPN-LRWC and HSWC when traffic load ischanged .

    Fig. 9. PLP of PFA and LFA algorithm in HSWC .

    LRWCs. The PWC can convert multiple wavelengths simulta-

    neously; so it can avoid the waste of LRWCs. Above all, a con-

    tending packet has a third better way to be converted by succes-

    sively using LRWCs and PWCs. This way can further reduce

    PLP because of using PWCs together LRWCs.

    Then we compare the performance of PFA and LFA sched-

    uling algorithm for the HSWC architecture in the Figs. 9 and 10.

    When increasing wavelength converters number, PFA and LFAalgorithm can be gradually close to the lowest packet loss prob-

    ability. Before reaching the lowest packet loss probability, we

    can obviously see the difference. The PFA algorithm can reach

    the lowest value earlier than LFA algorithm, and the PLP of

    PFA is always lower than that of LFA algorithm. That is to say,

    the PFA scheduling algorithm can maximize the utilization of

    wavelength converters. Therefore the PFA algorithm can take

    advantage of the benefit of PWCs that allow multiple wave-

    lengths to be simultaneously converted from one to another.

    We further study the influence of the LRWC conversion range

    on the HSWC architecture. As is shown in Fig. 11, the packet

    loss probability is a function of the LRWCs number when

    and . We can find that

    the lowest PLP of different conversion ranges are obtained by

    Fig. 10. PLP of PFA and LFA algorithm in HSWC .

    Fig. 11. PLP of HSWC when the conversion range of LRWC is changed.

    increasing conversion range. But we must note that the decre-

    ment is died out after because of no idle wavelength

    channels. And the decrements become smaller and smaller with

    the increasing conversion range. For example, the lowest PLP

    is , and corresponding to the

    conversion range , and , respectively. We can easily

    get that the decrements are and . However,

    the cost of LRWC is more expensive when increasing conver-

    sion degree. Therefore we should make compromise between

    the performance and the cost.

    V. CONCLUSION

    This paper presents an optical packet contention resolution

    scheme with hybrid shared wavelength conversion (HSWC) for

    all-optical packet switching nodes. And the parametric wave-

    length converterfirst available (PFA) algorithm is proposed for

    the HSWC architecture to achieve a maximum of wavelength

    utilization. The HSWC architecture can largely decrease the op-

    tical packet loss probability largely compared with SPN-PWC

    and SPN-LRWC, respectively. Therefore the HSWC architec-

    ture can improve the utilization of wavelength converters and

    reduce required wavelength converters number. It is helpful to

  • 8/13/2019 06117054

    7/7

    228 JOURNAL OF LIGHTWAVE TECHNOLOGY, VOL. 30, NO. 2, JANUARY 15, 2012

    design and implement the all-optical packet switching. More-

    over, we have to note that some factors would influence the PLP

    in HSWC, such as the ratio of LRWC number to PWC number,

    input port number of PWC, input fibers number and wavelength

    channels number.

    REFERENCES

    [1] N. Calabretta, W. Wang, T. Ditewig, O. Raz, F. G. Agis, S. Zhang, H.de Waardt, and H. J. S. Dorren, Scalable optical packet switche s formultiple data formats and data rates packets, IEEE Photon. Technol.

    Lett., vol. 17, no. 5, pp. 483485, Feb. 2010.[2] H. L. Liu and Q. B. Chen, Short packet first scheduling for vari-

    able length optical packet switch, J. Chongqing Univ. of Posts andTelecommun. (Natural Sci. Ed.), vol. 19, no. 1, pp. 7680, 2007.

    [3] D. Benhaddou and G. Chaudhry, Photonic switching techniques andarchitecture for next generation optical networks, Cluster Comput.,vol. 7, no. 3, pp. 281291, Jul. 2004.

    [4] J. C. Gao, A. P. Xiong, and S. Hu, A survivability-aware resourcereservation mechanisms for optical networks,J. Chongqing Univ. of

    Posts and Telecommun. (Na tural Sci. Ed.), vol. 21, no. 1, pp. 105109,

    Feb. 2009.[5] H. Buchta, C. M. Gauger, and E. Patzak, Maximum size andthroughput of SOA-based optical burst switching nodes with limitedtuning-range wavelength converters and FDL buffers, J. Lightw.Technol., vol. 26, no. 16, pp. 29191927, Aug. 2008.

    [6] A. Detti, G. Parca, V. Carrozzo, and S. Betti, Optical packet networkwith limited-range wavelength conversion: A novel formalization ofthe optimal scheduling problem,J. Lightw. Technol., vol. 27, no. 24,pp. 56075618, Dec. 2009.

    [7] X. D. Qin and Y. Y. Yang, Multicast connection capacity of WDMswitching networks with limited wavelength conversion, IEEE/ACMTrans. Netw., vol. 12, no. 3, pp. 526538, Jun. 2004.

    [8] T. Durhuus, B. Mikkelsen, C. Joergensen, S. L. Danielsen, and K. E.Stubkjaer, All-optical wavelength conversion by semiconductor op-tical amplifiers,J. Lightw. Technol., vol. 14, no. 6, pp. 942954, Jun.1996.

    [9] H.L. Liu and X.T. Bai,Performance studyof optical packetswitching

    node based on shared wavelength conversion, J. Opt. Commun., vol.30, no. 2, pp. 116119, May 2009.

    [10] S. J. B. Yoo, Wavelength conversion technologies for WDM networkapplications,J. Lightw. Technol., vol. 14, no. 6, pp. 955966, Jun.1996.

    [11] V. Eramo, M. Listanti, and M. Spaziani,Resources sharing in opticalpacket switches with limited-range wavelength converters,J. Lightw.Technol., vol. 23, no. 2, pp. 671687,Feb. 2005.

    [12] H. L. Li and I. L. J. Thng, Cost-saving two-layer wavelength conver-sion in optical switching network,J. Lightw. Technol., vol. 24, no. 2,pp. 705712, Feb. 2006.

    [13] V. Eramo, A. Germoni, C. Raffaelli,and M. Savi, Multifiber shared-per-wavelength all- optical switching: Architectures, control, and per-formance,J. Lightw. Technol., vol. 26, no. 5, pp. 537551, Mar. 2008.

    [14] V. Eramo, M. Listanti, and A. Germoni, Cost evaluation of opticalpacket switches equipped with limited-range and full-range convertersfor contention resolution, J. Lightw. Technol., vol. 26, no. 4, pp.390407, Feb. 2008.

    [15] C. Okonkwo, R. C. Almeida, R. E. Martin, and K. M. Guild, Per-formance analysis of an optical packet switch with shared parametricwavelength converters, IEEE Commun. Lett., vol. 12, no. 8, pp.596598, Aug. 2008.

    [16] N. Kitsuwan, R. Rojas-Cessa, M. Matsuura, and E. Oki, Performanceof optical packet switches basedon parametric wavelengthconverters,

    J. Op t. Commu n. Ne tw., vol. 2, no. 8, pp. 558569, Jul. 2010.[17] S. Watanabe, S. Takeda, and T. Chikama, Interband wavelength con-

    version of 320 Gb/s ( Gb/s) WDM signalusing a polarization-in-sensitive fiber four-wave mixer, in Proc. Eur. Conf. Opt. Commun.(ECOC), 1998, pp. 8586.

    [18] J. Yamawaku, H. Takara, T. Ohara, K. Sato, A. Takada, T. Morioka,O. Tadanaga, H. Miyazawa, and M. Asobe, Simultaneous 25GHz-spaced DWDM wavelength conversion of 1.03 Tbit/s (103 ch 10Gbit/s) signals in PPLN waveguide, Electron. Lett., vol. 39, no. 15,pp. 11441145, 2003.

    [19] J. Yamawaku, E. Yamazaki, A. Takada, T. Morioka, and K. Suzuki,Field demonstration of up to 640 Gb/s ( Gb/s) GWPswitching networks with a QPM-LN wavelength converter in JGNII test bed, IEEE J. Sel. Top. Quantum Electron., vol. 12, no. 4, pp.529535, Jul./Aug. 2006.

    [20] P. Devgan, R. Tang, V. Grigoryan, and P. Kumar, Highly efficientmultichannel wavelength conversion of DPSK signals, J. Lightw.Technol., vol. 24, no. 10, pp. 36773682, 2006.

    [21] J. Yu, M. Huang, and G. Chang, Polarization insensitive wavelengthconversion for Gbits/s polarization multiplexing RZ-QPSK sig-nals,Opt. Express, vol. 16, no. 26, pp. 2116121169, 2008.

    Huan-Lin Liu received the Ph.D. degree from Chongqing University,Chongqing, China, in 2008 and the M.S. degree from Chongqing University ofPosts and Telecommunications, Chongqing, China, in 2000.

    Her research interests include all-optical network research, all opticalswitch structure and scheduling algorithm research, information acquiring, andprocessing.

    Bin Zhangreceived the M.E. degree from the Chongqing University of Postsand Telecommunication, Chongqing, China, in 2011.

    His research focused on optical switching and scheduling algorithms. He iscurrently working toward the Ph.D. degree in the Beijing University of Postsand Telecommunications, Beijing, China.

    Song-Lei Shireceived the B.S. degree from the PLA Information EngineeringUniversity, Henan, China, in 2009. He is currently working toward the M.E. de-gree at the Chongqing University of Posts and Telecommunication, Chongqing,China.

    His research interests include all-optical wavelength conversions, opticalpacket switching systems, and optical multicast.