Elsevier

Computer Networks

Volume 41, Issue 1, 15 January 2003, Pages 41-55
Computer Networks

Algorithms for burst rescheduling in WDM optical burst switching networks

https://doi.org/10.1016/S1389-1286(02)00325-0Get rights and content

Abstract

Wavelength division multiplexing (WDM) is a promising technology for realizing terabit networks. Optical burst switching (OBS) is a way to efficiently support bursty traffic on WDM-based optical Internet networks. In OBS networks, the control (header) and payload (data) components of a burst are sent separately with a time gap. The control packet first traverses the burst switching nodes and reserves suitable wavelengths on the links for the corresponding data burst by using a scheduling algorithm. Our work is motivated from the observation that the existing scheduling algorithms either have low computational complexity or high performance in terms of burst dropping probability, but not both simultaneously. Since the arrival of bursts is dynamic, it is highly desirable that the scheduling is done as quickly as possible. We develop scheduling algorithms which integrate the merits of both low computational complexity and high burst dropping performance. The key idea is to reschedule an existing burst by assigning a new wavelength to it keeping the burst arrival and leaving time unchanged in order to accommodate the new burst. We propose computationally simple rescheduling algorithms called on-demand burst rescheduling and aggressive burst rescheduling. The effectiveness of the proposed algorithms and the signaling overhead are studied through simulation experiments.

Introduction

Wavelength division multiplexed (WDM) optical networks are a promising candidate for supporting the next generation backbone transport network traffic with hundreds of wavelengths, each operating at a different optical wavelength. The existing optical switching techniques can be broadly classified into optical circuit switching, optical packet switching, and optical burst switching techniques. Optical circuit switching networks provide circuit-switched lightpath services where lightpaths need to be established first from the source node to the destination node using a dedicated wavelength on each link along a physical path [1]. Such networks require a two-way reservation protocol to set up a circuit and they may not utilize the bandwidth efficiently to support bursty traffic. In optical packet switching networks, a fixed size packet is sent along with its header and the packet is optically buffered or delayed at the intermediate node while the header is being processed electronically [2], [3], [4]. Upon reaching a node, the header will be extracted and processed electronically. The optical packet switching technology is not yet mature mainly due to the technological limitations on optical packet synchronization, contention resolution, and buffering.

Optical burst switching (OBS), as described in [5], [6], [7] is able to achieve a balance between circuit and packet switching as it combines the advantages of both. There is no need for buffering and electronic processing of data in burst switching as in the case of optical circuit switching. At the same time, OBS ensures efficient bandwidth utilization on a fiber link just as optical packet switching by reserving bandwidth on a link only when data is actually required to be transferred through the link. It is therefore a promising solution towards the next-generation optical Internet with terabit optical routers and IP over WDM as the core architecture since it can more effectively exploit the capabilities of fiber optic transmission systems and could facilitate the transition of switching systems in which optical technology plays an important role [6].

In the literature, several burst switching techniques have been proposed, such as in-band-terminator (IBT), tell-and-go (TAG) [8], [9], and reserve-a-fixed-duration-based (RFD-based) protocol just-enough-time (JET) [10]. For burst switching, the basic switching entity is a burst. A burst is assembled at an ingress edge router by aggregating a number of IP packets which are destined to the same egress edge router. A burst has two parts called the control (header) and data (payload). They are referred to as control packet and data burst, respectively. In OBS, the control and data parts of a burst are transmitted separately with a time gap between the two parts, which is called the offset time. The offset time is a variable depending on the number of hops the burst has to traverse before reaching the destination node. This is done to ensure that the control part which carries the header information such as the source node address, destination node address, offset time, burst duration, and quality of service (QoS) requirements among others has already reserved resources for transmission of the data burst by the time it arrives at a node. By extending multi-protocol label switching (MPLS) capabilities to the OBS network, explicit routing can be used at the ingress nodes [11]. The burst offset time could also be adjusted to support QoS [12]. As they are sent over separate wavelengths through the OBS network, the control packet is processed electronically and the data burst is switched optically upon reaching every intermediate node before reaching the egress router.

OBS ensures efficient bandwidth utilization on a fiber link by reserving bandwidth on the link only when data is actually required to be transferred. If no wavelength is immediately available, the data burst is dropped. Upon successful reservation, the “RESERVE” control packet is forwarded to the next node along the selected outgoing fiber. When the data burst reaches the egress router, it is disassembled into a number of IP packets which are then transmitted to the appropriate access networks [13].

As arrival of bursts at a node is dynamic, scheduling which assigns an available wavelength to a burst for the entire duration of transmission in an efficient way is needed. If fiber delay lines (FDLs) are available, assignment of FDLs to a data burst is required when it could not be scheduled immediately. A few burst scheduling algorithms such as latest available unscheduled channel (LAUC) and latest available unused channel with void filling (LAUC-VF) have been proposed in the literature [6], [7]. While the LAUC algorithm is computationally simpler it performs poorer than the LAUC-VF algorithm. There may arise situations wherein the above algorithms fail to schedule a new data burst to some wavelength due to the non-availability of resources. Since scheduling is done well in advance before a data burst actually arrives, any changes to the allocated bursts are possible. The idea of burst rescheduling which assigns a scheduled data burst to other available wavelength to accommodate a new data burst is therefore a way to improve burst dropping performance.

The focus of this paper is to develop new burst scheduling algorithms which could realize effective optical burst switching for dynamically arriving requests with high performance close to that of LAUC-VF but with low computational complexity close to that of LAUC. Based on the idea of rescheduling, two algorithms namely, on-demand burst rescheduling (ODBR) and aggressive burst rescheduling (ABR) are developed. Simulation experiments show that the performance of the proposed algorithms which have low computational complexity outperform the existing LAUC algorithm in terms of burst dropping probability. The rest of the paper is organized as follows. In Section 2, related work on burst switching protocols and scheduling is discussed, followed by the details of the proposed burst rescheduling techniques in Section 3. Section 4 presents the results of performance study while Section 5 makes some concluding remarks.

Section snippets

Related work

There are several OBS protocols such as IBT, TAG [8], [9], and RFD-based protocol JET [10] available in the literature. These protocols could handle data bursts as short as a few kilobytes to several megabytes efficiently. Specifically, IBT and TAG are suitable for distributed control based on “open-ended” resource reservation. Open-ended resource reservation refers to reservation without the knowledge of the end time of reservation. The resources are reserved until the ‘release’ control signal

Burst rescheduling techniques

As both low computational complexity and low burst dropping probability do not coexist in the two existing scheduling algorithms namely LAUC and LAUC-VF, we propose new scheduling algorithms which combine the relative merits of both algorithms (high performance with low dropping and low complexity) with the motivation that a scheduled burst can be rescheduled to another available wavelength in order to accommodate a new request. This is possible as requests arrive dynamically and a control

Performance study

In this section, performance of the proposed ODBR and ABR algorithms is studied through simulation experiments. Twenty experiments are run for each algorithm and in each experiment, bursts on the order of 105 are generated to obtain accurate results with 95% confidence interval of ≈5% deviation from the reported mean value.

We compare the performance of these algorithms with that of LAUC and LAUC-VF. The performance metrics used are burst dropping probability, performance improvement, and

Conclusions

In this paper, we have proposed new rescheduling algorithms for supporting bursty traffic in optical networks with the objective of improving burst dropping performance keeping the computational complexity low. Since the rescheduling algorithm changes only the wavelength keeping the time unchanged, it does not pose any implementation problem. Also, it does not disrupt any traffic, because rescheduling takes place before a burst arrives. Simulation results have shown that the proposed ODBR and

Acknowledgements

This work was supported in part by the National University of Singapore Academic Research Fund under Grant R-263-000-173-112.

Siok Kheng Tan received her B.Eng. degree from Sheffield University, UK in 1999. She is currently working towards her M.Eng. degree in Electrical and Computer Engineering at National University of Singapore. She was an Engineer in Hewlett Packet Singapore and Micron Semiconductor Singapore during 1999–2000. Her current research interests focus on WDM optical networks.

References (13)

  • I Chlamtac et al.

    Lightpath communications: an approach to high bandwidth optical WANs

    IEEE Transactions on Communications

    (1992)
  • C Guillemot

    Transparent optical packet switching: the European ACTS KEOPS project approach

    IEEE/OSA Journal of Lightwave Technology

    (1998)
  • D.K Hunter

    WASPNET: a wavelength switched packet network

    IEEE Communications Magazine

    (1999)
  • D.K Hunter et al.

    Buffering in optical packet switches

    IEEE/OSA Journal of Lightwave Technology

    (1998)
  • C Qiao et al.

    Optical burst switching (OBS)––a new paradigm for an optical Internet

    Journal of the High Speed Networks

    (1997)
  • J.S Turner

    Terabit burst switching

    Journal of the High Speed Networks

    (1999)
There are more references available in the full text version of this article.

Cited by (17)

  • On burst rescheduling in OBS networks with partial wavelength conversion capability

    2010, Computer Networks
    Citation Excerpt :

    A majority of existing scheduling algorithms have been proposed for fully wavelength-convertible networks without WC-induced burst loss [2,8–10].

  • A performance comparison of OPS with variable-size packets and OBS

    2010, Optical Switching and Networking
    Citation Excerpt :

    In previous works, it has been shown that delayed reservation techniques have better channel utilization than immediate reservation techniques [22,23]. Upgrading channel utilization with on demand wavelength conversion and segmentation algorithms [24–27] is also possible. Although segmentation and on demand wavelength conversion mechanisms perform better than other reservation schemas, they increase computational or hardware complexity.

  • On ordered scheduling for optical burst switching

    2005, Computer Networks
    Citation Excerpt :

    This is illustrated by the scheduling scenario in Fig. 1(a). The burst rescheduling approach [10] is proposed to solve this problem. It still schedules a burst immediately when its header arrives at a node.

View all citing articles on Scopus

Siok Kheng Tan received her B.Eng. degree from Sheffield University, UK in 1999. She is currently working towards her M.Eng. degree in Electrical and Computer Engineering at National University of Singapore. She was an Engineer in Hewlett Packet Singapore and Micron Semiconductor Singapore during 1999–2000. Her current research interests focus on WDM optical networks.

G. Mohan received the Ph.D. degree in Computer Science and Engineering from the Indian Institute of Technology (IIT), Madras in 2000. From 1991 to 2000, he served as a Lecturer in the Department of Computer Science and Engineering, Regional Engineering College, Tiruchirappalli, India. From January 2000 to June 2000, he worked as a Senior Project Officer in the Department of Computer Science and Engineering, IIT, Madras. He joined the National University of Singapore in June 2000, where he is currently an Assistant Professor in the Department of Electrical and Computer Engineering. He has held a visiting position at Iowa State University, USA, during January–June 1999. His current research interests are in high speed multi-wavelength optical circuit and burst switching networks. He is the co-author of the textbook “WDM Optical Networks: Concepts, Design, and Algorithms” published by Prentice Hall PTR, NJ, USA in November 2001. He is a member of IEEE and SPIE technical group on optical networks.

Kee Chaing (KC) Chua received his Ph.D. degree in Electrical Engineering from the University of Auckland, New Zealand, in 1990. Following this, he joined the Department of Electrical Engineering at the National University of Singapore (NUS) as a Lecturer, became a Senior Lecturer in 1995 and an Associate Professor in 1999. From 1996 to 2000, he was seconded to be the Deputy Director of the Centre for Wireless Communications (now Institute for Communications Research), a national telecommunications R&D institute funded by the Singapore Agency for Science, Technology and Research. He is now on leave from NUS and works with Siemens Singapore where he heads the Mobile Core R&D department in the Information and Communications Mobile (ICM) group. He has carried out research in various areas of communication networks. His current interests are in quality of service provisioning in IP based networks. He is a member of the Institute of Electrical and Electronics Engineers, Inc. He is a recipient of an IEEE 3rd Millennium Medal for his contributions to professional activities.

View full text