WO2005022800A1 - Time-sliced optical burst switching - Google Patents

Time-sliced optical burst switching Download PDF

Info

Publication number
WO2005022800A1
WO2005022800A1 PCT/AU2004/001146 AU2004001146W WO2005022800A1 WO 2005022800 A1 WO2005022800 A1 WO 2005022800A1 AU 2004001146 W AU2004001146 W AU 2004001146W WO 2005022800 A1 WO2005022800 A1 WO 2005022800A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
burst
slices
means
time
Prior art date
Application number
PCT/AU2004/001146
Other languages
French (fr)
Other versions
WO2005022800A8 (en
Inventor
David James Moreland
Vijay Sivaraman.
Dean Economou
Diethelm Ironi Ostry
Original Assignee
Commonwealth Scientific And Industrial Research Organisation
Ip1 (Australia) Pty Ltd (Receivers And Managers Appointed) (Administrators Appointed)
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Priority to AU2003904705 priority Critical
Priority to AU2003904705A priority patent/AU2003904705A0/en
Application filed by Commonwealth Scientific And Industrial Research Organisation, Ip1 (Australia) Pty Ltd (Receivers And Managers Appointed) (Administrators Appointed) filed Critical Commonwealth Scientific And Industrial Research Organisation
Publication of WO2005022800A1 publication Critical patent/WO2005022800A1/en
Publication of WO2005022800A8 publication Critical patent/WO2005022800A8/en

Links

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L47/00Traffic regulation in packet switching networks
    • H04L47/10Flow control or congestion control
    • H04L47/21Flow control or congestion control using leaky bucket
    • H04L47/215Token bucket
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L47/00Traffic regulation in packet switching networks
    • H04L47/10Flow control or congestion control
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L47/00Traffic regulation in packet switching networks
    • H04L47/10Flow control or congestion control
    • H04L47/21Flow control or congestion control using leaky bucket
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04LTRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
    • H04L47/00Traffic regulation in packet switching networks
    • H04L47/10Flow control or congestion control
    • H04L47/22Traffic shaping
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04QSELECTING
    • H04Q11/00Selecting arrangements for multiplex systems
    • H04Q11/0001Selecting arrangements for multiplex systems using optical switching
    • H04Q11/0062Network aspects
    • H04Q11/0066Provisions for optical burst or packet networks
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04QSELECTING
    • H04Q11/00Selecting arrangements for multiplex systems
    • H04Q11/0001Selecting arrangements for multiplex systems using optical switching
    • H04Q11/0062Network aspects
    • H04Q11/0071Provisions for the electrical-optical layer interface
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04QSELECTING
    • H04Q11/00Selecting arrangements for multiplex systems
    • H04Q11/0001Selecting arrangements for multiplex systems using optical switching
    • H04Q11/0005Switch and router aspects
    • H04Q2011/0007Construction
    • H04Q2011/0033Construction using time division switching

Abstract

A method of transmitting traffic packets (101') over a data link in a time-slotted optical burst switching network is disclosed. The method comprises the steps aggregating the traffic packets (101') into a data burst (300), slicing the burst (300) into a plurality of data slices (318), establishing a spreading factor associated with the data burst (300) and allocating the data slices (318) to outgoing slots (309) on the all-optical data link according to the spreading factor.

Description

TIME-SLICED OPTICAL BURST SWITCHING Field of the Invention The present invention relates generally to optical communication networks, and

in particular, to networks based on optical burst switching. Background

In recent years the enormous bandwidth potential of fibre-optic networks has

been exploited through the emergence of Wavelength Division Multiplexing (WDM)

technology. Much work has been focused on solutions to the Lightpath Topology Design

(LTD) and Routing and Wavelength Assignment (RWA) problems. These issues relate to

optimal assignment of light-paths to optically routed WDM networks for the

establishment of end-to-end connections.

LTD/RWA solutions are particularly suited to off-line capacity planning and

scheduling for quasi-static networks that remain in a configured state for relatively long

periods of time (e.g. days, weeks or months). A problem with quasi-static optical

networks is that a wavelength's entire bandwidth is dedicated to a configured light -path

for as long as that light-path exists. This operating mode can be wasteful of network

capacity and potential revenue unless wavelengths are fully utilised. This potential

problem is exacerbated as bandwidth-per-wavelength, particularly in metropolitan and

core networks, continues to increase to lOGbit/s and above per wavelength. Recently there has been a shift in optical networking towards more dynamic

intelligent and statistically multiplexed optical infrastructures. This change requires fast changes in network configuration/provisioning, often within seconds, and more efficient utilisation of the available bandwidth per- WDM wavelength.

Optical Burst Switching (OBS), which operates in networks that are all-optical, at least as far as data connections, has been proposed to overcome some of the above

problems. However there remain significant technical hurdles that need to be overcome, not least of which is to find simple, effective and efficient strategies for resolving burst contention within switches. Burst contention occurs within a switch when more than one data burst competes for occupancy of the same switch output port at the same time on the same wavelength. Burst contention can increase the probability of bursts being dropped to above acceptable probability levels. Summary It is an object of the present invention to substantially overcome, or at least ameliorate, one or more disadvantages of existing arrangements. In the disclosed approach the objective is to reduce burst contentions by slicing and spreading data bursts, and varying the span of spreading using a synchronous fixed-length time-slotted method. This will be referred to as the "burst spreading technique". For a synchronous fixed-length time-slotted OBS network carrying variable sized traffic bursts, a method is disclosed for reducing network contention losses and switch complexity. The method involves OBS edge switch slicing of data bursts into fixed-length data slices and the regularly spaced allocation of these slices to the time-slots, where the span of these time-slots is varied to reduce, as a function of traffic and network attributes, the losses due to contention between different bursts in OBS core switches. According to a first aspect of the present invention there is provided a method of transmitting traffic packets over a data link in a time-slotted optical burst switching network, the method comprising the steps of: aggregating the traffic packets into a data burst; slicing the burst into a plurality of data slices; establishing a spreading factor associated with the data burst; and allocating the data slices to outgoing slots on the all-optical data link according to the spreading factor. According to another aspect of the present invention there is provided a method of receiving traffic packets that have been transmitted over a data link in a time-slotted

optical burst switching network, wherein the packets having been aggregated into data

bursts that have been sliced into data slices that have been allocated to outgoing slots on

the all-optical data link according to a spreading factor, the method comprising the steps

of: receiving, prior to receipt of the data slices, information comprising time-slot

positions in which data-slices will arrive, and the number of data-slices which constitute a

burst; and assembling the slices into corresponding traffic packets depending upon the received information.

According to another aspect of the present invention there is provided an

apparatus for transmitting traffic packets over a data link in a time-slotted optical burst

switching network, the apparatus comprising: means for aggregating the traffic packets into a data burst; means for slicing the burst into a plurality of data slices; means for establishing a spreading factor associated with the data burst; and means for allocating the data slices to outgoing slots on the all-optical data link

according to the spreading factor. According to another aspect of the present invention there is provided an

apparatus for receiving traffic packets that have been transmitted over a data link in a

time-slotted optical burst switching network, wherein the packets having been aggregated

into data bursts that have been sliced into data slices that have been allocated to outgoing slots on the all-optical data link according to a spreading factor, the apparatus comprising: means for receiving, prior to receipt of the data slices, information comprising time-slot positions in which data-slices will arrive, and the number of data-slices which constitute a burst; and means for assembling the slices into corresponding traffic packets depending upon the received information. According to another aspect of the present invention there is provided a method of communicating traffic packets over a data link in a time-slotted optical burst switching network, the method comprising, in regard to a transmitter, the steps of: aggregating the traffic packets into a data burst; slicing the burst into a plurality of data slices; establishing a spreading factor associated with the data burst; allocating the data slices to outgoing slots on the all-optical data link according to the spreading factor and transmitting the data slices over the network; and communicating, over the network, information comprising time-slot positions in which data-slices will arrive, and the number of data-slices which constitute a burst; the method further comprising, in regard to a receiver, the step of assembling the slices into corresponding traffic packets depending upon the received information. According to another aspect of the present invention there is provided a system for communicating traffic packets over a data link in a time-slotted optical burst switching network, the system comprising: a transmitter comprising: means for aggregating the traffic packets into a data burst; means for slicing the burst into a plurality of data slices; means for establishing a spreading factor associated with the data burst; means for allocating the data slices to outgoing slots on the all-optical

data link according to the spreading factor and transmitting the data slices over the

network; and means for communicating, over the network, information comprising

time-slot positions in which data-slices will arrive, and the number of data-slices which

constitute a burst; a receiver comprising; means for assembling the slices into corresponding traffic packets

depending upon the received information. According to another aspect of the present invention there is provided a

computer program product including a computer readable medium having recorded

thereon a computer program for implementing any one of the methods described above. Other aspects of the invention are also disclosed.

Brief Description of the Drawings Fig. 1 shows an Optical Burst Switching network in which the disclosed burst

spreading technique can be practiced;

Fig. 2 shows the source OBS switch from Fig. 1 in more detail; Fig. 3 illustrates how the burst spreading technique works; Fig. 4 shows the core OBS switch from Fig. 1 in more detail; Fig. 5 shows the destination OBS switch from Fig. 1 in more detail;

Fig. 6 shows the Optical Time Slot Interchange (OTSI) device from Fig. 4 in

more detail;

Fig. 7 depicts a leaky bucket approach that can be used to implement the burst spreader of Fig. 2; Fig. 8 shows an OBS network in regard to which a performance simulation of the

spreading technique is assessed; Fig. 9 shows a fragment of the network in Fig. 8 including representative loading parameters; Fig. 10 shows an example of packet loss probability vs spreading factor for the simulated network of Fig. 9; Fig. 11 shows packet delay vs spreading factor in the network of Fig. 9; Fig. 12 shows packet loss vs delay in the network of Fig. 9; Fig. 13 shows how bursts are assembled by the burst assembler of Fig. 2; Fig. 14 shows how burst slice spreading is determined by the burst spreader of Fig. 2; and Fig. 15 shows how burst slices are scheduled by the scheduler of Fig. 2 to effect the desired spreading. Appendix A presents definitions for abbreviations used in the description; Detailed Description including Best Mode It is to be noted that the discussions contained in the "Background" section and that above relating to prior art arrangements relate to discussions of documents or devices that form public knowledge through their respective publication and/or use. This is not a representation by the present inventor(s) or patent applicant that such documents or devices in any way form part of the common general knowledge in the art. Where reference is made in any one or more of the accompanying drawings to steps and/or features, which have the same reference numerals, those steps and/or features have for the purposes of this description the same function(s) or operation(s), unless the contrary intention appears. Fig. 1 shows a synchronous Optical Burst Switching network 100 in which the disclosed burst spreading technique can be practiced. Incoming traffic on an ingress port 101 of a source edge switch 102 is switched to an egress port 112 on a destination edge switch 109 over a data connection that traverses the source edge switch 102, an optical fibre 114, a core switch 113, an optical fibre 115 and the destination edge switch 109. In the network 100, data connections traverse a data link 106, 108 and control information traverses a control link 103, 110. Although Fig. 1 shows a single core switch 113 interposed between source and destination edge switches 102 and 109 respectively, in the general case a number of core switches in tandem will be interposed between any source edge switch and any corresponding destination edge switch. How traffic is carried Data traffic on the network 100 is carried in data bursts. Data traffic enters the network 100 in packets or frames (also referred to as Protocol Data Units or PDUs) that are assembled into data bursts. This is described further in regard to Fig. 13. Data bursts are then cut into data burst slices which are subsequently "spread" as will be described in regard to Fig. 3. Data is transported from the source switch 102 to the destination edge switch 109 over the network 100 using fixed-length time-slots. A burst data-slice is the amount of data that can be carried in a time-slot end-to-end across the network 100. The network 100 uses all-optical end-to-end data links, and uses a switched transmission system that is a hybrid of circuit and packet switching schemes. A data burst traversing the network 100 has access to the entire bandwidth of the wavelength assigned to a particular data link for the (very short) duration it takes to transmit the data burst over the designated end-to-end connection. Composition of data bursts The (electronic) ingress ports 209 of the source edge switch 102 typically receive Protocol Data Units (PDUs) for different packet protocols including Internet Protocol (IP), Asynchronous Transfer Mode (ATM), and Ethernet. PDUs are characterised by parameters including destination address and traffic descriptors. Generally each traffic protocol type will be assigned to a distinct electrical interface port (such as 101). A packet at such a port can however be routed to one or more than one destination. The routing functionality of the system 100 depends on the routing capability of the edge switch routing modules such as the module 200 in the source edge switch 102 of Fig. 2. Thus for example if the routing module 200 supports a multi-casting capability, then multi-casting can be supported by the system 100. This description presumes that comprehensive routing capability is provided by the routing module 200 and the other equivalent modules in the system 100. The source edge switch 102 is described in more detail in regard to Fig. 2. Data links The data links in the network 100 are pre-defined, on a quasi-static basis, between each designated source edge switch and each designated destination switch. The data links can be re-defined from time to time. A data link in this arrangement comprises a series of concatenated data link segments (such as 106) each spanning or transiting a pair of segment end-points. Accordingly the data link segment 106 uses a pre-defined optical wavelength on the optical fibre 114 that spans the segment end-point switches 102 and 1 13. Data connections traversing the optical switching fabric 107 of each switch 113 are time-slot switched from input to output in accordance with the time-slot scheduling that is effected by the electronic controller 105. Control links The control link in the described arrangement is an un-slotted (ie. asynchronous) one-way link. An exemplary control link in Fig. 1 comprises the control link segments 103 and 110. Between switch pairs (eg the switch pair 102/113 and the switch pair 113/109) the control link uses an optical medium comprising a specific WDM wavelength over a specific optical fibre. However, control link signals are converted to an electronic format in the controller 105 of the core switch 113 for information processing. The control link transports control information about corresponding bursts of data that are subsequently transmitted on the data link. Control links are predefined in the same manner as data links, however the pre-scheduling of time slot opportunities does not apply to control links which operate in an asynchronous manner. The term "pre- scheduling of time-slot opportunities" means that time-slots are pre-scheduled on the basis of available time-slot opportunities. This is explained in regard to the data link. Arrangement of control and data links In the described arrangement a control link uses one wavelength in any particular fibre segment and a data link uses a distinct wavelength in the same fibre segment. Alternately, the control and data links can be provided on different optical fibre segments. Furthermore, the control link can be implemented over an electrical rather than an optical transmission medium.

Transmission and Switching of an exemplary data burst - the source switch The source switch 102 aggregates from the port 101 a number of PDUs having the same parameters and assembles the PDUs into a data burst of information ready to be transmitted over a single wavelength which constitutes an optical data link to the designated destination. A burst is made up of PDUs having the "same parameters", which means for example that the PDUs share a common destination and have traffic descriptors that belong to a common Quality of Service (QoS) class. The source edge switch 102 assembles bursts of data, cuts up each data burst into a number of data slices, and inserts these data slices into specific time-slots for transmission across the network 100. All time-slot boundaries throughout the network 100 are synchronised to a common clock reference (not shown). A time-slot is the smallest (Fixed-length) unit of time in the data plane of the synchronous network 100. Control information is transmitted asynchronously. Transmission and Switching of an exemplary data burst - Pre-scheduling time-slot opportunities Before the source edge switch 102 can send an assembled data burst to the destination edge switch 109, time-slot opportunities must be pre-scheduled (ie., reserved) for the data burst along a data link that traverses the network 100. The term "pre- scheduling of time-slot opportunities" means that time-slots on the data link are pre- scheduled on the basis of available time-slot opportunities on the data link. In order to perform the time-slot pre-scheduling, the source edge switch 102 constructs a Burst Control Cell (BCC) that contains information about the assembled data burst. This control information includes the burst destination address, the time-slots used by burst slices, and the data-link wavelength to be used. The BCC can also contain information on the specific Quality of Service Class of the assembled data burst. The burst control cell is thus a "packet" that contains information about a corresponding data burst. The burst control cell is constructed by the source edge switch 102 and is used by subsequent switches (113 in Fig. 1) to pre-schedule the time-slot opportunities for the associated data burst. The source edge switch 102 temporarily stores the assembled data burst in a queue 204 (see Fig. 2) and forwards the burst control cell over a control link comprising the control link segment 103 to the core switch 113. The core switch 113 pre-schedules time-slot opportunities for switching a data burst to the data link segment 108. Transmission and Switching of an exemplary data burst - the core switch The core switch 1 13 consists of two main elements, the electronic controller 105 which performs optical-to-electronic and electronic-to-optical conversion of information, and an all-optical switch fabric 107. On receipt of the burst control cell the electronic controller 105 examines the information contained in the burst control cell that relates to the corresponding data burst. This information is used to pre-schedule time-slot opportunities through the all-optical core switch fabric 107 for the expected data burst. The core switch 113 thus pre-schedules switch resources, comprising switch output ports, fibre and wavelengths, in readiness to "cut-through" the subsequent data burst. The core edge switch 113 is described in more detail in regard to Fig. 4. The core switch 113 modifies the burst control cell information, and then forwards the control cell to the next switch toward the destination. In the described network 100 of Fig. 1, this "next switch" is the destination switch 109. In the general case however a number of core switches in tandem will be interposed between any source edge switch and the corresponding destination edge switch. The aforementioned process of modifying and forwarding the BCC is thus generally repeated by all core switches along the end-to-end data link, until all these core switches have been pre-scheduled. When the BCC reaches the destination edge switch 109, the switch 109 schedules its resources in accordance with the (modified) burst control cell information to prepare to re-assemble the burst slices into bursts. Segmentation-and-reassembly of the traffic packets are implemented in the edge switches 102 and 109 using electronic technology. Unidirectional nature of the data link The end-to-end data link is unidirectional, and no acknowledgment is returned to the source edge switch 102 indicating successful transmission of a data burst. When a data burst has been transmitted, the end-to-end network resource (i.e. the time-slots used for that data burst) become available for use by other data bursts. Examples of key parameters in the network 100 are shown in the following table with typical values:

Figure imgf000013_0001
Figure imgf000014_0001
Table 1.

Unidirectional nature of the control link The end-to-end control link is unidirectional, and no acknowledgment is returned on the control link to the source edge switch 102 indicating successful pre-scheduling of time-slot opportunities for data bursts. Although a particular network arrangement, utilising pre-defined data links each using a common end-to-end optical wavelength, forms the basis for describing the system 100, other arrangements which make use of wavelength conversion, tunable optical sources (ie., lasers), and other routing mechanisms can also be used with the disclosed burst spreading technique. Thus, for example, the optical burst switching architecture 100 can also make use of switching of burst slices to other wavelengths should wavelength conversion be sufficiently attractive from a commercial perspective. Fig. 2 shows the source switch 102 from Fig. 1 in more detail. The switch operates in the electrical domain between the input ports 209 and the schedulers 207,..., 216 and in the optical domain between the schedulers 207,..., 216 and the output ports 103, 106 ... 217, 218. Signal paths from the input ports 209 to the output ports 106, 217 depict data connections through the source edge switch 102. Signal paths from the schedulers 207,..., 216 to the output ports 103, 218 depict control connections through the source edge switch 102. The source edge switch 102 receives traffic from multiple asynchronous sources at the ports 101 and 111. The network arrangement 100 is capable of supporting synchronous inputs (eg. time division multiplex traffic) within the time-slotted optical burst switching structure. Asynchronous traffic generally comprises variable length packets or PDUs and is directed from the input port 101 to the route look-up module 200. In the described arrangement the source edge switch 102 has extensive routing capabilities that make use of pre-defined data and control links as described in relation to Fig. 1. The look-up information, which in one arrangement can be stored in the routing module 200, can be updated from time to time either by static configuration operations, or dynamically by routing protocols. For each incoming packet to the switch 102 a route look-up is performed to determine (a) which destination edge switch is the target destination, and consequently (b) which egress wavelength to assign to the data slice associated with the packet when switching the data slice from the switch 102 into the network 100. The switch 102 in Fig. 2 has "n" arrival queues 202, ..., 210 associated with the scheduler 207, and "n" arrival queues such as 212 associated with the scheduler 216. The scheduler 207 is associated with the wavelengths used for the data link segment 106 and the control link segment 103. The scheduler 216 is associated with wavelengths used for the data link segment 217 and the control link segment 218. The queues 202 through 210 relate to "n" traffic streams directed to n different destination edge switches such as 109, 125, 126 and 127 in Fig. 1. Each input packet at 101 is held, according to the information in the routing module 200, in an appropriate FIFO arrival queue such as 202. When the packet is in the queue 202, the packet is referred to as being in a queue designated as "q". Packets in the packet arrival queue 202 are output into a burst assembler 203 that assembles packets held in the arrival queue 202 into bursts. This is described further in regard to Fig. 13. The assembly of the packets into bursts uses, for instance, a combination of timer-based and threshold-based schemes. According to such an arrangement the burst assembler 203 deems a burst to be assembled when either (a) the length of the queued packets in the queue 202 is greater than or equal to Lπg bytes (where Ltng is a threshold packet length in bytes), or (b) the queue 202 has been non-empty for T^g milliseconds (where Ttrjg is a threshold time in milliseconds). Typical values of Ltrig and Ttr;g are given in Table 1. Assembled bursts are output from the burst assembler 203 and held in a FIFO burst queue 204. Information on the bursts held in the burst queue 204 is output, as depicted by a dashed arrow 208, to a burst spreader 205. The function of the burst spreader 205 is twofold. The spreader 205 first determines how to chop an incoming burst from the burst queue 204 into data-slices (where each data-slice fills a time-slot on the optical data link 106). The spreader 205 then determines how to "spread" the data-slices in

accordance with a spreading-factor φq (where 0 < φq < l-pq and q denotes the average

data arrival rate to the arrival queue 202). The burst spreader 205 outputs information on how to slice the bursts into slices, and how to spread the slices, to the scheduler as depicted by a dashed arrow 206. The spreading process is thus performed in two stages. Firstly the burst spreader 205 determined how data slices should be spread without reference to occupancy of time slots on the outgoing data link 106. This calculation is subject to modification by the scheduler which takes actual slot occupancy on the link 106 into account. This composite process effectively shapes the packets held in the input

burst queue 204 onto the outgoing data link 106. The packets held in the burst queue 204

are output, as depicted by an arrow 221, to the scheduler 207 which schedules the associated burst data slices onto the output data link 106 in a spread manner as

determined by both the burst spreader 205 and the scheduler 207. Data slice spreading is

implemented, for example, by shaping the data slices in the queue 204 at a rate equal to a

fraction l-φq of the channel capacity of the data link 106. As noted, the burst spreader

205 does not actually perform spreading of the burst slices, but performs a calculation that

determines the eligibility times (ie te ) for each data slice. This is described further in

regard to Fig. 14. The eligibility times are used by the scheduler 207 in order to allocate

time slot opportunities in a scheduler table, thereby implementing the spreading

calculated by the burst spreader. This is described further in regard to Fig. 15.

For φq=0, the queue 204 has access to the entire channel bandwidth on the data

link 106, and the data-slices belonging to a burst will be eligible for transmission onto

contiguous time-slots across the network 100. As φq approaches l- q, the burst slices will

be spread out to occupy time-slots that are spaced well apart. This data stream

consequently has an increased delay in the burst queue 204 due to the reduced shaping

rate.

For all burst queues 204, .. 214, typically φq = φ which means that the spreading

factor is the same for all burst queues in the network. There are however circumstances

in which different spreading-factors are used for different burst queues. Such

circumstances can arise when providing QoS differentiation or for other traffic

engineering reasons. The spreading in the burst spreader 205 can be achieved, for example, by means of a leaky bucket shaper as described in regard to Fig. 7. The packets held in the burst queues 204,..., 220 are input into the scheduler 207

as depicted, for example, by the arrow 221. The scheduler 207 maintains the table as described above, and vacant table locations represent time slot opportunities on the data link 106 to which data burst slices can be allocated. The "pre-scheduling" process allocates data slices from the burst queues 204,..., 220 to slot locations in the scheduler table, thereby allocating slots in accordance with the available opportunities. The scheduler maintains this table continuously, and sends allocated slices out over the appropriate data link according to the slot allocations in the table. There is a single scheduler per data wavelength per data link segment, and each scheduler stores in the internal table the data slices for a particular data burst that are to be transmitted on the data wavelength. The location of each data slice in the scheduler table corresponds to the timeslot in which the slice is to be transmitted onto the respective data link. In one arrangement, a circular buffer implementation can be used for the table in the scheduler 207. In this case a data slice scheduled for transmission in timeslot i is stored in location (i mod K), where K denotes the size of the circular buffer. In regard to each timeslot y the scheduler 207 transmits the data slice stored in location (j mod K), if one is present, otherwise the slot goes idle meaning that no data is transmitted in that time-slot. The scheduler 207 maintains this table and sends slices out in accordance with the scheduling information in the table. It is apparent that the scheduler table can either store the actual data slices, or can store references to memory locations at which the data slices are stored. For each data burst in the queue 204, the burst spreader 205 determines how to slice the burst into data slices. The burst spreader 205 receives information on each burst slice as depicted by a dashed arrow 208, according to one arrangement, computes how these slices would be processed through the leaky-bucket shaper as described in regard to Fig. 7. The burst spreader shaper determines the eligibility time te of each slice this being the earliest slot at which the burst slice can be transmitted on the data wavelength. The eligibility times of the burst slices are passed as depicted by a dashed arrow 206 by the burst spreader 205 to the scheduler 207. The burst slices themselves are passed as depicted by an arrow 221, to the scheduler 207. The computation for an entire burst is completed by the burst spreader 205 before the eligibility times for the burst data slices are passed from the burst spreader 205 to the scheduler 207. The scheduler 207 determines for each slice, from the scheduler table, the first idle slot on the outgoing data link 106 that is no earlier than the eligibility time of the slice. Each slice is scheduled for a corresponding time slot by inserting the slice at the location corresponding to the associated slot in the scheduler table. Once all the slices corresponding to the burst have been scheduled, the scheduler writes into the burst control cell the list of slots into which the burst slices have been scheduled. This is shown in more detail in regard to Fig. 13. The scheduler 207 loads the following burst data-slice information into the corresponding burst control cell.

• The time-slot positions into which the burst data-slices are to be inserted;

• The number of data-slices which constitute a burst i.e. the burst length;

• The fibre (eg 114 in Fig. 1) on which the data-slices are to be transmitted; • The wavelength (corresponding to the data path 106) on which the data- slices are to be transmitted on that fibre; and

• The network addresses of the source and destination edge switches. The burst control cell is sent in advance of the burst data-slices, where the time difference between sending the burst control cell from the source switch 102 and the transmission of the corresponding burst data-slices into the network 100 is defined as the offset time, T0tτSet (see Table 1). The burst data-slices and burst control cell are sent on separate wavelengths, 106 and 103 respectively. In summary, burst slices of a burst in the burst queue 204 are directed in accordance with the solid arrow 221 to the scheduler 207 in accordance with control information developed by the burst spreader 205. Information about the bursts in the burst queue 204 is directed in accordance with the dashed arrow 208 to the burst spreader 205. The burst spreader 205 determines eligibility times for each burst slice of a burst in the burst queue 204 according to a process 1400 that is described in relation to Fig. 14. Once the eligibility times for each slice in a particular burst in the queue 204 are calculated by the burst spreader 205, the table of eligibility times, that is maintained by the burst spreader 205 is past as depicted by the dashed arrow 206 to the scheduler 207. The scheduler 207 then uses a process 1500 that is described in relation to Fig. 15 in order to schedule the burst slices onto the data link 106. Fig. 3 illustrates how the burst spreading technique works. A first stream 101' of packets (an exemplary packet being depicted by 314) and a second stream 111 ' of packets represent asynchronous traffic stream fragments being input into the respective switch ports 101 and 111 in Fig. 1. The queue 202 is the arrival packet queue in Fig. 2. Packets in the arrival queue 202 are assembled into bursts held in the data burst queue 204. The burst queue 204 comprises data bursts 300 and 301, each of which comprise 6 packets in this example. In the described example, and using a first spreading factor φi, the first two packets (ie 318) of the data burst 300 are inserted, as depicted by an arrow 303, into a time slot position 309 in a scheduler table 207' (which is associated with the scheduler 207 in Fig. 2). The second two packets (ie 324) of the data burst 300 are inserted, as depicted by an arrow 305, into a time slot position 310 in the scheduler table 207'. The third two packets (ie 325) of the data burst 300 are inserted, as depicted by an arrow 306, into a time slot position 311 in the scheduler table 207'. Alternately, if a different (larger) spreading factor φι is used, the first two packets (ie 318) of the data burst 300 are inserted, as depicted by an arrow 302, into a time slot position 308, and the third two packets (ie 325) of the data burst 300 are inserted, as depicted by an arrow 307, into a time slot position 312. The packets in the packet stream 101' are internally labelled al-alO, and the packets in the stream 111 ' are internally labelled bl-b8. The time slot positions 321, 309,

322, 310, 323 and 311 have been internally labelled to depict the relative locations of the incoming packets in the outgoing time slots of the data path in the event that the two packet streams 101 ' and 111 ' are being sent to the same destination. Fig. 4 shows the core switch 113 from Fig. 1 in more detail. The core switch 113 contains Optical Time Slot Interchangers (OTSIs) the function of which is to interchange time-slots when burst slice contention (which is one form of traffic congestion) occurs within the switch 113. The OTSIs thus perform time domain data slice re-arrangement on a given link segment. The OTSI is described in more detail in regard to Fig. 6. An OTSI operates in the time domain and utilises only optical delay lines to buffer data when contentions arise in a switch. The use of expensive wavelength converters to resolve contentions can thus be avoided. Contention occurs within a switch such as 113 when more than one data burst slice (ie. occupied time slot) competes for occupancy of the same output time slot on the same output wavelength. The input fibres 114 and 119 can carry multiple separate data and control wavelengths as seen in Fig. 1. Similarly the output fibres 115 and 122 can carry multiple separate data and control wavelengths as seen in Fig. 1. Each incoming fibre such as 114 terminates on a synchronisation unit 400 the function of which is to synchronise incoming data signals to a local timing reference (not shown) that is slaved to the common timing reference of the system. This does not apply to the control signals, which are asynchronous. Each Sync unit 400 is followed by an associated OTSI 402, which is described in more detail in regard to Fig. 6. The OTSI 402 provides the required space and time domain switching of time-slots for each data wavelength in the associated fibre 114. The size of the time-slot used in the network 100 depends on the switching speed of the cross- connects in the optical crossbar(s) 107 (see Fig. 1). A small time-slot is typically desirable since it allows more fine-grained usage of bandwidth by the data bursts, and hence a larger potential for statistical sharing of the bandwidth on the WDM links 106, 108. However, the smaller the time-slots used, the faster the re-configuration of the optical crossbar switch fabric must be. Micro-Electro-Mechanical Systems (MEMS), Acousto-Optic and Thermo-Optic optical switching devices generally have relatively long re-configuration times typically of the order of tens of milliseconds. Chiaro™ Networks has however demonstrated a solid-state optical cross-connect with a switch re-configuration time of about 20ns. 64x64 non-blocking cross-connects can be realised using current technology, with larger devices to be expected as technology evolves. Using the Chiaro™ switch, or equivalent, as the optical crossbar in the OTSI (eg 608 in Fig. 6) and main switch (eg 413 in Fig. 4), a time- slot size of 500ns has been selected for the described arrangement. A guard-band of about 50ns at the start of each time-slot is reserved for optical cross-connect reconfiguration.

A data slice in a time slot on a wavelength λ (ie 403) which enters an incoming

port 414 of a core switch crossbar 413 can be switched to another time slot on another

instance of the wavelength λw (ie 416) which exits an outgoing port 417 of the core

switch crossbar 413. Under control 412 of the switch controller 105, WDM multiplexers 406 multiplex the data and control wavelengths onto the output fibre 115 that feeds into the next downstream OBS switch 109. The core switch intelligence resides in the switch controller 105 and is comprised of: • Scheduling algorithms for OTSI time-switching of time-slot contents to reduce the burst slice contention in the core switch optical crossbars such as 413; and

• Scheduler algorithms for space switching of input ports to output ports in the OTSI and main switch crossbars (such as 608 and 413 respectively; also see

Fig. 6). The switch controller 105 has scheduling functionality that makes use of the

information contained in each received BCC. This information informs the scheduler (not

shown) in the controller 105 of the time-slot numbers that the slices of a particular burst

occupy and the number of time-slots used by the burst. Using this information, from the

many BCCs received across all incoming fibres such as 114 and 119, the controller 105

schedules the time-slot usage in the OTSIs such as 402, so as to minimise contentions,

and schedules the space switching of the main switch crossbars (eg 413) and OTSI

crossbars (eg 608). After scheduling of time-slots in the core switch 113, the associated

controller 105 loads information into outgoing BCCs (thereby modifying the incoming

BCC information) indicating which time-slots the slices of a burst now occupy. These

(modified) BCCs are sent to the next downstream OBS switch 109 in Fig. 1.

Fig. 5 shows the destination OBS switch 109 from Fig. 1 in more detail. The

data connection through the destination edge switch 109 is described by reference

numerals 108 and 1 12. The control link that terminates in the switch 109 is depicted by

the reference numeral 110. The destination edge switch 109 receives burst control cells

on the control link 110. Each burst control cell contains the following information about its corresponding data burst:

• The time-slot positions in which data-slices will arrive; o The number of data-slices which constitute a burst i.e. the burst length; • The fibre on which the data-slices are received;

• The wavelength on which the data-slices are received; and

• The addresses of the source and destination OBS edge switches. When a de-scheduler 500 receives a burst control cell it uses the BCC information to reassemble the corresponding data burst by steering incoming burst data slices, arriving in time-slots on the data link 108, to an appropriate per-source queue 502, ..., 505. Re-ordering of burst data slices is not required since the data slices arrive in an orderly sequence. Although the described arrangement ensures that the ordered sequence of data slices that is initially established in the burst queue 204 is maintained as the data slice traverses the network 100, other ordering approaches can be used. The operation of the de-scheduler 500 is the inverse to that of the scheduler 207 described in relation to Fig. 2. Data slices originating from the source edge switch 102 are held in a corresponding arrival queue 505. Data slices originating from other source edge switches 116, 117, 118 are held in their own corresponding arrival queue such as 502. The queued data slices are then passed to corresponding per-source packet re- assemblers 506, ..., 504. The packet re-assembler modules 506, ..., 504 continually read data from their associated source buffers 505, ..., 502. The re-assembler modules 506, ..., 504 reassemble the data into the format (i.e. packets, cells or frames) in which the data arrived at the ingress side 209 of the source edge switch 102. As soon as a "packet" of information has been reassembled by one of the re-assembler modules 506, ..., 504, the packet can be passed to an appropriate upper layer protocol module (not shown) via an output interface 507. By operating in this mode the destination edge switch 109 need not know if an entire burst has arrived before passing on the "packets", contained in the burst (e.g. Gigabit Ethernet frames), to the higher layers, however the edge switch will only pass entire reassembled "packets" to the upper protocol layers module. Fig. 6 shows the Optical Time Slot Interchange (OTSI) device 402 from Fig. 4 in more detail. In the described arrangement a non-blocking architecture is used for the sake of simplicity of description. However, blocking OTSI architectures can also be used. The OTSI 402 separates out the WDM input 401 (which is carried by the fibre 114 in Fig. 4) into control and data wavelengths using a demultiplexer 606. Burst control cells, which are carried by control wavelengths, are forwarded via 408 to the switch controller 105 (see also Fig. 4). The separated data wavelengths such as 607 are forwarded on independent fibres each of which terminates on an optical crossbar such as 608. The optical crossbar 608 performs the required space division switching operation from the input port 607 to output ports such as 403. As seen from 607 and 609,

the cross bar 608 handles space switching of the λw wavelength. Accordingly, the cross

bar 608 can space-switch the input at 607 to an output at 403 or 609, noting that all inputs

and outputs of the cross bar 608 handle only the λw wavelength.

However, before any space switching of time-slots takes place, prior scheduling of the OTSI crossbars such as 608 is performed by switch controller control signals 409

(see also Fig. 4). On receipt of these control signals 409 the OTSI 402 schedules its crossbars in conjunction with delay lines 601-603 in the following way (considering the single crossbar 608). If there is no contention for a free time-slot on the crossbar output port 403 then data will be space switched through the crossbar 608, without any delay, to the central crossbar connection 403 that terminates on the core switch crossbar 413 (see Fig. 4). If, however, there is contention, then the following process applies: If incoming occupied time-slots to the crossbar 608 are in contention for a free time-slot on the crossbars output port 403 then the delay lines 601-603 having appropriate delays are selected in order to circulate the contents of all but one of the contending time- slots back into the input port 607 of the crossbar. The circulated time-slot(s) contents on re-entering the crossbar 608 will again vie for a time-slot on the output port 403. By this approach, the OTSI 402 time-switches the time-slot contents to avoid contentions. If a recirculated time-slot is again in contention, then the time-slot contents will be dropped by the OTSI 402. In this example, therefore, a given time-slot is allowed to be switched through at most one delay line, which means that it will undergo at most two switching operations in the OTSI. The "time- switching" performed by the OTSI maintains the order of the burst data slices in the order originally established in the burst queue 204 (see Fig. 2). However other arrangements in which the order is not maintained can also be used, noting that such arrangements would need to track ordering changes and thus would be more be more complex that the arrangement described. It is particularly important from a network performance perspective to minimise the number of re-circulations of data to keep signal degradation at a manageable level. In the OTSI 402 (see also Fig. 6) fibre-length scales linearly with D, where D is the maximum delay supported by the OTSI 402. Other architectures where for example fibre

length scales as VD , can also be used. Time slot information in an incoming BCC is time-switched by the OTSIs in the core switch as required on a per-burst-slice basis. This time-switching involves changing the slot position of a burst slice which is in contention, while maintaining the overall order of the slices for a particular burst in question. The table maintained in the BCC which reflects the time slots in which each particular burst slice are stored is updated according to this time-switching process. The OTSI performs time-slot interchange on a per burst-slice basis, this being reflected in the updating of the corresponding BCC. The core switch crossbar performs space switching which does not result in time-slot changes, and accordingly no BCC updating results therefrom. Fig. 7 shows an example of how the burst spreader 205 of Fig. 2 can be implemented using a leaky bucket approach. The leaky-bucket shaper process in the burst spreader 205 can be envisaged as a bucket 706 that is constantly being filled with tokens 701 at a rate of R tokens/second. The bucket depth 703 is D tokens, and any tokens arriving into a full bucket are lost. A data slice at 208 passing through the shaper 205 via the buffer 700 has to collect a token from the bucket 706. If a token is not available, the data has to wait in the buffer 700 until a token becomes available. In the described arrangement the bucket depth D is set to 0 tokens, the bucket occupancy is allowed to go negative, and a data slice can pass through if and only if the bucket occupancy is non- negative. This approach produces a more regularly spaced sequence of data slices as its output than a regular leaky-bucket. A "uniform" or "regularly spread" sequence describes a finite number of burst data-slices spaced out as evenly as possible on a transmission link. Thus, for example, if M data slices are allocated to N free consecutive time-slots (where N>M) on a transmission link then the gaps (i.e. number of free time-slots) between those time-slots carrying a data-slice typically differ by at most one time-slot. As noted, the burst spreader 205 performs the shaping calculation based, in the present example, upon the leaky bucket approach. Other burst spreading approaches can also be used. Fig. 8 shows an OBS network 800 in regard to which a performance simulation of the spreading technique is performed. User data terminals 801-803 are connected to an edge switch 902. User data terminals 804-805 are connected to an edge switch 910. The edge switches 902,..., 910 are connected by respective fibres 807,..., 809 to a core switch 915. The core switch 915 is connected by a fibre 812 to an edge switch 908. The core switch 915 is also connected by respective WDM fibres 811, 810 to core switches 815 and 819. The core switch 815 is connected by respective WDM fibres 814, 816 to an edge switch 813 and a core switch 817. The core switch 819 is connected by respective WDM fibres 833, 818 to an edge switch 820 and the core switch 817. The edge switch 820 is connected to user terminals 821-822. The core switch 817 is connected by respective WDM fibres 826-827 to edge switches 825, 828. The edge switch 828 is connected to user terminals 831-832. The edge switch 825 is connected to user terminals 829-830. The edge switch 813 is connected to user terminals 823-824. Fig. 9 shows a representative fragment of the network in Fig. 8 including representative loading parameters. The core switch 915 has 8 input links 911-912 on the fibres 807,..., 809, each connected to a respective source OBS edge switch 902-910. The core switch 915 has one output link 914 on the fibre 812 connected to the destination OBS edge switch 908. The source edge switch 902 terminates asynchronous electrical interfaces 916. The source edge switch 910 terminates asynchronous electrical interfaces 917. The destination edge switch 908 terminates asynchronous electrical interfaces 918. Only a single data wavelength is considered for the OBS network 900. Respective control links 903-904 connect the source switches 902, 910 to the core switch 915. A control link 906 connects the core switch 915 to the destination switch 908. An electrical interface 907 is operative between an electronic controller 905 and an optical cross- connect module 913 in the core switch 915. A performance simulation of the described spreading technique has been performed in regard to the network fragment in Fig. 9. In the simulation, each source OBS edge switch 902, 910 receives Poisson traffic (of fixed size packets of length 625 bytes) at an average rate of 0.5 Gbps. Since each wavelength operates at 10 Gbps, this corresponds to a loading 5% on each input link (911, 912) of the 8 x 1 core switch 913, and an average loading of 40% on the output link (914) of the core switch 913. Losses, if any, only occur due to contention in the core switch, which has limited all-optical delay lines of up to D time-slots. The OBS edge switches 902, 910 have enough electronic buffering such that they never have to drop traffic. The effect of varying the spreading-factor φ at the source OBS edge switches

902, 910 on the contention losses in the all-optical core switch and on the end-to-end delay of the packets is shown in Fig. 10, 11 and 12. Fig. 10 shows how packet loss probability varies with spreading factor φ for the simulated network of Fig. 9. The spreading factor φ for an input queue is defined as the fraction of the link capacity that is not accessible to the queue. The spreading factor therefore relates to the distribution of burst data-slices on a transmission link. For example, if the burst spreading factor indicates no spreading then all the data-slices of a burst occupy consecutive time-slots on a transmission link. If the burst spreading factor is any other setting up to a maximum, then the burst data-slices are spaced out as evenly as possible on the transmission link (i.e. a uniformly spread sequence). Fig. 10 shows for various D (ie the delay line lengths referred to in Fig. 6), the packet loss probability (log scale) as a function of the spreading-factor φ. First note that when D = 0, the contention losses are invariant to spreading (see 1001), as expected from analysis. When D = 1, the losses drop as spreading increases, but soon hit a saturation limit (see 1002) such that increasing the spreading does not result in any significant reduction in losses. As D increases, the saturation loss probabilities get progressively lower, and at D = 7 (see 1003), losses can be completely eliminated with sufficient spreading. This shows that as the spacing out of burst slices by the source edge switches 902-910 approaches regular spreading (as described in regard to Fig. 7), the contention losses drop drastically even with very limited buffering capability at the core switch 913. In fact, for the 8 input core switch 913 where each input link 911-912 carries regular traffic, D = 7 suffices to eliminate contentions completely. Fig. 11 shows at 1100 how packet delay varies with spreading factor. Spreading is achieved at the source edge switches 902, 910 by shaping the associated queues (such as 204 in Fig. 2) at a rate of 1 - φ. A consequence of having a high spreading factor φ is that the delays incurred by the packets at the ingress shaping queue (such as 202 in Fig. 2) grow. The 99.99%-ile end-to-end packet delay is measured as a function of the spreading factor. The end-to-end delay is counted from the time the packet is aggregated into a burst at the ingress (such as 916) to the time it is received in its entirety by the destination edge switch 908. Moreover, the fixed components (the propagation delay on the links, and the delay T0ffset for which a data burst is held while the control packet is sent through) are excluded. The 99.99%-ile delay value d is defined such that no more than 10"4 of the packets have a delay higher than d. The percentile value is indicative of the tail of the delay distribution, and is a more meaningful metric than the mean, especially for real-time traffic. Observe first that the plots 1102 for various D (see 1101) lie on top of each other in Fig. 11. This indicates that most of the delay is produced by the burst spreading process at the edge switch, whereas the core delay-buffering contribution is insignificant. Note that as the spreading factor increases to some limit (around 1103) the burst slices are spread out to thereby occupy an increasing number of "well spaced out" time-slots. This incurs increased delays in the queues (such as 202 and 204 in Fig. 2) due to a reduced shaping rate. The result is that delays rise sharply as can be seen at 1102. Fig. 12 shows how packet loss varies with delay. From Figs. 10 and 11 it is seen that increasing spreading φ reduces contention losses but increases end-to-end delay. These plots are now combined to show the loss versus delay trade-off. It is noted that the delay-loss trade-off relationship seems close to linear (where loss is on log-scale) as shown by the region 1201. The curves for D (0 to N-l; see 1202), where N denotes the number of input lines to the switch, peel off at their saturation loss rates, while those for

D > N peel off to drop to 0 losses. Also note that once D reaches 7, the incremental

advantages of increasing D are minimal. Fig. 13 shows one method 1300 for spreading bursts of time slices. The method

1300 is performed by the burst assembler 203 in Fig. 2. The process 1300 commences with a START step 1301, after which a testing step 1303 waits for a packet arrival. If a packet arrival is detected then following a YES arrow a testing step 1304 checks if the arrival queue 202 is empty. If this is the case, then following a YES arrow a step 1305 commences a Ttrig timer. Thereafter a step 1306 queues the packet in question in the arrival queue 202. Thereafter a testing step 1307 tests whether the size of the queue 202 is greater or equal to the value that has been set for the variable Lrig- If this is the case, then following a YES arrow a step 1308 stops the Ttπg timer. Thereafter a step 1309 assembles the packets in the queue 202 into a burst, after which a step 1310 moves the assembled burst to the burst queue 204. The process 1300 terminates at a "connection symbol" 1311 the details of which are described in relation to Fig. 14. Returning to the testing step 1303 if a packet arrival is not detected then the process 1300 is directed in accordance with a NO arrow to a testing step 1312 which determines if the Ttrig timer has expired. If this is the case then the process is directed in accordance with a YES arrow to the step 1308. If, on the other hand, the Ttπg timer has not expired then the process is directed in accordance with a NO arrow back to the step 1303. Returning to the testing 1307, if the size of the per-destination arrival queue 202 is not greater than or equal to Ltπg then the process 1300 is directed in accordance with a NO arrow back to the testing step 1303. Returning to the testing step 1304 if the arrival queue 202 is not empty, then the process 1300 is directed in accordance with a NO arrow to the step 1306. A connection point "C" designated 1302 is directed to the testing step 1303. The connection point "C" designated 1302 is described in relation to Fig. 15. Fig. 14 shows a process 1400 by which the burst spreader 205 determines the eligibility times according to which the burst slices in the burst queue 204 are spread. The process 1400 commences with the connection point 1311 from Fig. 13. Thereafter a step 1401 determines how a burst in the burst queue 204 is to be cut into slices. A subsequent 1402 sets an index "i" equal to 1. Thereafter a testing step 1403 checks whether the burst in question has a ith slice. If this is the case then the process 1400 is directed in accordance with a YES arrow to a step 1405. The step 1405 implements, according to the present arrangement, a leaky bucket approach which determines the eligibility time for the slice in question. In this approach the leaky bucket "rate" is set to

the value (1-φ) multiplied by the link rate (where the link rate is the variable "C" in table

1). The "depth" of the bucket is set to the value 0. The eligibility time is subject to the constraint that it must be greater than the present "actual" time plus the variable T0ffSet which is defined in Table 1. A subsequent step 1406 stores the calculated eligibility time for the slice in question in a table in the burst spreader 205. A following step 1407 increments the index "i" by a value of 1, and the process 1400 returns to the step 1402. Returning to the testing step 1403, if the burst in question does not have ith slice, then the process 1400 is directed in accordance with a NO arrow to a connection point "B" designated 1404 that is described in more detail with reference to Fig. 15. Fig. 15 shows a process 1500 in accordance with which the scheduler 207 operates (see Fig. 2) for the present arrangement. The process 1500 commences with the connection point "B" designated 1404 which emanates from the process 1400 in Fig. 14. Thereafter a step 1501 sets an index "i" to the value of 1. A subsequent testing step 1502 checks whether the burst in question has a "ith" slice. If this is the case, then the process 1500 is directed in accordance with a YES arrow to a step 1503. The step 1503 determines the minimum slot index "j" where j represents the lowest slot index in the table maintained by the scheduler 207 in relation to scheduled slots. This minimum value of "j" is subject to constraints, namely that j must be greater or equal to the eligibility time of the burst slice being considered, and the slot having an index j must be empty (meaning that the slot has not yet been allocated to a slice). Once the minimum value of "j" has been found a subsequent step 1504 allocates slot "j" in the table of the scheduler 207 to the slice having the presently considered index "i". Thereafter, a step 1505 allocates the corresponding slot position "j" to slice "i" in the BCC relating to the data slice being considered. Thereafter a step 1506 increments the index "i" and returns the process 1500 to the step 1502. th Returning to the step 1502 if the burst in question does not have a i slice then the process 1500 is directed by a NO arrow to a step 1507 that sends the burst control packet (BCC) into the network 100, after which the process 1500 is directed to the connection symbol "C" designated 1302 which connects back to the process 1300 in

Fig. 13. Industrial Applicability It is apparent from the above that the arrangements described are applicable to the telecommunications industry. The foregoing describes only some embodiments of the present invention, and modifications and/or changes can be made thereto without departing from the scope and spirit of the invention, the embodiments being illustrative and not restrictive. Thus, for example, the description is directed to traffic that enters the network in packets or frames (also referred to as Protocol Data Units or PDUs) to be subsequently assembled into data bursts that are then cut into data burst slices which are subsequently spread for transmission over the network using fixed-length time-slots. In the description a burst data-slice is the amount of data that can be carried in a time-slot end-to-end across the network. It is evident however that the inventive concept of cutting traffic entities of a generic type into slices and spreading the slices for transmission over the network is applicable to traffic entities other than packets, PDUs and data bursts.

Appendix A

Abbreviations

AON: All Optical Network

ATM: Asynchronous Transfer Mode BA: Burst Assembler

BCC: Burst Control Cell

BS: Burst Spreader

Demux: Demultiplexer

E/O: Electrical to Optical FIFO: First In First Out

GMPLS Generalised Multiprotocol Label S

LP Internet Protocol

LTD Lightpath Topology Design

MEMS Micro Electro-Mechanical Switch ms milliseconds Mux Multiplexer ns nanoseconds

N/W Network

OBS Optical Burst Switching

ODL Optical Delay Line

O/E Optical to Electrical

O/E/O Optical to Electrical to Optical

O/O Optical to Optical

OPA Optical Phased Array

OPS Optical Packet Switching

OTSI Optical Time-Slot Interchanger PDU Protocol Data Unit

PR Packet Reassembler

QoS Quality of Service

RAM Random Access Memory

RWA Routing Wavelength Assignment

SDH Synchronous Digital Hierarchy

SONET Synchronous Optical Network

Sync Synchronisation

TDM Time Division Multiplexing

TSOBS Time-Sliced OBS

TSS Time-Slot Switch

WDM Wavelength Division Multiplexin

Claims

Claims:
1. A method of transmitting traffic packets over a data link in a time-slotted optical
burst switching network, the method comprising the steps of: aggregating the traffic packets into a data burst; slicing the burst into a plurality of data slices; establishing a spreading factor associated with the data burst; and allocating the data slices to outgoing slots on the all-optical data link according
to the spreading factor.
2. A method according to claim 1, wherein each of the data slices comprises a
plurality of the traffic packets.
3. A method according to claim 1, wherein the establishing step establishes a
common spreading factor for successive groups of traffic packets, and the data slices
associated with each said successive group of traffic packets are allocated to outgoing
slots on the all-optical data link according to the common spreading factor.
4. A method according to claim 1, wherein the establishing step establishes a
distinct spreading factor for successive groups of traffic packets, and the data slices
associated with each said successive group of traffic packets are allocated to outgoing slots on the all-optical data link according to their corresponding distinct spreading factor.
5. A method according to claim 1, wherein: the traffic packets are transmitted from a source edge switch to a destination
edge switch; and the all-optical data link comprises at least one optical wavelength carried on a concatenated plurality of data link segments spanning the network between the source edge switch and the destination edge switch.
6. A method according to claim 1, wherein the aggregating step comprises the steps of: determining, for each of the traffic packets, traffic parameters establishing a corresponding Quality of Service Class; and aggregating traffic packets in different Quality of Service Classes into different corresponding data bursts.
7. A method according to claim 1, wherein prior to the allocating step, the method comprises pre-scheduling available time-slot opportunities on said data link.
8. A method according to claim 7, wherein the allocating step comprises, for a particular data slice, the steps of: determining the earliest slot position at which the data slice can be transmitted using a leaky bucket approach wherein the bucket rate is dependent upon the spreading factor; and allocating said data slice to a prescheduled time slot having a slot position no earlier than said earliest slot position dependent upon the pre-scheduled time-slot opportunities.
9. A method according to claim 6, wherein the aggregating step comprises, for a particular packet, the steps of: identifying a destination for the packet; allocating the packet to a queue depending on the destination and the associated Quality of Service Class; and assigning the packet to a data burst dependent upon at least one of a size of the queue and an elapsed time since a first said packet was allocated to the queue.
10. A method according to claim 1, wherein the aggregating step comprises, for a particular packet, the steps of: identifying a destination for the packet; allocating the packet to a queue depending on the destination; and assigning the packet to a data burst dependent upon at least one of a size of the queue and an elapsed time since a first said packet was allocated to the queue.
11. A method of receiving traffic packets that have been transmitted over a data link in a time-slotted optical burst switching network, wherein the packets having been aggregated into data bursts that have been sliced into data slices that have been allocated to outgoing slots on the all-optical data link according to a spreading factor, the method comprising the steps of: receiving, prior to receipt of the data slices, information comprising time-slot positions in which data-slices will arrive, and the number of data-slices which constitute a burst; and assembling the slices into corresponding traffic packets depending upon the received information.
12. A method of receiving traffic packets according to claim 11, wherein the received information further comprises an identification of the source switch from which the packets have been transmitted, and wherein subsequent to the receiving step and prior to the assembling step the method comprises, in regard to each said data slice, the further step of steering the data slice to a source-switch specific arrival queue according to the identification of the source switch.
13. A method of receiving traffic packets according to claim 12, wherein the received information further comprises a Quality of Service class for the packets, and wherein the steering step comprises steering the data slice to an arrival queue according to the received Quality of Service Class information.
14. An apparatus for transmitting traffic packets over a data link in a time-slotted optical burst switching network, the apparatus comprising: means for aggregating the traffic packets into a data burst; means for slicing the burst into a plurality of data slices; means for establishing a spreading factor associated with the data burst; and means for allocating the data slices to outgoing slots on the all-optical data link according to the spreading factor.
15. An apparatus according to claim 14, wherein the means for establishing the spreading factor establishes a common spreading factor for successive groups of traffic packets, and the means for allocating the data slices allocates the data slices associated with each said successive group of traffic packets to outgoing slots on the all-optical data link according to the common spreading factor.
16. An apparatus according to claim 14, wherein the means for establishing the spreading factor establishes a distinct spreading factor for successive groups of traffic packets, and the means for allocating the data slices allocates data slices associated with each said successive group of traffic packets to outgoing slots on the all-optical data link according to their corresponding distinct spreading factor.
17. An apparatus according to claim 14, wherein the means for aggregating the traffic packets comprises: means for determining, for each of the traffic packets, traffic parameters establishing a corresponding Quality of Service Class; and means for aggregating traffic packets in different Quality of Service Classes into different corresponding data bursts.
18. A method according to claim 14, wherein the means for allocating the data slices comprises means for pre-scheduling available time-slot opportunities on said data link.
19. An apparatus according to claim 18, wherein the means for allocating the data slices comprises: means for determining the earliest slot position at which the data slice can be transmitted using a leaky bucket approach wherein the bucket rate is dependent upon the spreading factor; and means for allocating said data slice to a prescheduled time slot having a slot position no earlier than said earliest slot position dependent upon the pre-scheduled time- slot opportunities.
20. An apparatus according to claim 17, the means for aggregating the traffic packets comprises: means for identifying a destination for the packet; means for allocating the packet to a queue depending on the destination and the associated Quality of Service Class; and means for assigning the packet to a data burst dependent upon at least one of a size of the queue and an elapsed time since a first said packet was allocated to the queue.
21. An apparatus according to claim 14, wherein the means for aggregating the traffic packets comprises: means for identifying a destination for the packet; means for allocating the packet to a queue depending on the destination; and means for assigning the packet to a data burst dependent upon at least one of a size of the queue and an elapsed time since a first said packet was allocated to the queue.
22. An apparatus for receiving traffic packets that have been transmitted over a data link in a time-slotted optical burst switching network, wherein the packets having been aggregated into data bursts that have been sliced into data slices that have been allocated to outgoing slots on the all-optical data link according to a spreading factor, the apparatus comprising: means for receiving, prior to receipt of the data slices, information comprising time-slot positions in which data-slices will arrive, and the number of data-slices which constitute a burst; and means for assembling the slices into corresponding traffic packets depending upon the received information.
23. An apparatus according to claim 22, wherein the received information further comprises an identification of the source switch from which the packets have been transmitted, and wherein the apparatus comprises: means which, in regard to each said data slice, steer the data slice to a source-
switch specific arrival queue according to the identification of the source switch.
24. An apparatus according to claim 23, wherein the received information further comprises a Quality of Service class for the packets, and wherein the steering means
comprise means for steering the data slice to an arrival queue according to the received
Quality of Service Class information.
25. A method of communicating traffic packets over a data link in a time-slotted
optical burst switching network, the method comprising, in regard to a transmitter, the
steps of: aggregating the traffic packets into a data burst; slicing the burst into a plurality of data slices; establishing a spreading factor associated with the data burst; allocating the data slices to outgoing slots on the all-optical data link according
to the spreading factor and transmitting the data slices over the network; and communicating, over the network, information comprising time-slot positions in
which data-slices will arrive, and the number of data-slices which constitute a burst; the
method further comprising, in regard to a receiver, the step of assembling the slices into
corresponding traffic packets depending upon the received information.
26. A system for communicating traffic packets over a data link in a time-slotted optical burst switching network, the system comprising: a transmitter comprising: means for aggregating the traffic packets into a data burst; means for slicing the burst into a plurality of data slices; means for establishing a spreading factor associated with the data burst; means for allocating the data slices to outgoing slots on the all-optical data link according to the spreading factor and transmitting the data slices over the network; and means for communicating, over the network, information comprising time-slot positions in which data-slices will arrive, and the number of data-slices which constitute a burst; a receiver comprising; means for assembling the slices into corresponding traffic packets depending upon the received information.
PCT/AU2004/001146 2003-08-29 2004-08-26 Time-sliced optical burst switching WO2005022800A1 (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
AU2003904705 2003-08-29
AU2003904705A AU2003904705A0 (en) 2003-08-29 Time-Sliced Optical Burst Switching

Publications (2)

Publication Number Publication Date
WO2005022800A1 true WO2005022800A1 (en) 2005-03-10
WO2005022800A8 WO2005022800A8 (en) 2005-05-12

Family

ID=34230056

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/AU2004/001146 WO2005022800A1 (en) 2003-08-29 2004-08-26 Time-sliced optical burst switching

Country Status (1)

Country Link
WO (1) WO2005022800A1 (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
FR2908577A1 (en) * 2006-11-10 2008-05-16 Canon Kk Quality of service resource managing method for e.g. domestic network, involves sending information representing accepted value to receiver device to initialize transmission of content based on window size equal to accepted value
CN104702397A (en) * 2013-12-05 2015-06-10 中兴通讯股份有限公司 Method and device for adjusting time slot length of OBTN (Optical Burst Transport Network) and node
CN107333190A (en) * 2017-06-01 2017-11-07 清华大学 Optic-electric hybrid switching structure with variable granularity

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2001084875A2 (en) * 2000-05-03 2001-11-08 Nokia, Inc. Robust transport of ip traffic over wdm using optical burst switching
US6721315B1 (en) * 1999-09-30 2004-04-13 Alcatel Control architecture in optical burst-switched networks
US20040120261A1 (en) * 2002-12-24 2004-06-24 Shlomo Ovadia Method and apparatus of data and control scheduling in wavelength-division-multiplexed photonic burst-switched networks

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US6721315B1 (en) * 1999-09-30 2004-04-13 Alcatel Control architecture in optical burst-switched networks
WO2001084875A2 (en) * 2000-05-03 2001-11-08 Nokia, Inc. Robust transport of ip traffic over wdm using optical burst switching
US20040120261A1 (en) * 2002-12-24 2004-06-24 Shlomo Ovadia Method and apparatus of data and control scheduling in wavelength-division-multiplexed photonic burst-switched networks

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
FR2908577A1 (en) * 2006-11-10 2008-05-16 Canon Kk Quality of service resource managing method for e.g. domestic network, involves sending information representing accepted value to receiver device to initialize transmission of content based on window size equal to accepted value
CN104702397A (en) * 2013-12-05 2015-06-10 中兴通讯股份有限公司 Method and device for adjusting time slot length of OBTN (Optical Burst Transport Network) and node
CN107333190A (en) * 2017-06-01 2017-11-07 清华大学 Optic-electric hybrid switching structure with variable granularity

Also Published As

Publication number Publication date
WO2005022800A8 (en) 2005-05-12

Similar Documents

Publication Publication Date Title
Chia et al. Packet loss and delay performance of feedback and feed-forward arrayed-waveguide gratings-based optical packet switches with WDM inputs-outputs
Listanti et al. Architectural and technological issues for future optical Internet networks
Duser et al. Analysis of a dynamically wavelength-routed optical burst switched network architecture
Chipalkatti et al. Protocols for optical star-coupler network using WDM: Performance and complexity study
US6778536B1 (en) Combined wavelength division multiplexing, time division multiplexing, and asynchronous packet switching with common time reference
US7801132B2 (en) Interface system and methodology having scheduled connection responsive to common time reference
Porter et al. Integrating microsecond circuit switching into the data center
Hunter et al. Approaches to optical Internet packet switching
Xiong et al. Control architecture in optical burst-switched WDM networks
Ramamirtham et al. Time sliced optical burst switching
US6757282B1 (en) Fast switching of data packet with common time reference
US6674754B1 (en) Wavelength division multiplexing combined with time division multiplexing using a common time reference
US20030189933A1 (en) Shared wavelength group to differentiate label switched paths for congestion control in optical burst switching networks
EP1177657B1 (en) Optical communications network
US6665495B1 (en) Non-blocking, scalable optical router architecture and method for routing optical traffic
Jue et al. Optical burst switched networks
Chlamtac et al. Quadro-star: A high performance optical WDM star network
Herzog et al. Metropolitan area packet-switched WDM networks: A survey on ring systems
Vokkarane et al. Burst segmentation: An approach for reducing packet loss in optical burst switched networks
US20030043430A1 (en) Optical packet switching apparatus and methods
Battestilli et al. An introduction to optical burst switching
Qiao et al. Optical burst switching (OBS)–a new paradigm for an Optical Internet^{1}
AU781132B2 (en) Control architecture in optical burst-switched networks
EP1578048B1 (en) Token-controlled data transmissions in communication networks
Veeraraghavan et al. Architectures and protocols that enable new applications on optical networks

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A1

Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BW BY BZ CA CH CN CO CR CU CZ DE DK DM DZ EC EE EG ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX MZ NA NI NO NZ OM PG PH PL PT RO RU SC SD SE SG SK SL SY TJ TM TN TR TT TZ UA UG US UZ VC VN YU ZA ZM ZW

AL Designated countries for regional patents

Kind code of ref document: A1

Designated state(s): BW GH GM KE LS MW MZ NA SD SL SZ TZ UG ZM ZW AM AZ BY KG KZ MD RU TJ TM AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HU IE IT LU MC NL PL PT RO SE SI SK TR BF BJ CF CG CI CM GA GN GQ GW ML MR NE SN TD TG

121 Ep: the epo has been informed by wipo that ep was designated in this application
CFP Corrected version of a pamphlet front page
CR1 Correction of entry in section i

Free format text: IN PCT GAZETTE 10/2005 UNDER (71) REPLACE "IP1 (AUSTRALIA) PTY LTD" BY "IP1 (AUSTRALIA) PTY LTD (RECEIVERS ANDMANAGERS APPOINTED) (ADMINISTRATORS APPOINTED)"

122 Ep: pct application non-entry in european phase