GB2474863A

GB2474863A - Optical buffer with controllable delay

Info

Publication number: GB2474863A
Application number: GB0918895A
Authority: GB
Inventors: Bart Steyaert; Carl Sys; Herwig Bruneel
Original assignee: Universiteit Gent
Current assignee: Universiteit Gent
Priority date: 2009-10-28
Filing date: 2009-10-28
Publication date: 2011-05-04
Also published as: GB0918895D0; WO2011051330A1

Abstract

An optical buffer (100) for buffering optical signals. The optical buffer comprises a plurality of fibre delay lines (110) of different lengths and a controller (120) adapted for controlling the delay of packet in a packet stream. The controller is adapted for delaying a packet by subsequently using different fibre delay lines and for using, for the delay of each packet, each fibre delay line at most once.

Description

Methods and systems for optical buffering

Field of the invention

The invention relates to the field of optical data communication. More particularly, the present invention relates to methods and systems for optical buffering using optical delay lines.

Background of the invention

In electronic switches, packets in contention are stored in a RAM and sent out one by one when the output is available.

The main incentive for industries to move to optical switching is the better performance that can be expected, when compared to electronic switching. Since optical switching necessitates some kind of buffering, optical buffers will play an essential role in years to come. In an optical switch, a different buffering approach is required because there is no ready-to-use random access optical memory.

In some applications, as described e.g. in US06934471, an optical to electronic and electronic to optical conversion is performed, so as to be able to use a conventional RAM memory for electronic signals. Currently, also optical cross connects (OXC) are being developed wherein optical buffering and (circuit, packet and burst) switching is obtained in the optical domain and will result in much higher network performance.

Optical buffering of packets relies on delaying packets by increasing their total transmission time, either by decreasing the group velocity (slow light buffers), or increasing the physical length to be travelled by the packet (delay line buffers). Slow light devices can be divided into two groups: devices using material-based resonances and those using coupled resonance structures (CRS). Burmeister et al. have indicated in A comparison of optical buffering technologies" in Science Direct 2007, that slow light buffers would degrade network performance by limiting the packet length and therefore would degrade the network load. They indicated that the technologies suffer from losses and dispersion which yield a low bandwidth-delay product, resulting in impractical bitrates and capacities.

Although optical buffering can be implemented in different ways, optical FDL buffering, by means of delay lines, is considered the only feasible option for networks. The most common proposed solution is buffering using a set of long fibre lines. Delay line buffers can also be divided into two subsets: feed forward and feedback. Data in a feed forward buffer is sent through exactly one of the delay lines of the buffer. In a feedback buffer, packets leaving the buffer may couple back to the same buffer. In a feedback buffer the length of the basic delay loop should in general be no less than the length of the packet payload.

For reasons of miniaturization, the use of fibre delay lines in a loop is often suggested.

Such loops are e.g. recirculation loops whereby feedback with one fibre delay line per buffer is used. Although fibre loops provide a solution to some extent, typically their capacity is very limited.

Two particular solutions are suggested by Beheshti et al. and Wang et al. Beheshti et al. describe in Packet Scheduling in Optical FIFO Buffers" a fibre delay line system wherein the number or length of the fibre delay lines can be reduced, by buffering arriving packets in a FIFO sequence in case the output is not always available due to some external process. The system is based on packets travelling through the same fibre delay lines for a number of times, i.e. recirculation occurs. The system is described for an environment wherein at most one packet arrival per time slot is present. Furthermore, the algorithm provided is for fixed-size packets, which is not always the case in optical data communication.

In Chinese patent application 2007/10092531, Wang et al. describe a system for using a set of fibre delay lines to realize a certain delay. The proposed system is based on fibre delay lines having the same length, resulting in the necessity for the optical packet to pass through a long series of fibre delay lines if a long delay is required.

Alternatively, the use of wavelength multiplexing techniques also is suggested for controlling packet delay. From e.g. US 6,819,870 it is known to use wavelength division multiplexing fibre delay line optical buffering and scheduling. In this technique a different wavelength may be assigned to different packets and routing to an optical buffer is performed taking into account the wavelength assignment.

Summary of the invention

It is an object of embodiments of the present invention to provide good methods and systems for optical buffering. It is an advantage of embodiments according to the present invention that compact and scalable fibre delay lines (FDL) systems can be obtained. It is an advantage of embodiments according to the present invention that, for the same setting of fibre delay lines, the algorithm results in low packet loss. The latter holds for different parameter settings for load, number of FDLs N, packet length L and granularity G. In other words, for the same target packet loss, a small number of fibre delay lines and small physical space is required.

It is an advantage of embodiments according to the present invention that low paket loss can be obtained, for different packet arrival patterns.

It is an advantage of embodiments according to the present invention that any set of fibre delay lines lengths can be used.

It is an advantage of embodiments according to the present invention that fibre delay lines can be used with a good efficiency, without requiring feedback and therefore without the need of complex data processing after each fibre delay line the packet has run through.

It is an advantage of embodiments according to the present invention that the packet sizes that can be buffered can vary in size. Furthermore, it is an advantage of embodiments according to the present invention that the optical packets may have unlimited length.

It is an advantage of embodiments according to the present invention that the loss probability is good, for a low number of fibre delay lines as well as for a larger number of fibre delay lines.

It is an advantage of embodiments according to the present invention that accurate contention detection can be performed.

It is an advantage of embodiments according to the present invention that the optical buffer route for a packet or the decision to drop a packet can be completely determined upfront, i.e. prior to entering of the packet in a first delay line.

It is an advantage of embodiments according to the present invention that routing can be performed for variable length packets.

The above objective is accomplished by a method and device according to the present invention.

The present invention relates to an optical buffer for buffering optical signals, the buffer comprising a plurality of fibre delay lines of different lengths, and a controller adapted for controlling the delay of packets in a packet stream, wherein the controller is adapted for delaying a packet by subsequently using different fibre delay lines and for using, for the delay of each packet, each fibre delay line at most once. It is an advantage of embodiments according to the present invention that various and large delays can be obtained with compact buffers. It is an advantage of embodiments according to the present invention that scalable optical buffers are provided.

The controller may be adapted for using the fibre delay channels in a predetermined order. The controller may be adapted for using the fibre delay lines in a decreasing order of length. It is an advantage of embodiments according to the present invention that by using a decreasing length, the longest fibre delay lines are again available most quickly, resulting in the possibility to provide longer delays for subsequently arriving packets.

The controller may comprise a memory and is adapted for storing the length of each of the plurality of fibre delay lines, the overall output scheduling horizon, and the output scheduling horizon for each of the plurality of fibre delay lines. It is an advantage of embodiments according to the present invention that the amount of data to be stored is limited, resulting in an easily scalable optical buffer system.

The controller may be adapted for operating the optical buffer as feed forward buffer. It is an advantage of embodiments according to the present invention that a good fibre delay line usage per packet can be obtained without the need for feedback.

The controller may comprise a processor for determining the set of different fibre delay lines subsequently used for delaying a packet before entering of the packet in the fibre delay lines. It is an advantage of embodiments according to the present invention that no intermediate evaluation and thus processing is required, resulting in the possibility of using less complex processors. It is an advantage of embodiments according to the present invention that by avoiding intermediate evaluation the algorithm used can be more simple. The processor may be adapted for retrieving possible combinations of fibre delay lines for predetermined delays based on a look up table. It is an advantage of embodiments according to the present invention that processing of the buffering to be applied can be based on conventional techniques, requiring less on-the-spot processing.

The controller may be adapted for determining the combination of fibre delay lines realizing the required delay by taking into account all combinations of fibre delay lines realizing the delay and selecting that combination of fibre delay lines wherein the longest delay line of the combination is the shortest.

The system furthermore may comprise a wavelength converter for converting an optical packet to one of a set of wavelengths distinguishably transmittable through the fibre delay line. It is an advantage of embodiments according to the present invention that the optical buffer techniques can be combined with wavelength multiplexing, thus allowing to substantially increase the capacity of the optical buffer.

At least one fibre delay line can be used for distinguishably transmitting different wavelengths through the fibre delay line, and the controller furthermore may be adapted for using wavelength selection as well as fibre delay line selection for buffering subsequent packet in a packet stream.

The present invention also relates to an optical data communication system, the optical data communication system comprising an optical buffering system as described above. The optical data communication system may for example be an optical switch fabric.

The present invention also relates to a method for optically buffering optical packets, the method comprising controlling the delay of a packet in an optical packet stream by subsequently using different fibre delay lines and by using, for the delay of each packet, each fibre delay line at most once.

The method may comprise using the fibre delay channels in decreasing length order.

The method may comprise determining a delay value D by which the packet at least needs to be delayed, determining a combination of fibre delay lines realizing the delay D by taking into account all combinations of fibre delay lines realizing the delay D, and selecting that combination of fibre delay lines wherein the longest delay line of the combination is the shortest.

The method may comprise, if the determined combination does not prevent contention, selecting a different combination allowing realizing the delay D. The method furthermore may comprise, if no different combination can be found, updating the delay value D using the granularity of the set of delay lines, and determining a combination of fibre delay lines realizing the updated delay.

The method may comprise scheduling the packet to the combination of fibre delay lines allowing obtaining a suitable delay.

The method may comprise storing the length of each of the plurality of fibre delay lines, the overall output scheduling horizon, and the output scheduling horizon for each of the plurality of fibre delay lines.

The method may operate an optical buffer as feed forward buffer.

The method may comprise determining the set of different fibre delay lines subsequently to be used for delaying a packet before entering the packet in the fibre delay lines.

The method may comprise using a look up table for determining a combination of fibre delay lines to be used.

The method may comprise determining a combination of fibre delay lines and a wavelength for transmission of the packet in the combination of fibre delay lines.

The present invention also relates to a controller for an optical buffer, the controller being configured for delaying a packet by subsequently using different fibre delay lines and for using, for the delay of each packet, each fibre delay line at most once.

The controller may be adapted for performing a method as described above.

The present invention also relates to a computer program product for, when executed on a computer, performing a method for optically buffering optical packets as described above.

The present invention also relates to a machine readable data storage device storing such a computer program product or to the transmission of such a computer program product over a local or wide area telecommunications network.

Particular and preferred aspects of the invention are set out in the accompanying independent and dependent claims. Features from the dependent claims may be combined with features of the independent claims and with features of other dependent claims as appropriate and not merely as explicitly set out in the claims.

These and other aspects of the invention will be apparent from and elucidated with reference to the embodiment(s) described hereinafter.

Brief description of the drawings

FIG. 1 illustrates an optical buffer system according to an embodiment of the present invention.

FIG. 2 illustrates a switch fabric for controlling contention, comprising a plurality of contention resolution blocks.

FIG. 3 illustrates a contention resolution block being a conventional feed forward

optical buffer, as known from prior art.

FIG. 4 illustrates a contention resolution block being an optical buffer according to an embodiment of the present invention.

FIG. 5 to 7 illustrate simulation results comparing loss probability for a basic feed forward system and a system according to an embodiment of the present invention for different traffic conditions and/or different fibre delay line configurations.

FIG. 8 illustrates an implementation of an algorithm for optically buffering an optical data packet according to an embodiment of the present invention, whereby only combinations of maximum two fibre delay lines are considered.

FIG. 9 illustrates an exemplary algorithm for optically buffering an optical data packet according to an embodiment of the present invention, whereby the granularity equals one.

FIG. 10 illustrates a second exemplary algorithm for optically buffering an optical data packet according to an embodiment of the present invention.

FIG. 11 illustrates a look up table listing all the combinations and their associated delay, as can be used in embodiments according to the present invention.

FIG. 12 illustrates the use of different wavelengths (WDM) in an optical buffering system.

FIG. 13 to FIG. 16 illustrate implementations and algorithms for exemplary optical buffering methods using wavelength division multiplexing, according to embodiments of the present invention.

The drawings are only schematic and are non-limiting. In the drawings, the size of some of the elements may be exaggerated and not drawn on scale for illustrative purposes.

Detailed description of illustrative embodiments

While the invention will be illustrated and described in detail in the drawings and the following description, such illustration and description are to be considered illustrative or exemplary and not restrictive. The invention is not limited to the disclosed embodiments.

Other variations to the disclosed embodiments can be understood and effected by those skilled in the art in practicing the claimed invention, from a study of the drawings, the disclosure and the appended claims. In the claims, the word "comprising" does not exclude other elements or steps, and the indefinite article "a" or "an" does not exclude a plurality. A single processor or other unit may fulfil the functions of several items recited in the claims. The mere fact that certain measures are recited in mutually different dependent claims does not indicate that a combination of these measures cannot be used to advantage. A computer program may be stored/distributed on a suitable medium, such as an optical storage medium or a solid-state medium supplied together with or as part of other hardware, but may also be distributed in other forms, such as via the Internet or other wired or wireless telecommunication systems. Any reference signs in the claims should not be construed as limiting the scope.

Where in embodiments according to the present invention reference is made to "voids", reference is made to the time in-between the transmission of two packets, during which the outgoing channel remains unused, despite of the fact that packets in the buffer are awaiting transmission.

Where in the present embodiments reference is made to the output scheduling horizon", reference is made to the time a new packet has to wait before being sent to the output (because it is reserved by previously scheduled packets). Where in embodiments of the present invention reference is made to the scheduling horizon of fibre delay lines", reference is made to the time a new packet has to wait before being sent to the fibre delay line (because it is reserved by previously scheduled packets).

Where in embodiments according to the present invention reference is made to the granularity" G, reference is made to a fixed period of time, whereby the fibre delay line lengths used are multiples of G. Where in embodiments according to the present invention reference is made to the gloss probability", reference is made to the probability that a packet is lost in data communication because it cannot be buffered by the system and it cannot be outputted.

In a first aspect, the present invention relates to a system and method for optical buffering of an optical packet stream. The optical packet stream may stem from any optical source and may be provided e.g. via an optical fibre, via a waveguide, directly initiated from a radiation source, etc. The optical buffering system may be part of an optical data communication system such as for example an all-optical packet switch fabric, an all-optical burst switch fabric, an all-optical router, etc. The optical buffering system according to embodiments of the present invention comprises a plurality of fibre delay lines of different time lengths. The number of fibre delay lines may be at least 2 such as for example 2, 3, 4, 5, 10, 20, 40 The time lengths of the different fibre delay lines may be equidistant, i.e. the lengths of the different fibre delay lines may be equally spaced, although embodiments of the invention are not limited thereto. The time length of the fibre delay line determines the time it takes to pass through the fibre delay line. The time lengths may be multiples of one time unit or multiples of a plurality of time units, in other words, the granularity may be one or a larger integer value.

The optical buffering system according to embodiments of the present invention furthermore comprises a controller adapted for controlling the delay of a packet in a packet stream. The controller may be performed in hardware as well as in software.

In the latter case, it may be implemented as a computer program product on a computer. The controller is adapted, e.g. configured or programmed, for delaying a packet of an optical packet stream by subsequently using different fibre delay lines, also referred to as FDLs. Furthermore, the controller is adapted, e.g. configured or programmed, for using each fibre delay line at most once, for the delay of each packet. In other words, if one packet is to be delayed, no fibre delay line is used twice.

By way of illustration, embodiments of the present invention not being limited thereto, an exemplary optical buffer system 100 is shown in FIG. 1, illustrating standard and optional components of an optical buffer system 100 according to an embodiment of the present invention. The optical buffer system 100 comprises a plurality of fibre delay lines 110 and a controller 120. The controller may comprise a memory 122 for storing data. In embodiments according to the present invention, the memory 122 advantageously may be sufficiently large for storing the length of each of the fibre delay lines 110 used and the output scheduling horizon as well as the fibre delay line scheduling horizon for each of the fibre delay lines 110. As indicated above, the scheduling horizon is the time a new packet has to be delayed before it can be sent to the output or the fibre delay line, because previously scheduled packets reserved the output or the fibre delay line. As the size of the information to be stored then is proportional to N, the memory 122 used may be a conventional ii memory component and no stringent requirements are posed on the memory component.

The controller may be adapted for operating the optical buffer as feed forward buffer. The latter implies that the determination of the combination of fibre delay lines to be used is determined up front. The latter is advantageous as it avoids the necessity for intermediate evaluation. The controller 120 may comprise a processor 124 for determining which set of different fibre delay lines should be subsequently used. The processor 124 may be a microprocessor or any other processor allowing evaluation of the different fibre delay lines. The processor may operate based on a predetermined algorithm, using neural networks, using look up tables, etc. The processor 124 may be adapted for determining the combinations of the fibre delay lines realizing the required delay. Multiple combinations of the fibre delay lines may produce the minimum delay required for resolving contention. The processor 124 may be adapted for selecting from the multiple combinations the combination of fibre delay lines wherein the longest delay line of the combination is the shortest. The controller 120 advantageously is adapted for leaving the lengthiest fibre delay lines unused. If the preferred combination of fibre delay lines does not allow avoiding conflicts, another combination realizing the same delay may be selected. As indicated above, the combination may be selected such that for each combination, each fibre delay line is used at most once. The processor furthermore may be adapted for selecting combinations of fibre delay lines corresponding with a larger delay, if none of the combinations realizing the minimum required delay results in a contention free solution. The delay may be increased until a suitable combination resulting in a contention free solution is found or until the largest delay of the optical buffer system is reached. In the latter case, the packet will be dropped. If a combination resulting in a contention free solution is found, the packet is routed using this combination of fibre delay lines. The processor 124 thereby is adapted for using the fibre delay lines in a predetermined order, e.g. a decreasing length order. The system may be adapted for operating at one wavelength only.

In some embodiments, the present invention relates to methods and systems for optical buffering as described above, whereby furthermore use is made of wavelength division multiplexing (WDM). In wavelength division multiplexing, the inputs and outputs carry different wavelengths, e.g. different wavelengths W. Packets can go through inputs and/or outputs at any of the different wavelengths W. It can be supposed that a packet can be transmitted at any wavelength without restrictions, i.e. whereby all wavelengths in a delay line are logically identical. From a buffering point of view the W wavelengths in the optical buffer behave as W identical, parallel optical buffers, one for each wavelength. Physically, there is only a single optical buffer. The wavelengths are considered independent because they do not interfere with each other when transmitted through the fibre delay lines.

The present invention can be combined with contention resolution in the wavelength domain. Different algorithms that can be used for choosing the optimum wavelength may be JLSO. (join the longest shortest queue) or other algorithms which may outperform JLSQ, but do some form of void filling and therefore are more complex.

Two exemplary combinations may be finding the optimum wavelength according to the JLSO. among all free paths with the best fibre delay line combination or finding the optimum wavelength according to the JLSO. among all free paths with the lowest delay. With free path" there is meant a combination of fibre delay lines at one particular wavelength that does not introduce a conflict with previously scheduled packets. The two exemplary combinations are shown further below.

It is to be noticed that both the wavelength division multiplexing algorithm and the sequential fibre delay line use algorithm can be easily combined resulting in a more effective scheduling algorithm.

According to some embodiments of the present invention, the optical buffer therefore may comprise a wavelength converter 130 for converting the wavelength of the optical package to the wavelength used for buffering. An additional converter may be used for re-converting the wavelength of the optical package to its original wavelength.

The optical buffer system 100 also may comprise an optical input port 140 for receiving the optical packet data stream and an optical output port 150 for outputting the optical packets, optionally after buffering. Such input and/or output ports may for example be in the form of waveguides or optical fibres.

The optical buffer system 100 also may comprise a powering system 160 for powering the optical buffer components such as for example the controller and/or the wavelength converter.

In some embodiments, the system is adapted for switching wavelengths in between the different fibre delay lines used. Although wavelength converters then should be present in between the fibre delay lines and although a more complex scheduling algorithm is required because the different wavelengths are not independent anymore, this could result in a lower loss probability compared to a system where no intermediate wavelength switch could be provided.

In a second aspect, the present invention relates to a method for optically buffering optical data packets in an optical data stream. The method may advantageously be performed using a system as described in the above aspect. The method comprises controlling the delay of a packet in an optical packet stream by subsequently using different fibre delay lines and by using, for the delay of each packet, each fibre delay line at most once. Other optional method steps may be steps expressing the functionality of the features of the system as described in the first aspect or may for example be as one or more features described with respect to the exemplary algorithms, descriptions thereof or implementations thereof.

In a third aspect, the present invention also relates to a controller for controlling optical buffering according to a method as described in embodiments of the second aspect or for controlling an optical buffering system as described in the first aspect of the present invention. Further features and advantages may be as described with respect to the optical buffer system, examples thereof or may be components adapted for performing the functionality described by methods according to the present invention.

In a fourth aspect, the present invention also relates to an optical data communication system comprising an optical buffer as described in embodiments of the first aspect.

By way of illustration, embodiments of the present invention not being limited thereto, a plurality of examples is discussed indicating standard or optional features and advantages of embodiments according to the present invention.

Simulation results are shown wherein a system according to an embodiment of the present invention is compared to a conventional basic feed forward optical buffer. In a first example, contention occurring in a switch fabric is illustrated. FIG. 2 illustrates contention resolution at the output, whereby a switch fabric provides different inputs to different outputs and whereby for a given output a contention resolution block by means of an optical buffer is used for controlling contention.

It is to be noticed that the configuration shown illustrates contention resolution blocks at the output. Other configurations where the contention resolution block is at the input port also exist and could be implemented using embodiments of the present invention. The system may be based on a control unit detecting contentions and sending messages to the contention resolution blocks to solve the contentions.

The control unit also may be the unit sending control signals to the switching elements in order to rout a packet via a predetermined output.

Where the optical buffer is a conventional feed forward system, exactly one fibre delay line is used or the packet is immediately sent to the output. Such a conventional system operates as follows: On each packet arrival, the output scheduling horizon H is updated. If H is larger than the largest FDL, the packet is dropped. Otherwise, the packet is sent to the shortest FDL with a delay of at least H. The particular example of system according to an embodiment of the present invention is based on a set of fibre delay lines wherein the lengths are consecutive multiples of the granularity G, as described above. The latter is also illustrated in FIG. 3, illustrating a conventional feed forward optical buffer as contention resolution block. In the particular example, the optical buffer contains three fibre delay lines with lengths 3, 2 and 1, and also a direct connection to the output. A packet is sent immediately to the output or it is delayed by sending it through exactly one FDL. For the same set of delay lines, in FIG. 4 the contention resolution block is shown according to an embodiment of the present invention whereby different delay lines may subsequently be used for the same packet. In this way, as can be seen longer delays are possible, thus illustrating an advantage of embodiments according to the present invention.

In further examples, simulation results are shown of the loss probability, as function of different parameters. In a first example in FIG. 5, the loss probability is shown as function of granularity for an optical buffer with 15 fibre delay lines, whereby the packet size is 50 time units and the load is 60%. Simulation results for the conventional feed forward system are indicated by circles, whereas for the system according to an embodiment of the present invention the simulation results are indicated with squares. It can be seen that the loss probability can be up to about one order or magnitude larger in the conventional feed forward system than in a system according to an embodiment of the present invention. The minima that appear for granularity equal to the average packet length A and the average packet length divided by two A/2 typically occurs for optical buffers. The global minimum is A for low loads and shifts to A/2, A/3, ... for higher loads. FIG. 6 illustrates the loss probability as function of the number of fibre delay lines N for a fixed packet size equal to time units, a load of 60% and a granularity of 50 time units (which is the optimum granularity for this load). It can be seen that the system according to an embodiment of the present invention has a significant lower loss probability compared to the conventional basic feed forward system, the gain being larger for a larger number of fibre delay lines. FIG. 7 illustrates the loss probability as function of the load. For each load the optimum granularity is chosen (i.e. a granularity of 50 time units for a load 40% and a granularity of 25 time units for higher loads). The packet size is 50 time units and the number of fibre delay lines is 10. Again it can be seen that the loss probability is much smaller for the system according to an embodiment of the present invention, the advantage being larger for smaller loads.

Another set of examples illustrates implementations and algorithms according to embodiments of the present invention. A first implementation is shown in FIG. 8 and a corresponding algorithm is shown in FIG. 9. The exemplary embodiments are illustrated for combinations of two fibre delay lines merely for ease of discussion and thus are not limited thereto. Different standard and optional steps of the exemplary method are now discussed in detail.

In a first step, the packet arrives at the optical buffer. The packet may be obtained from any source of optical data packets, such as an optical processing circuit, an optical fibre, a part of a switch fabric, etc. In a second step, the output scheduling horizon, indicated as delay[0], and all the fibre delay line scheduling horizons, indicated as delay[i], 1 «= i «= N are updated.

In a third step, which is a decision step, it is evaluated whether the output is idle and consequently whether the output scheduling horizon equals zero, in which case the packet can be sent directly, i.e. without optical buffering, as there is no conflict with previously scheduled packets and the output scheduling horizon is updated.

In a next step, if the packet cannot be sent directly, the delay value D is initialised by setting it to the value of the output scheduling horizon, i.e. D = delay[0]. As indicated above, the output scheduling horizon is the time the output is reserved by previously arrived packets. It is to be noticed that in case the granularity G for the fibre delay lines is not one, the initial value for the delay D can be calculated as D = r delay[0] I G 1 * G, whereby r x 1 means the lowest integer larger than or equal to x, e.g. r 4.01 1 = 5. Next values of the delay then will be determined by adding the granularity G to the current delay value, i.e. Dnew = D + G. In a subsequent step, all combinations of the fibre delay lines are chosen that realise a delay D, and from these, the one that is considered the best choice is selected.

From all the possible combinations realizing a certain delay, the best choice may be obtained by comparing the longest fibre delay line of each of the combinations and choosing that combination with the shortest one. For instance, in the case of an optical buffer consisting of a set of 10 delay lines with lengths 1, 2, 3, .., 10, there are three ways to realize a delay of 15 with two FDL5: the combination 8+7, the combination 9+6 and the combination 10+5. The longest FDLs in the respective combinations are 8, 9 and 10. The shortest of these longest FDLs is 8, so we first choose the associated combination, which is 8+7. The second choice is 9+6, the last choice 10+5. This leaves the longer fibre delay lines unused, and hence longer delays are possible for future packets.

In a further step, after the best combination of fibre delay lines is calculated for realizing the delay D, if scheduling the newly arrived packet through this combination of FDLs is free from collisions with previously scheduled packets, then the packet is scheduled through this combination of FDLs. In the special case that only combinations with maximum two FDLs are considered, the condition that the newly arrived and scheduled packet does not collide with a previously scheduled packet is equivalent with the condition "delay[i]=O". This condition is used in the algorithm implementation of FIG. 8, although embodiments of the invention are not limited thereto.

By way of illustration, the latter can be appreciated as follows. There is no conflict with previously scheduled packets if the first fibre delay line i is free and the second fibre delay line j is free by the time the packet leaves the first fibre delay line i. The second condition is always fulfilled if only combinations with maximum two fibre delay lines are considered. This can be seen as follows: The total delay realized by a chosen combination of delay lines i and j is at least the output scheduling horizon, as the new packet has to be delayed at least until the output becomes free, i.e. length[i] + length[j] »= delay[0] For every delay line k, 0 «= k «= N, the following relation holds: delay[k] + length[k] «= delay[0] since all previously scheduled packets leave the buffer before or at last at delay[O].

Combining the first equation with the second equation applied for delay line j: delay[j] + length[j] «= length[i] + length[j] Or, after striking out length[j] on both sides delay[j] «= length[i] This last equation actually indicates that the horizon of fibre delay line j is less than the length of fibre delay line i, thus, by the time the new packet leaves fibre delay line i, the horizon of fibre delay line j has become zero and thus is free.

If there is a conflict and the packet cannot be scheduled, in a subsequent step, it is tried to find the best combination of fibre delay lines among the remaining combinations of the fibre delay lines that realize the delay D. If such another combination realises the delay D, it is again evaluated whether or not there is a conflict with previously scheduled packets, and if not the packet is scheduled. If no other combination can be found or the other combinations also give rise to a conflict, the delay is increased with the granularity and the process for calculating the best combination of fibre delay lines realizing the updated delay and, if required, calculating further combinations is repeated.

The process is repeated until a suitable non conflicting combination is found, resulting in the package being scheduled, or until the delay is larger than a maximum delay value, e.g. predetermined by the user, resulting in the package being dropped.

Whereas the flow chart and description provided above has been mainly described for a granularity G of one, as indicated this can be easily extended to an arbitrary granularity. Furthermore, the flow chart and description can also easily be extended to optical buffers wherein non-equidistant fibre delay lines are used. In the last case, there is no general formula for the calculation of G, but a table could be used listing the possible FDL combinations and their respective delays as an ordered sequence.

Furthermore, whereas the algorithm has been described for a maximum of two fibre delay lines, the algorithm is not restricted thereto.

In a second example, a similar algorithm as in the first exemplary algorithm is considered, wherein the different allowed combinations of fibre delay lines are not calculated during scheduling, but are permanently stored in a memory. An example of a corresponding algorithm is shown in FIG. 10. The different combinations are known in advance because they only depend on the hardware, i.e. the fibre delay line lengths. All the combinations can be stored in an ordered list. The combinations can be ordered from low delay to high delay. For combinations with the same delay, there is an additional ordering from better' to worse' combinations. To schedule a packet, if the output is not free, the first combination in the list with a delay greater than or equal the output scheduling horizon is tried first. If it leads to a conflict, then the next combination in the list is tried, and so on... By way of illustration, the ordering of the combinations table in the special case where only combinations with maximum two fibre delay lines are allowed is shown. The first ordering criterion can be the delay D, as before. The second ordering criterion could be the longest fibre delay line length in the combination (thus ordering from the smallest longest fibre delay line length to the longest longest fibre delay line length). Consider for example an optical buffer with equidistant fibre delay lines and granularity G = 1, the data being illustrated in FIG. 11.

The right table contains all possible combinations with 2 fibre delay lines for a configuration with 10 fibre delay lines with lengths 1, 2, 3, ... 10. The highest possible delay is 19. The third and fourth columns of the right table show the first and second ordering criterion. These columns do not need to be part of the table or stored in the memory. Combinations with the same delay are ordered according to ascending longest delay line. To use this combinations table in the algorithm, the first combination that realizes a certain delay, e.g. delay[0], is needed. To retrieve this, a second (shorter) table can be used that lists for each possible delay, the combinations table index of the first combination that realizes this delay. Suppose for example that the required delay is 15. The left table indicates that the first (and best) combination realizing this delay is found at index 47 in the combinations table. In the combinations table at position 47 is combination 8+7. It first can be checked if this combination leads to some conflict with previously scheduled packets. If not, then send the packet through fibre delay line 8 and fibre delay line 7. If there is a conflict, the next combination can be tried, i.e. fibre delay line 9 and fibre delay line 6. The process can be continued so on... If none of the combinations is free from conflict, then the packet is dropped.

In still another set of examples, the use of a JLSO. algorithm is illustrated, whereby a comparison of the combination of JLSO. with a conventional feed forward optical buffer on the one hand and the combination of JLSO. with a buffer system according to examples of the present invention on the other hand is shown.

First the combination of the known algorithm JLSQ. with the conventional feed forward optical buffer is discussed. In a memory of/for the buffer, the output scheduling horizon value for each wavelength is stored and this is updated when a new packet arrives. Upon packet arrival, the JLSQ algorithm assigns wavelength i to the transmission of the packet such that two conditions are met: the first condition being that the selected wavelength i offers minimum delay (shortest queue) and the second condition being that among the wavelengths that offer minimal delay, the wavelength i has the longest scheduling horizon (longest shortest queue). The packet is converted to the chosen wavelength i and routed to the fibre delay line realizing the above mentioned minimal delay.

By way of illustration, operation of the JLSO. algorithm is shown in FIG. 12, illustrating the use of 6 different wavelengths for a plurality of scheduled packets. At time t' a new packet arrives and needs buffering since no wavelength is free (although wavelengths 1 and 6 have a void, these are not filled) Due to the granularity of an optical buffer, the packet can only be scheduled at times t-i-nG'. The minimal delay t-i-G' resulting in an allowable transmission is realized with the wavelengths 1, 2, and 5. Among these three, the one with the longest scheduling horizon is chosen. Thus, the packet is converted to wavelength 5. This choice results in the creation of the smallest void (among the wavelengths with minimal delay), which is beneficial because the excess utilization is minimized. Creating a void is bad because it is the creation of a period of time when the output will not be used while there are packets waiting to be sent. This is called excess utilization. It is to be noticed that converting the packet to wavelength 4 would result in an even smaller void, but this is not done because the delay is not minimal.

By way of illustration, two algorithms combining JLSQ. with optical buffering using a plurality of fibre delay lines as described in the above embodiments are discussed. In the algorithm, for each wavelength, there is an output scheduling horizon and a set of scheduling horizons being a scheduling horizon for each fibre delay line. The number of fibre delay lines used is referred to as N, the number of wavelengths used is referred to as W. Typically, in a memory following data may be stored: -length[i], 1 «= i «= N, being the length of each fibre delay line i, -delay[0, ii, 1 «= j «= W, being the output scheduling horizon for each wavelength j -delay[i, ii, 1 «= i «= N, 1 «= j «= W, being the scheduling horizon for each fibre delay line i for each wavelength j In a first exemplary algorithm that will be shown, priority is given to a good fibre delay line combination (shortest longest fibre delay line among the combinations with shortest delay). In the second exemplary algorithm that will be shown, priority is given to a good wavelength (the longest shortest queue = shortest void among the wavelengths with shortest delay).

A possible implementation of the first exemplary algorithm is shown in FIG. 13, whereas a flow chart of the first exemplary algorithm is shown in FIG. 14. By way of illustration, standard and optional features of the exemplary algorithm will be discussed in more detail.

In a first step, upon arrival of a packet to be routed, the output scheduling horizon for each wavelength, as well as the scheduling horizons per fibre delay line for each wavelength are updated. There are W output scheduling horizons, one for each wavelength and there are NxW fibre delay line scheduling horizons, one per fibre delay line per wavelength. So, in total, there are now (N+1)W scheduling horizons to be updated.

In a second step, for each of the wavelengths or until for one of the wavelengths the condition is met, it is checked whether or not the output sheduling horizon delay delay[O, w] is zero. If one of the wavelengths is free, i.e. the corresponding output scheduling horizon is zero, then the packet is converted to that wavelength and sent to the output. If more than one wavelength is free, the packet may be converted to the first free wavelength found.

In a third step, it is attempted to minimise the buffer time for the packet, in other words a solution is sought advantageously for a minimum delay, and if no solution can be found, for higher delay values. The delay D thereby starts at D = rmin{delay[o,w]} I G 1 * G and subsequent values for the delay D are calculated as adding the granularity to the current delay D + G. The attempt to minimise the buffer time for the packet then checks all possible combinations of the fibre delay lines whereby the total length equals the delay D, the delay lines here called i and j where the length of fibre delay line i is larger than the length of fibre delay line j, such that the total delay a packet experiences when it travels through both fibre delay lines equals D. According to the present embodiment, one starts with the "best" combination. In the case that only combinations with maximum two fibre delay lines are used, this best combination could be the one with the shortest longest fibre delay line.

A wavelength selection variable, in the present example referred to as "optw", is used for indicating the best wavelength so far, i.e. the one with the longest output horizon among the wavelengths with the shortest delay. This is according to the JLSQ method: Join the Longest Shortest Queue. It gets the initial value 0, which means no solution, since the wavelengths are numbered starting with one. If a contention free fibre delay line combination for a particular wavelength is found, the value stored in the variable optw will be different from 0.

The for loop checks the current fibre delay line combination for each wavelength. If the condition "delay[i,w]==0" is true, then the current combination of fibre delay lines (fibre delay line i and fibre delay line j) for the current wavelength (w) is contention free and it is checked if this wavelength is a better solution than the one stored in optw. There are two cases: (1) If the value of the variable "optw" is zero, then the current wavelength is the first solution and thus the only solution so far. Thus, it is stored in the variable optw".

(2) If the output scheduling horizon from the current wavelenght (w) is larger than the output scheduling horizon from the wavelength "optw" (delay[0, w] > delay[0, optw]), then the current wavelength offers a better solution. Thus, it is the best wavelength so far and stored in the variable optw".

If the value stored in "optw" is not zero, then a contention free solution has been found. The output scheduling horizon, the fibre delay line i scheduling horizon, and the fibre delay line j scheduling horizon for wavelength optw" are updated, the packet is converted to wavelength optw" and sent first through fibre delay line i, then throught fibre delay line j, and finally to the output.

In a further step, if no solution is found, other combinations are tried. If there are no more combinations that realize the current delay, higher delays are tried.

In a final step, either the packet is outputted if an appropriate delay is found or dropped if at last not one of the possible hardware combinations leads to a solution.

A possible implementation of the second exemplary algorithm is shown in FIG. 15, whereas a flow chart of the second exemplary algorithm is shown in FIG. 16. The algorithm comprises similar steps as discussed for the first exemplary algorithm discussed above, but instead of an optimal wavelength for a particular combination of fibre delay lines an optimal wavelength for a certain delay is sought. The algorithm therefore is adapted for finding the best wavelength for a group of combinations with the same delay.

The attempt to minimise the buffer time for the packet in the third step is adapted compared to the first exemplary algorithm. Although all possible combinations of the fibre delay lines whereby the total length equals the delay D is determined in a similar way as in the first algorithm and a variable "optw" is kept for indicating the wavelength corresponding to the best solution found, also two additional variables opti" and optj" are stored for identifying the corresponding individual delay lines obtained for the best solution identified. If a contention free wavelength is found for a set of delay lines, further checks will be performed to identify whether a solution (comprising a wavelength and a combination of fibre delay lines) resulting in a smaller void can be found.

Comparison between the two algorithms learns that the second algorithm puts more effort in reducing the void length. The latter can be seen when, e.g., the example of an optical buffer with 10 equidistant fibre delay lines having a granularity G = 1 is considered. If the minimum output horizon among the different wavelengths D equals 15, i.e. D = min{ delay[0,w], 1 «= w «= W} = 15, and if there exists a contention free wavelength for the best combination 8+7, then the first algorithm will choose the combination 8+7 with the best wavelength. The second algorithm will check if there exists a smaller void with the combinations 9+6 and 10+5.

Thus, the first algorithm will, in general, result in a better fibre delay line combination (a shorter longest FDL) and the second algorithm will result in a smaller void (a longer shortest queue). Both algorithms could be used according to embodiments of the present invention.

In one aspect, embodiments of the present invention also relate to computer-implemented methods for performing optical buffering or for controlling optical buffering according to embodiments of the invention as described above.

Embodiments of the present invention therefore also relate to corresponding computer program products. The methods may be implemented in a computing system. They may be implemented as software, as hardware or as a combination thereof. The hardware may be dedicated hardware. Such methods may be adapted for being performed on computer in an automated and/or automatic way. In case of implementation or partly implementation as software, such software may be adapted to run on suitable computer or computer platform, based on one or more processors.

The software may be adapted for use with any suitable operating system such as for example a Windows operating system or Linux operating system. The computing means may comprise a processing means or processor for processing data. The computing means also may be a controller as described in an aspect above.

According to some embodiments, the processing means or processor may be adapted for performing optical buffering of optical packets according to any of the methods as described above. The processor therefore may be adapted for controlling the use of a plurality of fibre delay lines in a particular manner as described above and/or for determining the used based on input regarding the optical packet stream, i.e. the length of the packet to be buffered, information regarding the previous and/or following packet. Besides a processor, the computing system furthermore may comprise a memory system including for example ROM or RAM, an output system such as for example a CD-rom or DVD drive or means for outputting information over a network. The memory or memory system may be configured to store length of delay lines, the scheduled output scheduling horizon and the scheduled scheduling horizon for the fibre delay lines.

Conventional computer components such as for example a keyboard, display, pointing device, input and output ports, etc also may be included. Data transport may be provided based on data busses. The memory of the computing system may comprise a set of instructions, which, when implemented on the computing system, result in implementation of part or all of the standard steps of the methods as set out above and optionally of the optional steps as set out above.

The foregoing description details certain embodiments of the invention. It will be appreciated, however, that no matter how detailed the foregoing appears in text, the invention may be practiced in many ways, and is therefore not limited to the embodiments disclosed. It should be noted that the use of particular terminology when describing certain features or aspects of the invention should not be taken to imply that the terminology is being re-defined herein to be restricted to include any specific characteristics of the features or aspects of the invention with which that terminology is associated.

Claims

Claims 1.-An optical buffer for buffering optical signals, the buffer comprising -a plurality of fibre delay lines of different lengths, and -a controller adapted for controlling the delay of packets in a packet stream, wherein the controller is adapted for delaying a packet by subsequently using different fibre delay lines and for using, for the delay of each packet, each fibre delay line at most once.
2.-A system according to claim 1, wherein the controller is adapted for using the fibre delay channels in a predetermined order.
3.-A system according to claim 2, wherein the controller is adapted for using the fibre delay lines in a decreasing order of length.
4.-A system according to any of the previous claims, wherein the controller comprises a memory and is adapted for storing the length of each of the plurality of fibre delay lines, the overall output scheduling horizon, and the output scheduling horizon for each of the plurality of fibre delay lines.
5.-A system according to any of the previous claims, wherein the controller is adapted for operating the optical buffer as feed forward buffer.
6.-A system according to any of the previous claims, wherein the controller comprises a processor for determining the set of different fibre delay lines subsequently used for delaying a packet before entering of the packet in the fibre delay lines.
7.-A system according to claim 6 wherein the processor is adapted for retrieving possible combinations of fibre delay lines for predetermined delays based on alook up table.
8.-A system according to any of the previous claims, wherein the controller is adapted for determining the combination of fibre delay lines realizing the required delay by taking into account all combinations of fibre delay lines realizing the delay and selecting that combination of fibre delay lines wherein the longest delay line of the combination is the shortest.
9.-A system according to any of the previous claims, wherein the system furthermore comprises a wavelength converter for converting an optical packet to one of a set of wavelengths distinguishably transmittable through the fibre delay line.
10.-A system according to any of the previous claims, wherein at least one fibre delay line can be used for distinguishably transmitting different wavelengths through the fibre delay line, and wherein the controller furthermore is adapted for using wavelength selection as well as fibre delay line selection for buffering subsequent packet in a packet stream.
11.-An optical data communication system, the optical data communication system comprising an optical buffering system according to any of claims 1 to 10.
12.-A method for optically buffering optical packets, the method comprising controlling the delay of a packet in an optical packet stream by subsequently using different fibre delay lines and by using, for the delay of each packet, each fibre delay line at most once.
13.-A method for optically buffering according to claim 12, the method comprising using the fibre delay channels in decreasing length order.
14.-A method for optically buffering according to any of claims 12 to 13, the method comprising -determining a delay value D by which the packet at least needs to be delayed, -determining a combination of fibre delay lines realizing the delay D by taking into account all combinations of fibre delay lines realizing the delay D, and selecting that combination of fibre delay lines wherein the longest delay line of the combination is the shortest.
15.-A method for optically buffering according to claim 14, wherein the method comprises, if the determined combination does not prevent contention, selecting a different combination allowing realizing the delay D.
16.-A method for optically buffering according to claim 15, wherein the method furthermore comprises, if no different combination can be found, updating the delay value D using the granularity of the set of delay lines, and determining a combination of fibre delay lines realizing the updated delay.17.-A method for optically buffering according to any of claims 12 to 16, wherein the method comprises scheduling the packet to the combination of fibre delay lines allowing obtaining a suitable delay.18.-A method according to any of claims 12 to 17, the method comprising storing the length of each of the plurality of fibre delay lines, the overall output scheduling horizon, and the output scheduling horizon for each of the plurality of fibre delay lines.19.-A method according to any of claims 12 to 18, the method operating an optical buffer as feed forward buffer.20.-A method according to any of claims 12 to 19, wherein the method comprises determining the set of different fibre delay lines subsequently to be used for delaying a packet before entering the packet in the fibre delay lines.21.-A method according to any of claims 12 to 20, the method comprising using a look up table for determining a combination of fibre delay lines to be used.22.-A method according to any of claims 12 to 21, the method comprising determining a combination of fibre delay lines and a wavelength for transmission of the packet in the combination of fibre delay lines.23.-A controller for an optical buffer, the controller being configured for delaying a packet by subsequently using different fibre delay lines and for using, for the delay of each packet, each fibre delay line at most once.24.-A controller according to claim 23, the controller being adapted for performing a method according to any of claims 12 to 22.25.-A computer program product for, when executed on a computer, performing a method for optically buffering optical packets according to any of claims 12 to 22.26.-A machine readable data storage device storing the computer program product of claim 25.27.-Transmission of the computer program product of claim 25 over a local or wide area telecommunications network.