Method of allocating bandwidth in an optical network

This invention is based on a priority application EP 12 305 052.8 which is hereby incorporated by reference.

The present invention relates to the field of transparent optical networks. More specifically, the present invention relates to a system and method of channel allocation in a transparent optical network.

The steady increase of capacity demand in transport core networks requires cost-effective solutions for increasing the amount of traffic that can be carried by fibre infrastructures which have already been deployed. Current network architectures based on a single grid wavelength allocation scheme (e.g. 50 GHz or 100 GHz grids as defined by the International Telecommunication Union's Telecommunications Standardisation Sector, or ITU-T) can result in congestion of certain links and blocking in parts of a network, while spectral bandwidth is still available in other parts of the network. In order to accommodate an ever-growing capacity demand in a fibre-limited environment, it is therefore necessary to push network capacity beyond that allowed by standard wavelength grids.

This problem is compounded by the desire to keep the cost and energy consumption of the network to a minimum. Increasing transmission via transparent optical data path generally reduces the amount of energy consumption needed for optical transmission in the optical layer but may cause increased energy consumption in other/higher layers of the network. That is to say that the higher the degree of transparency, the fewer optical-electrical-optical (OEO) conversions will be present in the network. Several methods and systems have been developed to attempt to address this problem.

One such method involves the use of a uniform wavelength grid with smaller-than-50 GHz spacing, as used for instance in submarine transmission links to increase their spectral efficiency. In meshed networks, this solution requires changing all wavelengths selective switches (WSS) to ensure compatibility with the new spacing. In addition, the smaller spaced grid causes increased physical impairments (e.g. cross-talk and cross-phase modulation, or XPM) and forces the use of numerous OEO regenerators even for demands passing through links which are far from congestion.

Another known method consists of dividing the spectrum of each link into two or more bands, with different channel spacings on each band. The main drawback of such a solution is its lack of flexibility, because the spectrum of all links has to be split in bands of the same width regardless of their level of congestion to avoid fragmentation problems. As is the case with the previous solution, the use of narrow spacing where not needed reduces the transparent reach and increases the number of required OEO. Moreover, this solution is not compliant with most current WSS architectures.

Another known approach is to divide the spectrum of the network into a transparent band and an opaque band and use as close a spacing as physically feasible (depending on link length) in the opaque band. The channel spacing (and hence link capacity) can thus be adjusted on a link-per-link basis, avoiding wavelength continuity issues while retaining some degree of transparency. This approach however, wastes transparent bandwidth in uncongested links and opacity requires a high number of OEO resources. This makes it difficult to control the increase in price/energy consumption per connection for a target capacity increase.

In order to solve the problems associated with the prior art, the present invention provides a method for allocating bandwidth in an optical network. The method comprises different steps.

A bandwidth grid containing first bandwidth slots of a lower bandwidth and higher bandwidth slots of a higher bandwidth, which is twice the lower bandwidth, is provided, wherein one of the higher bandwidth slots represents a same bandwidth on said bandwidth grid as two of the lower bandwidth slots.

Furthermore, a set of requests for respective data transmission connections with respective data rates and respective shortest path metrics is provided.

Even furthermore, shorter distances, over which a data signal may be transmitted transparently at the respective data rate within the lower bandwidth, and longer distances, over which a data signal may be transmitted transparently at the respective data rate within the higher bandwidth, are provided for the data transmission connections.

The method comprises furthermore the step of allocating bandwidth slots of the lower bandwidth for those of the data transmission connections, whose respective shortest path metric is smaller than or equal to their respective shorter distance.

Finally, bandwidth slots of the lower or the higher bandwidth are allocated for the further of the data transmission connections, whose respective shortest path metric is greater than their respective shorter distance, using respective first and second numbers of necessary optical-electrical-optical regenerations for transmitting data signals at the respective data rates via the further data transmission connections using the lower bandwidth or the higher bandwidth respectively.

As will be appreciated, the present invention provides several advantages over the prior art. For example, the present invention allows a network design tool to optimize the allocation of bandwidth slots of fixed wavelength grids on links in an optical network, such that transparent optical transmission paths are realized at least partially in order to keep the cost and energy consumption of the network at a minimized level, by taking into account the number of necessary optical-electrical-optical (OEO) conversions into account. Thus, it is achieved to carry more traffic than allowed by fixed wavelength grids.

The present invention proposes to pack transmission channels close together on a bandwidth grid, thus achieving the use of high spectral efficiencies.

Moreover, the present invention is also cost-efficient in that the amount of signals allocated onto slots of a bandwidth grid is controlled.

Furthermore, the complexity of the method of the present invention is low and it can be preferably implemented in, for example, a GMPLS control plane.

FIG. 1 is a diagram representing the relationship between 50 GHz and 25 GHz grids as defined by the International Telecommunication Union (ITU).

FIG. 2 is diagram representing a 25 GHz grid, and an example of allowed spectrum allocation for 50 GHz-wide channels, in accordance with the present invention.

FIG. 3 represents a diagram of a network for use with an embodiment of the present invention.

FIG. 4 is a table of links lengths in the network represented in FIG. 3 .

FIG. 5 is a table containing an example of traffic request information relating to the network of FIG. 3 .

FIG. 6 is a diagram representing the paths associated with the various traffic requests contained in the table of FIG. 5 .

FIG. 7 shows a flow chart of the proposed algorithm according to an embodiment.

FIGS. 8 to 12 show tables containing desired transmission together with respective values.

Specific non-limiting technical embodiments of the present invention will now be described.

FIG. 1 is a diagram representing the relationship between 50 GHz and 25 GHz grids as defined by the International Telecommunication Union (ITU) and FIG. 2 is diagram representing a 25 GHz grid, and an example of allowed spectrum allocation for 50 GHz-wide channels, in accordance with the present invention.

The bandwidth grid as defined by the ITU is such, that the centre frequency of an n-th 50 Ghz bandwidth slot corresponds to the centre frequency of an (2*n−1)-th 25 Ghz bandwidth slot.

Thus, a 50 Ghz bandwidth slot overlaps at least partially with three other 25 Ghz bandwidth slots.

The bandwidth grid as proposed herein is such, that a centre frequency of a 50 Ghz bandwidth slot lies between the centre frequencies of two other 25 Ghz bandwidth slots. Thus, a 50 Ghz bandwidth slot overlaps, preferably exactly, with two other 25 Ghz bandwidth slots.

FIG. 3 represents a diagram of a network for use with an embodiment of the present invention. The network shown in FIG. 3 is a Wavelength Division Multiplexed (WDM) network. The optical spectrum that can be used for data transmission on each link is limited and therefore bandwidth slots of this spectrum have to be allocated on the links for demanded transmission paths.

The network shown comprises a plurality of nodes A, B, C, D, F, G and H and a plurality of links there between A/B, A/C, B/C, B/F, C/D, F/G, D/F, D/H and G/H. Each node comprises a plurality of optical input ports, a plurality of optical output ports, and means capable of transferring an optical signal from its input to its output without conversion to an electrical signal, which is called a transparent transmission. Moreover, each node also comprises means capable of receiving an optical signal, converting the received optical signal into an electrical signal, thereby allowing the signal to be regenerated in the electrical domain, and converting the electrical signal back to an optical signal, before transmitting the optical signal through an output port, which is called an opaque transmission.

FIG. 4 is a table of the link lengths of the network of FIG. 4 . As will be appreciated, the values in this example have been chosen in order to simply the understanding of the present invention and do not have units. In practice, the units of most WDM networks would be in the region of kilometers, or hundreds of kilometers. The length of the span will affect the signal quality of a received signal. In general terms, the longer the ultimate path of a given optical signal, the poorer the quality of the received signal will be. Conversely, the shorter the path of the signal, the higher the quality of the received signal will be. Moreover, the modulation format that is used also affects the span that a data signal can travel without exceeding a maximum bit error ratio (BER) at the receiver. On may assume that a data signal with a predefined data rate that is transmitted within a first bandwidth slot of a first bandwidth by means of a predefined modulation format with a predefined symbol rate, this data signal will be able to travel over a span preferably called a first reach without OEO conversion and without exceeding a maximum bit error ratio at the receiver. This first reach is amongst other things limited due to the fact of cross-talk and cross-polarization between neighbouring bandwidth slots. If the same data signal with the same data rate is transmitted within a second bandwidth slot of a second bandwidth, greater than the first bandwidth, by means of the same general modulation format with the same predefined symbol rate, this data signal will be able to travel over a span preferably called a second reach without OEO conversion and without exceeding a maximum bit error ratio at the receiver, wherein this second reach is greater than the first reach.

Thus, by knowing the properties of the optical link over which the data signal is to be transmitted and the chosen modulation format, one may determine a first reach over which the signal may be transmitted transparently when using a first bandwidth slot of a first bandwidth, as well as a second bandwidth slot of a second bandwidth without OEO conversion and without exceeding a maximum bit error ratio at the receiver.

FIG. 5 represents an example of information relating to requested transmission paths or transmission connections in regard to the present embodiment. Each request d 1 , d 2 , d 3 , d 4 , d 5 , d 6 , denoted as d i with index i, comprises a demand, e.g. from node B to node H, denoted as “B→H”, a demanded data rate, a shortest path, e.g. from node B to node F to node D and then finally to node H, denoted as “B-F-D-H”, and a shortest path value L i , which represents the sum of the link lengths included in the shortest path. This non-limiting example has been chosen in order to provide a simple way of explaining the invention. Accordingly, the invention is not limited in any way to the specific features of the above traffic requests.

As mentioned above, a signal of a given data rate transmitted over a given bandwidth will be associated with a specific distance, called span or reach, over which that signal can travel transparently. For the purposes of simplicity, it is assumed that the reach over which an optical signal can be transmitted transparently travel while maintaining an acceptable error rate is determined primarily by it spectral efficiency, i.e. the ratio between the data rate and the bandwidth of the optical spectrum that the signal occupies. It is a well documented fact that the reach tends to decrease rapidly when the spectral efficiency increases.

Each node (transponder) can tune the amount of bandwidth allocated to the signal, carrying data at a constant data rate they transmit. These nodes are thus capable of trading transparent reach for spectrum efficiency or the other way around, according to which feature (cost-efficiency or network throughput) is most desirable.

In this example, 10 Gb/s non-return to zero (NRZ) channels can be multiplexed on 50 GHz grids or 25 GHz grid and 100 Gb/s traffic can be routed through Polarization-Division-Multiplexed Quaternary Phase Shift Keying at 28 Gbaud, requiring 50 GHz of spectrum, or through Polarization-Division-Multiplexed 16 Quadrature Amplitude Modulation at 14 Gbaud, requiring 25 GHz of spectrum.

Moreover, it is assumed that the reach of a 100 Gb/s signal transmitted over a 50 GHz channel is 10 units of length, preferably called Reach2 for 100 Gb/s, the reach of a 100 Gb/s signal transmitted over a 25 GHz channel is 5 units of length, preferably called Reach1 for 100 Gb/s; the reach of a 10 Gb/s signal transmitted over a 50 GHz channel is 15, preferably called Reach2 for 10 Gb/s, and the reach of a 10 Gb/s signal transmitted over a 25 GHz channel is 6, preferably called Reach1 for 10 Gb/s. Thus, for each demanded transmission connection d_i with index i respective reaches Reach1_i and Reach2_i can be determined depending on the demanded data rate of the respective connection.

FIG. 6 is a graphical representation of the shortest paths for the requested traffic connections of FIG. 5 applied to the network of FIG. 3 . As can be seen from FIG. 6 , not all links will need to support the same number of signals. For example, link B/F will need to carry three signals, link C/D will need to carry two signals and link D/H will need to carry four signals. In order to route this traffic demands with maximum efficiency, the method of the present invention will be used.

The demanded transmission connections d_i may be listed in a list of demands to be routed through the network as list D={d_i}. This list D contains the demanded transmission connections, as for example shown in the table of FIG. 5 .

FIG. 7 shows steps of the proposed algorithm A. The algorithm is provided with respective values Reach1_i and Reach2_i for each demand with index i. In a step S 1 , shortest-paths are determined for all demanded connections d_i. The resulting paths in this example are shown in FIG. 6 . Clearly, the link LK — 4 from the node D to the node H needs to carry the most demanded connections di, in this case four connections.

Coming back to FIG. 7 , bandwidth slots with a lower bandwidth BW_low, in this example BW_low=25 GHz, are allocated for those demanded connections, for which the corresponding shortest path value L_i is smaller than the corresponding smaller reach Reach1_i for a 25 Ghz slot. The allocation of these bandwidth slots may be carried out using messages of the Resource Reservation Protocol (RSVP). Thus, with these allocations transparent data transmission along the corresponding transmission connections is ensured.

In a step S 3 , it is determined whether an allocation of bandwidth slots with a larger bandwidth BW_high, in this example BW_high=50 GHz, is necessary for ensuring transparent data transmission along transmission connections, for which the corresponding shortest path value L_i is greater than the corresponding smaller reach Reach1_i. It may be the case, that, when trying to establish transparent data connections using bandwidth slots with a larger bandwidth BW_high, the overall bandwidth of a link LK_i is exceeded.

In this example, the link LK — 4 shall have a maximum bandwidth of 125 GHz. The desired transmission connections d1, d4, d5 and d6 that shall be routed via the link LK — 8 are shown in the table of FIG. 8 . It is clear, that when wishing to establish these transmission connections d1, d4, d5 and d6 as pure transparent connections without any OEO conversion, an overall bandwidth of 175 Ghz is needed. Since the link LK — 8 has a maximum bandwidth of 125 Ghz, this is not possible. If all these transmission connections d1, d4, d5 and d6 were established by means of 25 GHz slots only, then only 100 Ghz bandwidth would be used while a maximum of 4 OEO conversions would be necessary, which is not desirable.

Thus, in a step S 4 , it is determined that link LK_g with index g, for which the greatest number of desired data connections is blocked when trying to achieve pure transparent transmission without OEO conversions. In this case, this is the link LK — 8 with g=8 connecting the nodes D and H.

In a step S 5 , it is determined for each desired transmission connection d1, d4, d6, which cannot be established transparently with the smaller bandwidth BW_low via this link LK — 8, the number N1_i of necessary OEO conversions when using the smaller bandwidth BW_low and the number N2_i of necessary OEO conversions when using the larger bandwidth BW_high. In this example, the desired transmission connections d1, d4, d6 together with the necessary numbers of OEO conversions for BW_low=25 GHz and BW_high=50 GHz are shown in the table of FIG. 9 . Furthermore, for each of these desired transmission connections d1, d4, d6 the difference ΔN_i=N1_i−N2_i is determined, as shown in FIG. 9 .

The desired transmission connections d1, d4, d6, which cannot be established transparently with the smaller bandwidth BW_low via this link LK — 8, may then be sorted in an ascending order depending on their corresponding values of ΔN_i, as shown in the table of FIG. 10 .

In a step S 6 , a bandwidth slot of the lower bandwidth BW_low is then allocated on the examined link LK_g=LK — 8 for that transmission connection d1, which has the smallest values of ΔN_i, out of the remaining desired transmission connections d1, d4, d6. In this example, the transmission connections d1 and d6 both have a same smallest value of ΔN_i=+1, but only the connection d1 as one of these transmission connections d1 and d6 is chosen for allocating bandwidth slots of the lower bandwidth BW_low on the examined link LK_g=LK — 8.

The allocation of this bandwidth slot may be carried out using messages of the Resource Reservation Protocol (RSVP). Thus, in this example, the transmission connection d1 is allocated a bandwidth slot of BW_low=25 GHz, leading to the need of one OEO conversion for each of them. This leads to an allocation of 25 Ghz bandwidth for the connection d1, and thus to an intermediate overall allocation of 50 GHz on the link LK — 8. Since the link LK — 8 has a maximum bandwidth of 125 GHz, there is still a 75 GHz bandwidth available on the link LK_ 8 .

The remaining desired transmission connections, in this case the transmission connections d6 and d4, for which no bandwidth slots are yet allocated, may then be stored in a table shown in FIG. 11 .

In a step S 7 , a necessary allocation of bandwidth slots with a larger bandwidth BW_high is then determined for the remaining demanded transmission connections, when trying to achieve transparent data transmission, which in this example are the connections d6 and d4. If such an allocation is possible, this is carried out via the messages of the Resource Reservation Protocol (RSVP). An allocation of 50 GHz for each of the connections d6 and d4 would demand an allocation of 100 GHz for the connections d6 and d4. Since there is only still a bandwidth of 75 GHz bandwidth available on the link LK — 8, it is not possible to allocate a 50 GHz slot for each of the demanded connections d6 and d4.

In a step S 8 it is checked, if there are still links remaining in the network, for which desired transmission connections cannot be established as pure transparent connections. Such links may be called congested links. If this is not the case, the algorithm A ends. If this is the case, the algorithm returns to the step S 3 . Re-starting from the step S 3 , only links that have not yet been examined within the steps S 4 to S 7 will be considered. In this example, it is determined, that the link LK — 8 is still congested, since and allocation of a 50 GHz slot for each of the demanded connections d6 and d4 is not possible on the link LK — 8. Thus, in this example, the algorithm A returns to the step S 3 from where it starts a further iteration.

In this further iteration, the algorithm A reaches once again the step S 6 , in which in this example the demand d6 is identified as the demand with the smallest number of ΔN_i, as it can be seen in FIG. 11 . Thus, a bandwidth slot of the lower bandwidth BW_low is then allocated on the examined link LK_g=LK — 8 for that transmission connection d6. Thus, an intermediate overall allocation of 75 GHz is now given on the link LK — 8. Since the link LK — 8 has a maximum bandwidth of 125 GHz, there is still a 50 GHz bandwidth available on the link LK — 8. The remaining demanded transmission connection is the connection d4.

In a step S 7 , a necessary allocation of bandwidth slots with a larger bandwidth BW_high is then determined for the remaining demanded transmission connections when trying to achieve transparent data transmission, which in this example is the connection d4. If such an allocation is possible, this is carried out via the messages of the Resource Reservation Protocol (RSVP). An allocation of 50 GHz for each of the connection d4 is possible in this example, and thus carried out. This leads to an overall bandwidth allocation on the link LK — 8 of 125 GHz. Thus, it is achieved by the proposed method to maximise the allocated bandwidth on the link LK_g=LK — 8, while minimizing the number of necessary OEO conversions.

The table shown in FIG. 12 illustrates the chosen bandwidth allocation for each of the demanded transmission connections on the link LK — 8 together with the resulting number of OEO conversions for the given example.

In a step S 8 it is checked, if there are still links other than the examined link LK_g=LK — 8 remaining in the network, for which desired transmission connections cannot be established as pure transparent connections. Such links may be called congested links. If this is not the case, the algorithm A ends. If this is the case, the algorithm returns to the step S 3 . Re-starting from the step S 3 , only links that have not yet been examined within the steps S 4 to S 7 will be considered.