METHOD AND SYSTEM FOR MULTIPLEXING DATA FOR

TRANSMISSION

CROSS-REFERENCE TO RELATED APPLICATIONS

[01] The present application claims priority of US provisional patent application no. 60/869,966 filed on December 14, 2006, and of US provisional patent application no. 60/978,501 filed on October 9, 2007.

TECHNICAL FIELD

[02] The invention relates to digital communications networks, and more particularly concerns a method and system for multiplexing data packets over network communication links.

BACKGROUND

[03] Fibre optic networks were first designed to accommodate voice telephony circuits. Legacy protocols like SONET/SDH were instrumental for the deployment of fibre optic networks. These protocols possess multiple features for network management and for network protection, and offer guaranteed bandwidth because of their circuit-switched nature with static bandwidth allocation. Low rate traffic streams (e.g. STS-3) are time-domain multiplexed according to a digital signal hierarchy to form high rate optical signals to be transported over a fibre optic infrastructure (e.g. OC-48, OC-192, OC-768).

[04] The need to accommodate more and more data traffic in addition to voice traffic made it necessary to develop new protocols targeting an efficient use of installed fibre infrastructures. Data traffic being packet-based and relying on protocols like Ethernet and IP, first attempts to transport data traffic simply consisted in hardwiring packets to SONET/SDH circuits. Point-to-Point Protocols (PPP) over High-Level Data Link Control (HDLC) for Packet over SONET (PoS), and X.86 encapsulation for Ethernet over SONET/SDH (EoS) are examples of protocols

developed to allow transport of packet-based data traffic over SONET/SDH circuit- switched fibre optic infrastructures.

[05] Although deterministic performance could be achieved, hardwiring packets to SONET/SDH circuits resulted in rigid and inefficient point-to-point data transport. Because data traffic is usually made of bursts of data that need to be dynamically routed, static allocation of SONET/SDH bandwidth resulted in an inefficient use of the infrastructure. The next natural step was to evolve to Switched-Ethernet over SONET/SDH (S-EoS) where data packets destination addresses are processed at wire speed before being allocated to the appropriate SONET/SDH circuits. By doing so, efficient use of infrastructure bandwidth can be obtained because of the statistical multiplexing gain offered by wire speed packet switching.

[06] Since the advent of S-EoS, other attempts to increase the efficiency in fibre optic infrastructure bandwidth included the native transport of Ethernet data traffic where Ethernet packets are directly transmitted on an optical wavelength without any other protocol encapsulation. Although attractive from an efficiency perspective, native transport of Ethernet data traffic lacks the network management and network protection capabilities of SONET/SDH.

[07] To allow efficient use of the fibre optic infrastructure for data packet transport while keeping advantages associated with SONET/SDH technology, the Resilient Packet Ring (RPR) was developed. In addition to being a ring topology that inherently offers advantages for network protection, RPR can consider multiple traffic priority classes when time multiplexing data traffic flows, consequently approaching the ideal deterministic case of pure circuit switched technologies in terms of quality of service for the higher priority classes. With RPR, the whole bandwidth of a given wavelength can be used by multiple packet-based traffic flows using statistical multiplexing. RPR is designed to operate on single wavelength architectures.

[08] The growth of packet-based traffic with different quality of service (QoS) requirements is putting pressure on networking equipment. Wavelength division multiplexing (WDM) is used to meet the increasing demand in bandwidth by creating

as many independent channels as there are available WDM channels, i.e. WDM wavelengths.

[09] In fibre optic WDM multiple access networks, one or more WDM channel is typically dedicated to each customer for carrying data traffic associated with each customer in a static bandwidth allocation scheme. The given data bandwidth the customer pays for is thereby guaranteed. This wavelength dedication system is however inefficient under strong traffic dynamics. The total system bandwidth in such a WDM system is achieved only when all WDM channels are used at a maximum capacity, i.e. when all customers use their dedicated WDM channel(s) at a maximum capacity.

SUMMARY

[10] There is a need for a multiplexing method and system that breaks the static one-to-one relationship between deployed Wavelength Division Multiplexing (WDM) channels and customers in telecommunication networks, such as access networks over optical fibre

[11] There is provided a method and system for multiplexing data for transmission over a communication link having a plurality of communication channels. The proposed multiplexing method finds application in access networks over WDM optical fibre links. The multiplexing method provides a dynamic reconfiguration of the WDM channels such that one of or any combination of the WDM channels may be assigned to one client at any one given time. This allocation is dynamically reconfigured by the network as a function of a level of priority of the data packets to be transmitted. This allows multiple wavelengths of a WDM optical fibre link to be considered as a single high capacity unified channel. This technology can be used over ring, mesh, linear or point-to-multi point telecommunication infrastructures.

[12] The method and system described herein breaks the one-to-one relationship between the WDM channels and customers, which allows network operators to efficiently use the optical bandwidth. The network can also be

reconfigured to add customers to an existing network without upgrading the optical transport layer, and the network operators are allowed to grow their networks by gradually adding physical capacity when limits of the installed capacity are reached.

[13] According to one aspect, there is provided method for multiplexing data for transmission over a communication link having a plurality of communication channels. A cluster of data packets associated with a client is provided. The data packets each comprises user data to be transmitted to the client through the communication link, and each comprises control information indicative of a priority level associated thereto. A value of a number n of the channels to be used for transmitting the cluster is determined. The value is determined according to at least the priority level. N specific channels selected among the plurality of channels are assigned to the cluster. The cluster of data packets is split among the n specific channels for transmission over the communication link.

[14] According to another aspect, there is provided a method for multiplexing data for transmission over a communication link having a plurality of communication channels. A data packet comprising user data to be transmitted to a client through the communication link, and each comprising control information indicative of a priority level associated thereto is provided. A value of a number n of the channels to be used for transmitting the data packet is determined, the value being determined according to at least the priority level, n specific channels selected among the plurality of channels are assigned to the data packet. The data packet is split among the n specific channels for transmission over the communication link using the specific channels.

[15] According to still another aspect, there is provided a telecommunication interfacing module for multiplexing data for transmission over a communication link having a plurality of communication channels. The interfacing module comprises: a service interface for receiving a cluster of data packets comprising user data to be transmitted to a client through the communication link, and control information indicative of a priority level associated with the user data; an encapsulating unit for determining a value of a number n of the channels to be used for transmitting the

cluster of data packets according to at least the priority level, and for encapsulating the data packets of the cluster according to a protocol; a multiplexer for assigning to the cluster n specific channels selected among the plurality of channels and for splitting the data packets of the cluster among the n specific channels; and a transport layer interface for transmitting the split data packets on the n specific channels for transmission over the communication link.

[16] According to still another aspect, there is provided a software definable encoding unit for multiplexing data for transmission over a communication link having a plurality of communication channels. The multiplexing unit comprises: an input for receiving a cluster of data packets each comprising user data to be transmitted to a client through the communication link, and control information indicative of its priority level; an encapsulating unit for determining a value of a number n of the channels to be used for transmitting the cluster of data packets according to at least the priority level, and for encapsulating the data packets of the cluster according to a protocol; a multiplexer for assigning to the cluster n specific channels selected among the plurality of channels and for splitting the data packets of the cluster among the n specific channels; and an output for outputting the split data packets for transmission on the n specific channels over the communication link.

[17] According to still another aspect, there is provided a telecommunication system for serving a plurality of clients. The network comprises: a communication link having a plurality of communication channels, a central interfacing module and a node interfacing module. The central interfacing module is for multiplexing data traffic flows corresponding to the clients for transmission over the channels such that data traffic flow associated with at least one of the clients is split for transmission using any number n of the channels. The data traffic flows comprises data packets having a priority level. A value of the number n being determined according to the priority level. The node interfacing module is for demultiplexing data traffic flow associated with the clients after transmission over the communication link and for switching to the each of the clients data traffic flow corresponding thereto.

[18] According to still another aspect, there is provided a method for multiplexing data for transmission over a communication link having a plurality of communication channels, in which a cluster of data packets associated with a client is provided, the data packets each comprising user data to be transmitted to a client through the communication link, and each comprising control information indicative of its priority level, characterized in that : a number of the channels to be used being determined according to at least the priority level, the number of any of the channels among available ones of the channels is assigned to the cluster and the cluster is split among the assigned channels for transmission over the communication link, the number of the channels to be used for transmitting the cluster being determined as a function of the priority level.

[19] According to still another aspect, there is provided a method for multiplexing data for transmission over an optical fibre link having a communication channel corresponding to a first wavelength, in which clusters of data packets associated with clients are provided, the data packets each comprising user data to be transmitted to a client through the optical fibre link, and each comprising control information indicative of its priority level, traffic flows associated with the clients being multiplexed in the communication channel, the improvement comprising : providing a plurality of Wavelength Division Multiplexed (WDM) communication channels in the optical fibre link for providing service to a plurality of clients; determining a value of a number n of the WDM channels to be used for transmitting a cluster of data packets associated with one client, the value being determined according to at least the priority level; assigning to the cluster n specific WDM channels among the plurality of WDM channels; and splitting the cluster among the n specific WDM channels for transmission over the optical fibre link.

[20] According to still another aspect, there is provided a method for multiplexing data for transmission over an optical fibre link having a plurality of Wavelength Division Multiplexed (WDM) channels, in which each of the WDM channels is being statically allocated to a single client, a cluster of data packets associated with one client is provided, the data packets each comprising user data to be transmitted to a

client through the communication link, and each comprising control information indicative of its priority level, the cluster being transmitted to the client using its allocated WDM channel(s), the improvement comprising : determining a value of a number n of the WDM channels to be used for transmitting a cluster of data packets associated with one client, the value being determined according to at least the priority level; assigning to the cluster n specific WDM channels selected among the plurality of WDM channels; and splitting the cluster among the n specific WDM channels for transmission over the optical fibre link.

BRIEF DESCRIPTION OF THE DRAWINGS

[21] Fig. 1 is a block diagram showing an example application of Software- Definable Wavelength Agile Networking (SD-WAN) in an access network using a ring topology;

[22] Fig. 2 is a graph illustrating data multiplexing by comparing three different multiplexing schemes, wherein section A illustrates a Coarse Wavelength Division Multiplexing (CWDM) scheme in which each CWDM channel (i.e. each wavelength) is allocated to a different client, section B illustrates a single wavelength communication link using time division multiplexing, and section C illustrates a SD-WAN multiplexing scheme on a CWDM communication link;

[23] Fig. 3 is a block diagram illustrating the main components of an example SD-WAN interfacing module to be used in a SD-WAN application such as the application illustrated in Fig. 1 ;

[24] Fig. 4 is a block diagram illustrating the SD-WAN interfacing module of Fig. 3, wherein the SD-WAN encoding/decoding unit is shown in more detail;

[25] Fig. 5 is a flow chart illustrating a method for multiplexing data for transmission over a SD-WAN link;

[26] Fig. 6 is a block diagram illustrating an example implementation of the SD- WAN encapsulating unit and the decapsulating unit of the SD-WAN interfacing module of Fig. 4;

[27] Fig. 7 is a block diagram illustrating an example implementation of the SD- WAN MUX and the optical receiving unit interface of the SD-WAN interfacing module of Fig. 4;

[28] Fig. 8 is a block diagram illustrating an example implementation the SD- WAN interfacing module of Fig. 4;

[29] Fig. 9 is a block diagram illustrating another example of the SD-WAN interfacing module to be used in a point-to-point topology;

[30] Fig. 10 is a graph showing a comparison of the SD-WAN technology with other technologies in terms of the number of served customers as a function of the total installed capacity, wherein the squares curve is for a SD-WAN link, the diamonds curve is for a 10 Gigabit Ethernet link; the triangles curve is for a Resilient Packet Ring (RPR) link; the cross curve is for a Switched-Ethernet over SONET link (S-Eo); and the stars curve is for an Ethernet over SONET link;

[31] Fig. 11 is a graph showing SD-WAN statistical multiplexing gain over RPR (triangles curve) and 10 Gigabit Ethernet (squares curve) technologies as a function of the grade of service;

[32] Fig. 12 is a graph showing SD-WAN statistical multiplexing gain over 10 Gigabit Ethernet technology as a function of the maximum bandwidth capacity offered per customer, wherein the stars curve is for a 40 Gb/s infrastructure and the dots curve is for a 20 Gb/s infrastructure;

[33] Fig. 13 is a block diagram illustrating an example of a Multiple Systems Operator (MSO) access point-to-point topology which uses SD-WAN technology;

[34] Fig. 14 is a block diagram illustrating another example of a MSO access topology which uses SD-WAN technology and wherein the SD-WAN interface module is located in the premises of one of the commercial customers; and

[35] Fig. 15 is a block diagram illustrating still another example of a MSO access topology which uses SD-WAN technology, wherein ring is formed with multiple client nodes;

[36] It will be noted that throughout the appended drawings, like features are identified by like reference numerals.

DETAILED DESCRIPTION

[37] The proposed data multiplexing method will be referred to herein as the Software-Definable Wavelength Agile Networking (SD-WAN) technology. While the SD-WAN technology finds herein an application in access networks over a WDM optical transport layer, the present description is not limited to this application. The technology may also be used in other applications using optical fibre or not. For example, the plurality of channels that are dynamically reconfigured may be a plurality of frequency channels in a wireless application, a plurality of physical cables, such as a plurality of optical fibres, a plurality of coaxial cables, a plurality of Code Division Multiple Access (CDMA) channels, etc. Furthermore the SD-WAN technology is herein described for application in access networks over a Wavelength Division Multiplexing (WDM) optical transport layer and typically operated by Multiple System Operators. It is however noted that the SD-WAN technology could also find applications in other levels of a network, such as in the metro edge or the metro core, and in networks operated by Telco operators for example.

[38] Now referring to Fig. 1 , there is shown a ring access network 10 using SD- WAN technology and which constitutes an example application of the proposed technology in an access network using a ring topology. The ring access network 10 is composed of multiple network nodes including a SD-WAN master node 12 and multiple SD-WAN client nodes 14. Each node 12 and 14 has two bidirectional

transport ports 15, i.e. XENPAK transceiver modules, for connecting the nodes in a ring configuration using a SD-WAN link 16. The SD-WAN link 16 generally consists of two optical fibre rings for carrying Coarse WDM (CWDM) channels, one optical fibre link for each direction of the ring, clockwise and counter-clockwise. Each node 12 and 14 further has a service port 17 for connection to a client service node or Customer Premise Equipment (CPE) 19 to be served using a client service link 18, or for connection to another network such as another area network. In this case, the SD- WAN master node 12 is connected upstream to a Metro Area Network (MAN) 20, through a MAN node 22 pertaining to the metro area network 20. Connection between the metro area network 20 and the ring access network ring 10 is made through an interconnection service link 24 connected to the service port 17 of the SD-WAN master node 12. Typically, the client service link 18 and the interconnection service link 24 are Ethernet links according to the IEEE 802.3 standard. Data traffic between the metro area network 20 and the SD-WAN master node 12, data traffic between the client nodes 14, and data traffic between a client node 14 and its associated client are bidirectional.

[39] Each node 12 or 14 receives data traffic from its two transport ports 15 and from its service port 17. Data traffic consists of data packets. Packets incoming to a node 12 or 14 are switched such that they are either dropped at the node if their destination address matches the address of the node or are forwarded to the next node (transit) through the transport port 15. Local routing tables are generated and stored in each of the nodes 12 and 14. Each local routing table identifies the destination address of a received data packet and selects the proper address to use for sending a packet through a given output port 15 or 17.

[40] The master node 12 also has the responsibility of management network control processes such as topology discovery, fairness, protection, etc. Each client node 14 also supports management network control processes.

[41] The main embodiment described herein is designed in conformity with the IEEE 802.3 and 802.17 standards and as such, includes functionalities such as Media Access Control (MAC), priority management, dynamic bandwidth allocation,

protection mechanisms, etc. Standard mechanisms associated with the Resilient Packet Ring (RPR) technology are also included but will not be described herein.

[42] According to the IEEE 802.3 standard, a priority level is attributed to each incoming data packet, based on Class of Service, Quality of Service and Service Level Agreements.

[43] As will be described in more detail hereinbelow, a priority level is assigned to each data packet when it first enters the ring access network 10, i.e. at its originating node. Dynamic management of active data packet addresses is also done at the originating node. Other management algorithms, such as dynamic bandwidth allocation and protection mechanisms, can be implemented at each node 12 and 14. The SD-WAN protocol includes the physical layer management and the data link/Media Access Control (MAC) layer managements and provides maximum optical layer flexibility via an optical-electronic-optical switching configuration.

[44] It is noted that in other embodiments, the network architecture can take on various other forms, i.e. point-to-point, daisy chain, point-to-multipoint, mesh, etc. The ring architecture shown herein is used for illustration.

[45] Fig. 2 illustrates the SD-WAN multiplexing method by comparing three different multiplexing schemes by comparing the timing of the multiplexed data packets as they are transmitted. On every sections A, B and C of Fig. 2, the time line goes on from left to right.

[46] Section A of Fig. 2 illustrates a CWDM scheme using four independent CWDM channels at 2.5 Gbps, in which each CWDM channel (i.e. each wavelength) is allocated to a different client. Packets pio, Pn, P ₁₂ and p ₁₃ are transmitted on channel 1 and are drawn with the same type of filling to shown that they are associated with the client 1. Packet p ₁₃ is a low priority jumbo packet. Packets p ₂ o, P ₂₁ , P ₂₂ and p ₂₃ are transmitted on channel 2 and are all associated with another client, i.e. client 2. Packets p ₃₀ , P31 and p ₃₂ are transmitted on channel 3 and are all associated with still another client, i.e. client 3. Packets p ₄₀ and p ₄ i are transmitted on channel 4 and are

all associated with still another client, i.e. client 4. Packet p ₁₀ and p ₄₁ are high priority packets.

[47] Section B of Fig. 2 illustrates a single wavelength high capacity channel at 10 Gbps used for transmitting the data packets shown in section A on a single high capacity channel using time division multiplexing. All packets transmitted on the four independent CWDM channels of section A are time division multiplexed for transmission on the single channel of section B. The arrows from section A to section B show the delay incurred by the data packets pi ₂ , P ₂₂ and p ₄ i due to the time division multiplexing.

[48] Referring to section B of Fig. 2, it can be seen that when multiplexing data from four independent channels (section A) to a single high capacity channel (section B), a low priority jumbo packet such as packet pi ₃ can jam the link for the duration of the transmission of the low priority jumbo packet, thereby pushing back the transmission of a high priority packet, p ₄ i for instance, received on another independent channel, until the end of the transmission of the low priority jumbo packet

[49] Section C of Fig. 2 illustrates a SD-WAN multiplexing scheme on four CWDM channels operating at 2.5 Gbps. All packets transmitted on the four independent CWDM channels of section A are multiplexed according to the SD-WAN scheme on four CWDM channels. According to the SD-WAN multiplexing scheme, any packet can be transmitted over any one or any combination of the four CWDM channels using a dynamic reconfiguration. According to this SD-WAN protocol, higher priority data packets or clusters, such as Pi ₀ or P ₄₁ , can be divided for transmission over a large number of CWDM channels, thereby increasing the effective bandwidth in the case of high priority packets. Low priority packets, such as P- ₁₃ and P ₂₂ , are rather transmitted using a low number (typically a single) CWDM channel. The other channels thus remain available for transmission of higher priority packets, such as P ₄₁ in the illustrated example. This scheme reduces the latency for transmission of high priority packets. The number of channels to be used for transmission of a data packet or cluster is made as a function of the priority level of the data.

[50] The implementation of the SD-WAN multiplexing method is made by encapsulating each data packet incoming to a service port of a node with header information that depends on a priority level associated with the data packet and which is typically determined according to its class of service and service level agreements (quality of service, etc.). The node then decides on which set of wavelengths from the SD-WAN link to send the encapsulated packet. The choice is made dynamically and depends on the traffic load processed by the node at that moment. In other words, the number of wavelengths selected for packet encoding is reconfigurable and is software definable for each packet. Higher priority packets will be given priority over lower priority packets.

[51] A data packet to be transmitted may be broken into parts and transported over multiple wavelengths at once. As a consequence, any residual capacity of a wavelength can be used to transport packets, resulting in a better statistical multiplexing gain than with any technology operating on a per wavelength basis. The statistical multiplexing gain will be further discussed herein below. It is noted that each client has access to full network capacity in the case of low traffic associated with other clients. The SD-WAN multiplexing thus provides a high bursting capacity.

[52] When more capacity is required, additional channels may be added. The additional capacity is consequently shared among the customers. Furthermore, this configuration allows the registering of a new customer without modification at the optical transport layer.

[53] It is also noted that the non-predictable nature of the SD-WAN multiplexing technique provides an additional level of encryption which improves data transport security. It is thus more difficult for hackers to break in and get unauthorized access to information. Data information is transported over multiple wavelengths following various network parameters like instantaneous network traffic loads. The data is transmitted following an unpredictable wavelength spreading and hopping pattern which requires more complex systems to hack and recover data than in a standard WDM multiplexing network. Only authorized SD-WAN network users should be able to decrypt and reconstruct the data packets.

[54] Fig. 3 shows the main components of an example SD-WAN interfacing module 100 to be used in a SD-WAN application. In the access network of Fig. 1 , each SD-WAN master node 12 and each SD-WAN client node 14 include one SD- WAN interfacing module 100. The SD-WAN interfacing module 100 is designed to be used in a ring configuration and therefore has two optical transport layer interfaces 600a and 600b, i.e. one for each side of the ring, for transmission and reception of SD-WAN multiplexed data on the SD-WAN link 16, on both sides of the ring access network 10. The SD-WAN interfacing module 100 also has a service interface 400 for inputting/outputting data in/from the access network and from/to a client service node or CPE 19 or from/to a network node in another network level, e.g. a MAN node 22.

[55] The SD-WAN Interfacing module 100 further has a SD-WAN encoding/decoding module 200 which cross-connects data packets from/to the transport layer interfaces 600a and 600b and the service interface 400 and multiplexes/demultiplexes data for transmission on the SD-WAN link according to the SD-WAN multiplexing technology described with reference to Fig. 2. The SD-WAN encoding/decoding unit 200 will be described in more detail with reference to Fig. 4.

[56] In this embodiment, the SD-WAN link 16 consists of a pair of optical fibre links carrying CWDM channels, one for each direction on the ring. On the receiver side, the SD-WAN interfacing module 100 receives the optical signals incoming from the SD-WAN link 16 and converts them into electrical signals for the SD-WAN encoding/decoding unit 200. On the transceiver side, the SD-WAN interfacing module 100 converts the electrical signals coming from the SD-WAN encoding/decoding unit 200 into optical signals for sending them over the SD-WAN link. Each optical transport layer interface 600a and 600b has a transport port output 15O connected to one optical fibre of the SD-WAN link 16 for sending SD-WAN multiplexed data and a transport port input 15I connected to the other optical fibre of the SD-WAN link 16 for receiving SD-WAN multiplexed data.

[57] In this case, the transport layer interfaces 600a and 600b consist of LX4 XENPAK four-WDM-channel transceiver modules. Each XENPAK module comprises multiple wavelengths optical sources 604 and receivers 602, an optical WDM

multiplexer 606 and a demultiplexer 608 for injection to and receiving from the WDM optical fibre link. Each LX4 XENPAK module supports four independent wavelengths modulated at 3.125 Gb/s on the optical side, i.e. the SD-WAN link, and 10 Gigabit Attachment Unit Interface (XAUI) on the electrical side, i.e. to/from the SD-WAN encoding/decoding unit 400. An 8b/10b encoding scheme is used on the optical signal, so that a 3.125 Gb/s rate is required for transmission at 2.5 Gb/s. In order to bypass the XAUI Synchronization Mode and its built-in pattern detector, word aligner, and XAUI state machines, an adaptation layer was designed to allow independent operation of each wavelength which leads to an additional throughput penalty of 5% for larger packets. It is noted that while the XENPAK transceiver is used in this embodiment, other transceiver modules may used as well. For example, single wavelength independent modules like CWDM small form-factor pluggable modules, or CWDM 10 Gigabit small form factor pluggable modules, can be used. Of course, the number of total WDM channels may vary.

[58] The Service interface 400 comprises transceivers for connection to the service link 18 through the service port 17. The Service interface 400 receives data packets to be transmitted on the access network on the service link 18 and sends data packet that are dropped from the access network on the service link 18. In this embodiment, the service port 17 is an eight-channel Ethernet protocol port as defined by the IEEE 802.3 standard.

[59] Now referring to Fig. 4, the SD-WAN encoding/decoding unit 200 of the SD- WAN interfacing module 100 of Fig. 3 is shown in more detail. The SD-WAN encoding/decoding unit 200 comprises a SD-WAN packet encapsulating unit 210, a decapsulating unit 250, a node cross-connect 220, two SD-WAN multiplexers 230a and 230b and two optical receiving units 240a and 240b, i.e. one for each side of the access network ring, and a medium access control unit 260. The letter indexes "a" and "b" are respectively used to designate components assigned to sides "a" and "b" of the SD-WAN interfacing module 100 for connection to both sides of the access network ring.

[60] The SD-WAN packet encapsulating unit 210 receives, from the service interface 400, data packets to be transmitted on the access network. The data packets are typically received as clusters of data packet. The WAN packet encapsulating unit 210 encapsulates the received data packets of cluster for transmission according to SD-WAN multiplexing. As a function of the priority level of the received data packets, each data packet is encapsulated in one packet for transmission of the data packets of a same cluster using a number of WDM channels. Each data packet may also be further split apart and encapsulated in a number of sub-packets such that the sub-packets can be later transmitted in parallel on the same number of WDM channels. The number of WDM channels to be used for the transmission of a cluster or of a single data packet is determined as a function of the priority level which takes into account the service level agreement, the traffic type and its related class of service.

[61] The format of the received data packets conforms to the Ethernet standard. Each data packet thus includes user data to be transmitted to one or more client nodes, and a header comprising control information. Among others, the control information includes data indicative of the priority level associated with the data packet, such as the class of service. While the priority level may simply be based on the class of service, in this embodiment it is determined by a combination of priority level factors also including the quality of service and the service level agreement. These last factors are part of the network configuration and not included in the data packet header. They are rather stored in configuration tables. The Virtual Local Access Networking ID (VLAN ID) of the data packet, which is found in its header, is used in combination of configuration tables to determine those other priority level factors. The SD-WAN encapsulating unit 210 thus reads the control information in the data packets including the destination address, the VLAN ID and the class of service and evaluates their destination using the address and their associated service level agreement, traffic type and class of service using the configuration tables. It then determines the priority level to be associated with each data packet based on those factors. The data packets are then encapsulated with an SD-WAN internal header. A data packet can be encapsulated in one encapsulated packet or can be split in a

plurality of encapsulated packets for later SD-WAN multiplexing. The number of encapsulated packets in which a data packet or cluster in split depends on the priority level determined by the service level agreement, the traffic type and the class of service. The SD-WAN internal header includes SD-WAN management restrictions associated with the data packets that are later used in SD-WAN multiplexing and which indicate the number of WDM channels that should be used to multiplex the encapsulated packets.

[62] In this embodiment, the SD-WAN network uses Virtual Local Access Networking (VLAN) to achieve tunnelling of the data to be transported. Various types of virtual private networks, such as VLAN ₁ Virtual private LAN service, pseudo-wire and multiprotocol label switching, can be used to separate the traffic of different users over an SD-WAN network. For a given SD-WAN network, the Service packet to be transported is mapped in the appropriate VLAN tunnel. All the VLANs that are defined, are listed in a VLAN control table. Each packet coming from a client interface has its VLAN ID (12 bits) and its Class of Service ID (3 bits). A packet without an ID can be rejected or have default values associated to the service interface following the network parameters.

[63] In this embodiment, two headers, i.e. an optical header and an internal header, are added to the received Ethernet data packets by the encapsulating unit 210 in the encapsulation process. The optical header is used to encapsulate the received Ethernet data packets in order to transport them on the access network. The optical header is used in the optical layer of the system. The internal header is rather used to support the data packet processing within the SD-WAN interfacing module 100.

[64] The optical header contains various fields including the following: Control, Destination Node ID, Source Node ID, Time to Live, i.e. a hop count to destination, Packet Number and Splitting Ratio. The optical header should be distinguished from the optical trailer which is added by the SD-WAN multiplexer 240, just before optical transmission and which generally contains a Terminating Packet Field and a Cyclic Redundancy Check (CRC) field which is used as an error detecting code.

[65] In the optical header, the Control field comprises various control bits including: Identifier for Protocol Update, Class of Service, VLAN ID, Resent Packet Tag, Wrapped Packet Tag, Message Type. The Message Type field specifies the type of payload content. The Resent and Wrapped Tags identify if a packet has been resent or wrapped.

[66] The internal header is inserted in addition to the optical header and is used to facilitate data packet processing within the SD-WAN interfacing module 100 and contains various fields including: Packet Size, VLAN ID, Optical Ring Side (forced side under protection), and a Time Stamp calculated relative to the Ethernet arrival time from the service interface. The internal header also includes SD-WAN management restrictions which include the following fields: Wavelength ID, and Maximum Burst ID. The Wavelength ID indicates the WDM channel associated to the encapsulated packet if the channel should be forced. The Maximum Burst ID indicates whether any WDM channel can be used for transmission and whether the encapsulated data packet can be split upstream for transmission.

[67] The encapsulated packets are transmitted to the node cross-connect 220 which switches the encapsulated data packets toward the proper side of the access network ring. The node cross-connect 220 also receives data from both sides of the network ring and in transit to the other side, and data received from one side and to be dropped to service. The node cross-connect 220 also reads the destination address of the encapsulated packets received and connects the received packets to the proper side of the network based on the shortest path and determined using network configuration tables. The node cross-connect 220 queues and switches all encapsulated packet received. Packets with highest priority are always processed first. When two or more packets have the same priority, the Time Stamp is used to process the oldest packets first.

[68] Encapsulated packets to be transmitted on the network ring on either side of the SD-WAN interfacing module 100 are received from the node cross-connect 220 by the proper SD-WAN multiplexer 230a or 230b. The SD-WAN multiplexers 230a and 230b switch the incoming encapsulated packets to the appropriate WDM channel

using the wavelength ID. The SD-WAN multiplexers 230a and 230b use a control table which is read, along with SD-WAN management restrictions inserted in the SD- WAN internal header by the packet encapsulating unit 210, for determining on which WDM channel a packet should go. The SD-WAN multiplexer 230a, 230b may either force a packet to a given WDM channel or operate in real time mode wherein a packet is cross-connected to the emptiest channel. The Maximum Burst field indicates if the packet can use any wavelength such that it is to be sent to the first available channel. It is noted that under given conditions, encapsulated packets may be split and re-encapsulated by the SD-WAN multiplexer 230a or 230b for transmission over multiple WDM channels. These conditions are determined by the packet size, VLAN ID and CoS ID, in combination with control tables.

[69] The SD-WAN multiplexer 230a or 230b also takes off the SD-WAN internal header of the packet and adds the optical trailer to the packet.

[70] At the output of the SD-WAN multiplexer 230a or 230b, data packets are arranged in four channels which correspond to the four WDM channels on which the data packets will be transmitted by the transport layer interface 600a or 600b, or XENPAK. Accordingly, the SD-WAN multiplexer 230a or 230b also includes an adaptation layer unit to format the encapsulated packets in a format compatible with the transponders of the XENPAK.

[71] The electrical signal which is provided by the SD-WAN multiplexer 230a or 230b and exits the SD-WAN encoding/decoding unit 100 to the transport layer interface 600a or 600b is a 10 Gigabit Attachment Unit Interface (XAUI) electrical signal. This electrical signal is converted into four-wavelength WDM optical signals by the transport layer interface 600a or 600b, for transmission over the SD-WAN link.

[72] Similarly, optical signals received by the transport layer interface 600a or 600b from other nodes and through the SD-WAN link are detected and converted into a 10 Gigabit Attachment Unit Interface (XAUI) electrical signal to be used by the SD- WAN encoding/decoding unit 200. The electrical output of the transport layer interface 600a or 600b is connected to the SD-WAN encoding/decoding unit 200 and is

received by the corresponding optical receiving unit 240a or 240b. The optical receiving unit 240a or 240b reads the destination address of each encapsulated packet received and determines the target node of the encapsulated packet according to network configuration tables. Packets to be dropped to the current node are seperated from packets to transit to the other side of the node from transmission along the access network ring. Packets to be dropped and packets to transit are provided to the node cross-connect 220 on different ports. The node cross-connect 220 connects encapsulated packets to be dropped toward the service interface and encapsulated packets to transit toward the transport layer interface 600a or 600b on the opposite side.

[73] At the output of the node cross-connect 220, encapsulated packets to be dropped are switched to the decapsulating unit 250. The decapsulating unit 250 removes SD-WAN encapsulation and reconstructs data packets that were split for SD-WAN multiplexing. The decapsulating unit 250 thus restores the data packets in the Ethernet format for transmission over the service link.

[74] The service interface 400 receives the Ethernet format data packets and transmits them over the service link according to the Ethernet protocol. The service link has eight different Ethernet service channels. The service interface 400 transfers the data packets to the appropriate Ethernet service channel.

[75] The SD-WAN encoding/decoding unit 200 also has a medium access control unit 260 for monitoring and packet inspection to support Medium Access Control (MAC) mechanisms and different control planes from upper protocol layers. The unit dynamically reconfigures the cross-connecting parameters according to the MAC mechanisms such as protection, fairness, topology and operations- administration-maintenance, provided by IEEE standards. The unit 260 also has the capability to send and receive SD-WAN control frames. The SD-WAN control frames are data frames sent by one MAC unit to another MAC unit to exchange data related to MAC mechanisms. This communication channel is created using a VLAN to broadcast over the network to exchange. The SD-WAN encoding/decoding unit 200 also has queuing capabilities. It is noted that the standard mechanisms provided by

the IEEE standards on Ethernet and optical fibre communication, such as VLAN, CoS, MAC-in-MAC and Q-in-Q are typically supported by the SD-WAN interfacing module 100.

[76] The medium access control unit 260 of the master node provides the VLAN control table which lists the members (clients) of each VLAN in terms of node and client interface destination. The SD-WAN master node is in charge of updating the VLAN control table of each SD-WAN client node. The VLAN control table is based on programmed parameters considering service requirements. According to existing standards, there are typically 4096 possible VLANs and 8 possible CoSs. For each of the 4096 VLANs listed in the first column of the table, a list of each member (SD-WAN Station ID and Client Service Port ID) of the VLAN is provided in the following columns. When a data packet arrives with an associate VLAN ID, the table provides the network parameters, i.e. the SD-WAN Station ID and Client Service Port ID, to reach the others members of this VLAN.

[77] The medium access control unit 260 also comprises the bandwidth allocation control table which provides the data rates and the wavelengths to be used for the transmission of each packet following its VLAN ID and its CoS ID. The bandwidth allocation control table is based on programmed parameters following service requirements and existing network conditions such as congestion, link failure and statistics. For each of the 4096 VLANs and each of the 8 CoSs listed in the first two columns, a list of bandwidth parameters are given in the following columns. When a data packet arrives with an associated VLAN ID and CoS ID, the bandwidth allocation control table provides the network parameters in term of priority, wavelength splitting ratio, minimum/maximum bandwidth allowed, optical transmission side and the list of available wavelengths.

[78] The medium access control unit 260 of a client node supports a delivery control table which corresponds to the client node version of the VLAN Control Table. Each node (SD-WAN Station ID) has its specific delivery control table which is constructed from the VLAN control table. The delivery control table identifies which Client Service Port IDs are allowed to receive data packets incoming to the client

node. For each of the 4096 VLANs, a list of Client Service Port ID's which are associated to the VLAN are given.

[79] Finally, the topology table provides the entire topology map (in hop count) by giving the network position of each node, identified by its SD-WAN node ID, in the network. For each SD-WAN node listed, all SD-WAN node ID's are given. The order of the list follows the clockwise ring direction.

[80] Each node of the access network ring includes one SD-WAN interfacing module 100. Each SD-WAN interfacing module 100 can be set to operate either as a SD-WAN master node 12 or as a SD-WAN client node 14. In an access network ring, one SD-WAN interfacing module 100 is set as a master node while all others are set as client nodes. The SD-WAN interfacing module 100 set as a SD-WAN master node 12 controls the MAC mechanisms.

[81] The access network ring is configured by the master node 12 which assigns a node ID to each client node 14 by broadcasting a message. During a ring topology discovery process, each client node 14 receives, in turn, the message packet initiated by the master node 12. Each client node 14 appends its unique ID to the message and forwards it to the next node in order to map the relative position of each client node 12 and to determine the ring topology. The master node 12 receives the ring topology in the message which comes back, and saves this information which is also broadcasted to all the client nodes 14.

[82] Fig. 5 shows the SD-WAN multiplexing method used for multiplexing data for transmission over a communication link, e.g. the SD-WAN link, having a plurality of communication channels, e.g. WDM channels. It is noted that SD-WAN multiplexing may be performed by either splitting clusters of data packets among a number n of channels for transmission, or by splitting large unitary data packets into the number n of channels. In the former case, each split sub-packet is an integral data packet, while in the latter case, each large data packet is split in a number n of sub- packets for transmission. The sub-packets are then recombined after transmission. In the embodiment described with reference to Fig. 4, both types of splitting are

possible, i.e. clusters can be split and large data packets can also be split. A large data packet within a cluster that is split can also be split.

[83] In step 501 , a data packet or a cluster of data packets is provided. Referring to the example embodiment described herein with reference to Figs. 3 and 4 for illustration only, the data packet or the cluster of data packets is received on the service link. Each data packet includes user data to be transmitted to a client using the communication link, e.g. using the SD-WAN link, and control information, e.g. in the data packet header, indicative of a priority level associated to the data packet.

[84] In step 502, the number of channels to be used for transmitting the cluster or data packet is dynamically reconfigured by determining the number n of channels to be used according to the priority level of the data packets. In the embodiment of Figs. 3 and 4, the number n of channels to be used is determined by the packet encapsulating unit 210. Each data packet of a cluster is encapsulated. If a large data packet is split into sub-packets, each sub-packet is encapsulated individually. The SD-WAN encapsulation internal header associates the sub-packets or the cluster together and includes SD-WAN restrictions associated to each data packet or to each sub-packet to indicate the number n of WDM channels to be used for the transmission. Encapsulated sub-packets will be called encapsulated packets in the following. For instance, the SD-WAN multiplexer 230a or 230b receives encapsulated packets and selects the n specific WDM channels to be used for transmission the encapsulated packets using the restrictions and considering the available channels.

[85] In step 503, the encapsulated packets are assigned to n specific channels among the available channels. In the embodiment of Figs. 3 and 4, this is done by the SD-WAN multiplexer 230a or 230b which selected the WDM channels to be used for transmitting the traffic flow.

[86] In step 504, the data packets of the cluster or the split sub-packets, i.e. the encapsulated packets, are split apart by distributing them on the n specific channels selected for transmission over the communication link. In the embodiment of Figs. 3 and 4, this is done by the SD-WAN multiplexer 230a or 230b which multiplexes the

encapsulated packets by distributing the traffic flow among the specific WDM channels.

[87] Figs. 6, 7 and 8 illustrate an example implementation of the SD-WAN encoding/decoding unit 100. This implementation is performed using a Field- Programmable Gate Array (FPGA) with an embedded PowerPC processing unit 802 (see Fig. 8) and which is programmed to apply the SD-WAN multiplexing method. The SD-WAN encoding/decoding unit 100 is composed of various elements implemented in the FPGA and which are now described.

[88] Fig. 6 shows an example implementation of the packet encapsulating unit 210 and of the decapsulating unit 250 of the SD-WAN interfacing module of Fig. 4. In this embodiment, the encapsulating unit 210 comprises eight Bcm5466 receivers 212, eight header formatters 214, an Ethernet path merger 216 and an interface converter 218. There are as many Bcm5466 receivers 212 and header formatters 214 as there are Ethernet ports, i.e. eight in this case. The packet encapsulating unit 210 also uses a number of buffers illustrated as shaded and boxes with stripes.

[89] The Bcm5466 Receiver 212 receives the Ethernet formatted data packets from the service interface 400. It is the receiving MAC layer of the Ethernet port. It handles the incoming data to isolate the data packets. It handles the Cyclic Redundancy Check (CRC) function by striping off the CRC32 data from the data packet and checking its validity.

[90] The header formatter 214 receives the user data of the data packets from the Bcm5466 receiver 212, along with its control information included in the headers of the data packets. Control information includes the size of the packet, the CRC validity, the VLAN ID, the class of service, etc. A first one of the tables 215, i.e. the VLAN control table, is read to determine if the packet must go further, and if so what value must be written in the DestinationlD field. If the CRC is not valid, the VLAN ID is not asserted, or the next buffer free space is insufficient, the packet is discarded; otherwise, the SD-WAN internal and optical header are added to encapsulate the data packet which is propagated toward the node cross connect 220. The bandwidth

allocation table is also read in tables 215. If the bandwidth allocation table allows packet splitting (based on packet size and Service Level Agreement), the packet is split following the splitting ratio given by the table. Each sub-packet is encapsulated with the appropriate numbering for further packet reconstruction.

[91] The Ethernet path merger 216 merges the streams of encapsulated packets coming out of the buffers at the output of the header formatters 214. The Ethernet path merger 216 monitors the upstream buffer free space to delay the transfer until enough space is available. Packet prioritization is also performed based primarily on the class of service of the packets in the buffers to merge, higher value first. When the class of service is the same, the time stamp is used, lower value first.

[92] The interface converter 218 at the output of the Ethernet path merger 216 transfers the encapsulated packets from a simple FIFO buffer interface to a cross- connect device 221 part of the node cross-connect 220.

[93] The decapsulating unit 210 comprises an interface converter 219, an Ethernet steerer 252, eight header strippers 254 and eight Bern 5466 formatters 256. There are as many header formatter 254 and Bern 5466 formatter 256 as there are Ethernet ports, i.e. eight.

[94] The Ethernet steerer 252 steers the encapsulated packets received from the node cross-connect 220 to the appropriate Ethernet port. The tables 215 also include the delivery control table described herein and which is read to determine on which port the packet should go. The Ethernet steerer 252 transfers the encapsulated data packets received simultaneously to the Ethernet ports specified by the delivery control table. The Ethernet steerer 252 also monitors the free space of the target buffer to delay the transfer until enough space is available.

[95] The header strippers 254 receive encapsulated data from the Ethernet steerer 252 and strip the SD-WAN internal header from the packet for transmission on to the Ethernet port.

[96] The Bern 5466 formatters 256 are the transmission MAC layer of the Ethernet ports. They perform clock domain transition and CRC32 computation. They handle the control signalling specifying when a packet starts or stops. They also manage the duration of the Inter-Packet Gap to keep it to a minimum while respecting the gab required by the standard, i.e. 12 bytes.

[97] Fig. 7 illustrates an example implementation of the SD-WAN multiplexer 230 and the optical receiving unit 240 of the SD-WAN interfacing module 100 of Fig. 4.

[98] The SD-WAN multiplexer 240 comprises an interface converter 241 , a Tx path merger 242, an optic steerer 243, four channel shapers 244 and four Tlk3114 formatters 245. There are as many channel shapers 244 and Tlk3114 formatters 246 as there are WDM channels, i.e. four in this case.

[99] The interface converter 241 at the output of the node cross-connect 220 transfers the encapsulated packets from a switch merger 222 to a simple FIFO buffer interface.

[100] The path merger 242 merges two streams coming out of simple buffers, i.e. out of the interface converter 241 and from a transit path merger 236 (described below) of the optical receiving unit 230. The path merger 242 monitors the buffer inside the optic steerer 243 and transfers packets when enough buffer space is available.

[101] The optic steerer 243 steers the incoming packets from the path merger 242 to the appropriate WDM channel port. A channel table 248, which corresponds to the bandwidth allocation table, is read, along with SD-WAN management restrictions in the SD-WAN internal header, to determine on which port the packet should go. The optic steerer 243 monitors the free space of the target buffer to delay the transfer until enough space is available. It also supports internal wrapping, to the optical port of the other side, when no channel of the actual side is enabled.

[102] Each channel shaper 244 receives the packets corresponding to its WDM channel port and takes off the internal header of the packet, adds an optic trailer to

the packet, and computes a 32-bit CRC before transferring the packet to the next TIk 3114 formatter 245.

[103] Each Tlk3114 formatter 345 receives the packet from the channel shaper 244. It is the transmission MAC layer of the WDM channel optical port. It handles control codes specifying when a packet starts or stops. It also manages the various idle codes pseudo-random sequences. The Tlk3114 formatter 245 is also responsible for transmitting the VLAN packet transition information, which allows the use of a unified communication trunk, as if it would consist of several independents channels.

[104] The output of the TIk 3114 formatter 245 goes to a serialiser/deserialiser before going the transport layer interface (XENPAK) 600.

[105] The optical receiving interface 230 comprises four Tlk3114 receivers 231 , four packet reshapers 232, four header formatters optic 233, a downstream path merger 234, a down stream interface converter 235, a transit path merger 236 and an interface converter 237. There are as many Tlk3114 receivers 231 , packet reshapers 232 and header formatters optic 233 as there are WDM channels, i.e. four in this case.

[106] Each Tlk3114 receiver 231 receives data from the transport layer interface 600 thought the serialiser/deserialiser 246. Each WDM channel port is associated with one Tlk3114 receiver 231. Each Tlk3114 receiver 231 is the receiver MAC layer of the optical port. It handles the incoming data to detect control codes and recover the received packets.

[107] Each packet reshaper 232 receives data from its corresponding Tlk3114 receiver 231. It takes off the packet transition information, the trailer and the 32 bits CRC from the data. It then computes the CRC on the remaining data and validates if it matches with the received CRC.

[108] Each header formatter optic 233 receives the user data of the data packets received from its corresponding packet reshaper 232, along with the control information, i.e. header information, including the size of the packet, the CRC validity,

etc. If the CRC is not valid, or the next buffer free space is insufficient, the packet is discarded; otherwise, a SD-WAN internal header is added to the packet, and it may then be propagated further. The header formatter optic 233 also drops data packets that should be dropped to this node. Data packets to be dropped are determined according to the destination address associated with the packet. The packet will also be dropped if it is a broadcast coming from another node, or if it is a unicast targeting the actual node. The packet will be propagated in transit if the TimeToLive has not expired and either the packet is a broadcast coming from another node, or it is a unicast targeting another node. The header formatter optic 233 is responsible for decrementing the TimeToLive, once per node, along the optical journey. A special sequence on two control bits in the optical header prevents double delivery of a packet in wrapping conditions.

[109] The downstream path merger 234 merges the streams of data packets to be dropped and coming out of buffers at the output of the header formatters optic 233. The downstream path merger 234 monitors the upstream buffer free space to delay the transfer until enough space is available. The packet transfer is based on each input buffer used size, higher value first.

[110] The transit path merger 236 merges the streams of data packets in transit and coming out of buffers at the output of the header formatters optic 233. The transit path merger 235 monitors the upstream buffer free space to delay the transfer until enough space is available. The packet transfer is based on each input buffer used size, higher value first.

[111] The interface converter 237 at the output of the transit path merger 236 transfers the data packets from a simple FIFO buffer interface to a cross-connect device 221 part of the node cross-connect 220.

[112] Fig. 8 is illustrates an example implementation of the SD-WAN interfacing module 100 of Fig. 4 in a FPGA with PowerPC. The node cross-connect 220 includes several 2x2 cross-connects 221 connected together to constitute the node cross- connect 220 along with switch mergers 222, traffic validators 223 and traffic

generators 224. A processing unit 802, i.e. PowerPC, included on the FPGA, controls different network management functionalities. The FPGA also includes an external memory controller 803 connected to an external flash memory 808, a SDRAM controller 804 connected to an external SDRAM 809, a controller 805 connected to a low-voltage differential signalling ring 810, a Universal Asynchronous Receiver/Transmitter (UART) connected to a host processing unit 811 , a UART 807 connected to a debug processing unit 812 and a Management Data Input/Output controller 814 connected to the processing unit 802. The SD-WAN interfacing module 100 also comprises an FPGA configuration device 813 connected to the processing unit 802.

[113] Each cross-connect device 221 merges and/or splits all streams of encapsulated packet received. The merge includes packet reordering in the transmit queue, based primarily on the class of service (highest to lowest) of the streams to merge. When the class of service is the same for both streams, a time stamp (oldest to newest) is used. When the cross-connect is used to split a stream, two transmit queues are necessary. The switching is done following the destination parameters include in the internal header.

[114] While the SD-WAN technology is described herein in an access network, the same technology may also be applied to other network levels such as in a metro core or metro edge network. Furthermore, other topologies can also benefit from SD- WAN multiplexing. Examples of such other topologies are point-to-point, hub and spoke, flat ring, daisy chain and passive optical network topologies. Accordingly, the SD-WAN interfacing module 100 may include only one or more than two transport layer interfaces 600, XENPAK, for use respectively in a point-to-point or a star topology for example.

[115] Fig. 9 illustrates another example of the SD-WAN interfacing module 100' which is adapted for use in a point-to-point topology for example. The SD-WAN interfacing module 100' has a single transport layer interface 600 for connection with a single other node. Similarly to the SD-WAN interfacing module 100 of Figs. 3 and 4, the transport layer interface 600 is connected to an SD-WAN multiplexer 230 and to

an optical receiving unit 240. The SD-WAN interfacing module 100' also has a service interface 400 connected to a SD-WAN encapsulating unit 210 and a decapsulating unit 250 similarly to the SD-WAN interfacing module 100 of Fig. 3 and 4. The difference lies in that, as there is a single transport layer interface 600.

[116] Statistical multiplexing gain

[117] In order to appreciate the statistical multiplexing gain that can be achieved using SD-WAN technology, a statistical model was developed. Just like earlier traffic models (e.g. Erlang loss formula developed for call blocking probability in telephone networks), the present model links network capacity, traffic demand and realized performance.

[118] Consider a communication channel with capacity C. Knowing that each customer has a mean bandwidth consumption c _mea n with an instantaneous maximum bandwidth consumption c _max , the maximum number of customers n _sim that can be served simultaneously is given by:

« , _m =— (D c

[119] The fact that a customer wants to use the channel with instantaneous capacity Cm _3x is a binomial statistical variable for which the probability of occurrence p is as follows:

P = (2)

[120] Of course, P ^≤ as ^c ™ ^« ™ ^{~ c} ™ ^x , the equality corresponding to the case where the customer uses the channel all the time. The probability P _r that r customers want to use the channel simultaneously when a total of n customers are served by the same channel is then given by a binomial probability distribution:

P = . " ^{' X} ^ x?" ^" " (3) rιx(n - r)\

where 1 ^~ ^ ^~ P represents the probability that a given customer does not want to use the channel.

[121] The notion of statistical multiplexing gain implies that a probability threshold for which customers can expect to use the channel without any blocking is defined. This is what is referred to as the grade of service (GoS). To determine the maximum number of customers n _max that can be served by a communication channel while respecting the grade of service, there should be found the maximum value of n such that the cumulative probability of having n _sim customers wanting to access simultaneously the channel is greater than or equal to the grade of service GoS. Mathematically, n _max is the maximum value of n such that:

∑P _r ≥ GoS (4) r=0

[122] By doing so, it is ensured that blocking, when the number of customers wanting to use the channel is greater than the maximum number of customer that can be served instantaneously n _S i _m , happens with a probability lower than or equal to 1 - GoS consequently respecting the grade of service expected by customers.

[123] As an example, suppose that eight wavelengths are used on a fibre link, each wavelength being modulated at 2.5 Gb/s. Suppose that each customer consumes a bandwidth capacity c _mea n of 1 Mb/s with a maximum consumption c _max of 100 Mb/s, which corresponds to the case of a Fast Ethernet port. The total channel capacity being 8 * 2.5 Gb/s = 20 Gb/s, the maximum number of customers n _S i _m that can be served instantaneously is

^sm c _max 100 Mb/s [124] Considering a 97% grade of service GoS, it can be shown, that the GoS is respected per equation (4) as long as the total number of served customers n is not greater than 17558.

[125] In the maximal case of 17558 served customers, statistical multiplexing gain is such that a bandwidth capacity of only 1.14 Mb/s is reserved for each customer,

even though each of them may have a 100 Mb/s instantaneous consumption. The theoretical limit, which will never be reached in practice, is to reserve a bandwidth capacity of c _mean for each customer. The ratio R between the bandwidth capacity reserved for each customer and bandwidth capacity c _mea n is an interesting metric by which we can appreciate the statistical multiplexing gain - the lower it is, the more efficient is the fibre optic infrastructure usage. In our example, ratio R is 114%.

[126] It should finally be mentioned that the developed statistical model does not take into account any encapsulation overhead to keep the model as simple as possible. Including encapsulation overhead in the model would have been difficult as the relative importance of the overhead depends on the length of packets being transported. The model, just like the Erlang loss formula, is independent of call duration and is independent of the packet length. The priority management capability of SD-WAN technology has not been considered for the same reason.

[127] If the network relies on a single wavelength, SD-WAN should show the same performance as RPR in terms of statistical multiplexing. With multiple wavelengths, the SD-WAN technology is more efficient.

[128] Fig. 10 compares SD-WAN technology with other technologies in terms of the number of served customers as a function of the total installed capacity The squares curve is for a SD-WAN link, the diamonds curve is for a 10 Gigabit Ethernet link, the triangles curve is for a Resilient Packet Ring (RPR) link, the cross curve is for a switched Ethernet over SONET link (S-EoS), and the stars curve is for an Ethernet over SONET link (EoS). It is assumed that customers are each provided a Fast Ethernet (i.e. 100 Mb/s) port with a 1-Mb/s mean bandwidth consumption. Grade of service is 97%. For all technologies with the exception of 10 Gigabit Ethernet (10GE), the modulation rate of a single wavelength is 2.5 Gb/s. Of course, 10 Gb/s per wavelength is assumed for 10GE technology. In the particular case of S-EoS, we arbitrary chose an eight nodes ring as S-EoS statistical multiplexing gain cannot be determined without knowing the number of nodes. Fig. 10 shows the statistical multiplexing advantage of SD-WAN over any technology. Under the assumptions at 40 Gb/s total capacity, SD-WAN has 32% more effective throughput than the best

performing 2.5 Gb/s technology (RPR). It also has a 9% effective throughput advantage over 10GE, even though SD-WAN is used on a 2.5 Gb/s per wavelength architecture. In fact, the individual modulation rate per wavelength has no impact on the SD-WAN statistical gain as the total bandwidth capacity can be shared among customers.

[129] Fig. 11 shows SD-WAN statistical multiplexing gain over RPR (triangles curve) on a 40Gb/s infrastructure (16 wavelengths * 2.5 Gb/s), and 10 Gigabit Ethernet (squares curve) technologies on a 40Gb/s infrastructure (4 wavelengths * 10 Gb/s) as a function of the grade of service. The more the network operator is required to offer a high grade of service, the more SD-WAN technology becomes attractive. At 99,999% grade of service, the SD-WAN effective throughput is 109% higher than the RPR effective throughput while it is still a respectable 27% higher than 10GE effective throughput. As a consequence, increasing the grade of service offered to customers is much less impacting with SD-WAN technology than with other technologies.

[130] Fig. 12 shows SD-WAN statistical multiplexing gain over 10 Gigabit Ethernet technology as a function of the maximum bandwidth capacity offered per customer, wherein the stars curve is for a 40 Gb/s infrastructure and the dots curve is for a 20 Gb/s infrastructure. On a 20 Gb/s infrastructure, SD-WAN technology effective throughput gain vs. 10GE technology goes from 2% at 10 Mb/s per customer (regular Ethernet ports) to 6% at 100 Mb/s per customer (Fast Ethernet ports) to 17% at 1000 Mb/s per customer (Gigabit Ethernet ports). On a 40 Gb/s infrastructure, where the total number of wavelengths is doubled with respect to the 20 Gb/s infrastructure, SD- WAN technology effective throughput gain vs. 10GE technology goes from 3% to 10% and to 32% for each of the previously mentioned cases respectively. Consequently, it is shown that as maximum bandwidth capacity per customer is increased and as the number of wavelengths in the infrastructure is increased, SD-WAN technology is more and more efficient in comparison to any other technology.

[131] The effective throughput gain obtained by statistical multiplexing is an advantage of SD-WAN technology. SD-WAN technology also offers a high flexibility in

bandwidth management, allowing a totally arbitrary bandwidth allocation amongst customers and taking into account traffic classes for priority management. In case of a failure on a particular wavelength, traffic flows transported on the failing wavelength are automatically reassigned to remaining wavelengths. In such a case, the grade of service of all customers is impacted but no customer is left without services. Also, it makes it possible for the network operator to grow his network as needed by traffic demand, consequently avoiding punctual and substantial capital expenditures and inefficient network overbuilds.

[132] Application to Multiple System Operators (MSOs) networks

[133] Any multi-wavelength network can benefit from the SD-WAN technology. The case of Multiple System Operators (MSOs) networks is an example of such a multi-wavelength network.

[134] MSO networks are typically made of a fibre optic network for the part of the network that is closer to the central office (also known as head end) and a cable TV RF network, using coaxial cable, for the part of the network that is closer to the customers. At the junction of the fibre optic part and the cable TV RF part, electronic devices known as fibre nodes are used to transfer the transported information from the optical domain to the RF domain and vice-versa. On the fibre optic part, multiple TV signals, each on a particular RF carrier corresponding to the TV channel, are transported using an optical signal typically at wavelength of 1310 nm. Transition from the optical domain to the RF domain simply requires photodetection at the fibre node and amplification of the resulting RF signal.

[135] Nowadays, MSOs are more and more interested in increasing their service offer to customers. As an example, MSOs offer Internet connectivity in addition to the conventional cable TV service. Some technologies allow the transport and distribution of the digital information over the RF network. These technologies are usually suitable for residential type of services.

[136] At some point, MSOs are also interested in providing new services not only for residential customers but also for commercial customers e.g. small and medium- size businesses. The two main differences between residential customers and commercial customers are a higher need for bandwidth, in terms of capacity and flexibility, and a higher need for reliability of service.

[137] To serve commercial customers, some MSOs began deploying multiple optical signals over their fibre optic network. By doing so, they increase the bandwidth capability of their fibre optic network, consequently enabling new services for commercial customers. One technology to do so is the Coarse Wavelength Division Multiplexing (CWDM) technology. This technology allows the transport of multiple independent optical signals over a given fibre optic network because each optical signal has its own allocated wavelength. As an example, a MSO may transport TV signals on the 1310 nm wavelength while transporting digital data for three commercial customers on the 1490 nm, 1510 nm and 1530 nm wavelengths. A simple CWDM optical filter is used, at the fibre node, to demultiplex optical signals in order to provide connectivity to the commercial customers without disturbing the cable TV service for residential customers.

[138] By adding an overlay of CWDM wavelengths to serve commercial customers, capital investments are kept to a minimum. However, exactly one wavelength is used for each commercial customer served. As the number of CWDM wavelength is limited (there can be a maximum of 18 CWDM wavelengths in a fibre optic network per ITU-T G.694.2, although typical systems tend to stay at 8 or less wavelengths), the number of customers that can be served is also limited. That limitation is seen even though each customer does not necessarily need the complete bandwidth that may be transported with a given wavelength. As a result, substantial residual bandwidth remains unexploited.

[139] SD-WAN multiplexing breaks that one-to-one relationship between deployed CWDM wavelengths and customers and exploits the residual bandwidth. Fig. 13 shows an example of application of the SD-WAN technology on a Multiple Systems Operator (MSO) access point-to-point topology.

[140] In this embodiment, a SD-WAN link 1016 is created between a central office 1002 and a client node CO. Both the central office 1002 and the client node CO includes an SD-WAN interfacing module 100 is used, which are respectively designated as the central interfacing module 1004 and the node interfacing module 1006. The client node CO is actually located in an outside plant. The signal transmitted on the SD-WAN link 1016 includes SD-WAN WDM signals at 1479, 1490, 1510 and 1530 nm, and an additional TV signal at 1310 nm. The node interfacing module 1006 includes WDM optical filter to separate the TV signal from the SD-WAN signals. The TV signal is to be transmitted to a cable TV RF network 1022 using a coaxial cable TV link 1020, while the SD-WAN signals are to be demultiplexed by the node interfacing module 1006 before routing the data packets to the appropriate commercial customer C1 , C2, C3, C4, C5, C6 or Cn in a star configuration using customer optical links 1018. The customer optical links 1018 which connect the node interface module 1006 to the each commercial customers may use various technologies, one of them being the use of CWDM wavelengths. It all depends on the available resources for a particular situation.

[141] The central interfacing module 1004 multiplexes the traffic flows to be transmitted to commercial customers. These traffic flows are processed according to SD-WAN technology and sent over the fibre optic SD-WAN link 1016 or network using a lower number of CWDM wavelengths than the number of served commercial customers. At the client node CO, the node interfacing module 1006 receives the CWDM wavelengths and demultiplexes the traffic flows according to SD-WAN technology. In this illustrative case, colorless signals over optical fibres are used to send the disaggregated traffic flows from the node interface module 1006 to commercial customers C1 to Cn. Any other appropriate technology may also be used (e.g. copper wire pair (VDSL), CWDM, wireless (WiMax), etc.). Available bandwidth is not necessarily evenly allocated between customers as SD-WAN technology allows a totally arbitrary and dynamic allocation of bandwidth between customers.

[142] It should be noted that in this and the following examples the complete network is a bidirectional network, i.e. upstream traffic flows coming from the

commercial customers are aggregated and processed according to SD-WAN technology by the node interface module 1006 while the central interface module 1004 disaggregates these upstream traffic flows. For simplicity, figures only show the downstream direction of the network.

[143] In the MSO context, SD-WAN technology breaks the one-to-one relationship between the number of CWDM wavelengths and the number of served commercial customers. It then enables MSOs to increase substantially the number of commercial customers that can be served from their fibre network, with minimal capital investment and without disturbing the TV signal part of their service offering for residential customers. It also allows the MSOs to grow their network gradually as they can simply add a CWDM wavelength each time the total capacity of the installed wavelengths is consumed. That growth of the network may be required as the reserved bandwidth per customer increases or as the total number of customers increases or by a combination of these two options. The bandwidth of the added wavelength becomes readily available to the whole network and it remains to the MSO to decide how this additional bandwidth is to be allocated.

[144] The proposed access network architecture exploiting the SD-WAN method breaks the one-to-one relationship between the number of CWDM wavelengths and the number of served commercial customers. It may enable MSOs to increase substantially the number of commercial customers that can be served from their fibre network, with minimal capital investment and without disturbing the TV signal part of their service offering for residential customers. Not to be neglected, it may also allow the MSOs to grow their network gradually as they can simply add a CWDM wavelength each time the total capacity of the installed wavelengths is consumed. That growth of the network may be done by increasing the reserved bandwidth per customer, by increasing the number of customers or by a combination of these two options. The bandwidth of the added wavelength becomes readily available to the whole network and it remains to the MSO to decide how this additional bandwidth is to be allocated.

[145] Many other embodiments are also possible. Fig. 14 shows an embodiment similar to the one of Fig. 13 but where the node interface module 1006 is located in the premises of one of the commercial customers, i.e. C1. In this case, no active device is located in an outside plant. Only optical filters are deployed in the outside plants to separate the incoming SD-WAN WDM signals at 1479, 1490, 1510 and 1530 nm, from the TV signal at 1310 nm. The TV signal is converted into a RF signal at the outside plant and transmitted to the cable TV RF network 1022 using a coaxial cable TV link 1020, similarly to the embodiment Fig. 14. In the embodiment of Fig. 15, the first commercial customer C1 where the node interface module 1006 is located may be a big bandwidth consumer while residual bandwidth can be used by the other customers C2 to Cn. The links 1018 from that first commercial customer to the other ones may use various technologies, one of them being the use of CWDM wavelengths. It all depends on the available resources for a particular situation.

[146] Fig. 15 shows another topology. In this case, the network includes a ring access network formed between multiple node interfacing modules 1006 located in the premises of first-level customers C1 , C2, C3 and C4, and the central interface module 1004. Some additional customers C11 , C12, C13, C21 , C22, C23, C31 , C32, C33, C41 , C42 and C43 are also linked to the first-level customers in a radial topology. The ring topology provides improved robustness against failure like a fibre cut. Similarly to the embodiments of Figs. 13 and 14, optical filters are deployed in an outside plants to separate the incoming SD-WAN WDM signals at 1479, 1490, 1510 and 1530 nm to be transmitted on the ring, from the TV signal at 1310 nm to be transmitted to the cable TV RF network 1022. The first-level customers C1 , C2, C3 and C4, each having a node interfacing module 1006, are major nodes. They can be seen as hubs for the other customers which typically require less bandwidth. The link between a first-level customers and its radially connected customers may use any appropriate technology. The topology of Fig. 15 also breaks the one-to-one relationship between the number of CWDM wavelengths and the number of served commercial customers. It offers total flexibility for dynamic bandwidth allocation without having to consider bandwidth on a per wavelength granularity.

[147] While illustrated in the block diagrams as groups of discrete components communicating with each other via distinct data signal connections, it will be understood by those skilled in the art that the illustrated embodiments may be provided by a combination of hardware and software components, with some components being implemented by a given function or operation of a hardware or software system, and many of the data paths illustrated being implemented by data communication within a computer application or operating system. The structure illustrated is thus provided for efficiency of teaching the described embodiment.

[148] The embodiments described above are intended to be exemplary only. The scope of the invention is therefore intended to be limited solely by the appended claims.