WO2005117300A1 - Method and system for data transfer - Google Patents

Abstract

A method for transferring data in a communication system, which comprises a first communication node and at least one user unit, is disclosed. The method includes, individually for each user unit, establishing and maintaining two distinct data transfer connections simultaneously, each on a different path between the first communication node and the user unit. One of the data transfer connections is assigned as a primary connection and the other data transfer connection is assigned as a secondary connection between the first communication node and the user unit. Data is transferred over a primary communication path on which the primary connection is established and transfer of the data is switched instantaneously from over the primary communication path to over a secondary communication path on which the secondary connection is established, when transfer of data over the primary communication path is no longer feasible. A system that implements the above method is also disclosed.

Description

METHOD AND SYSTEM FOR DATA TRANSFER

BACKGROUND

[0001] This invention relates generally to a method and a system for transfer of data in an access network. More particularly, this invention relates to a method and system for transfer of data in an optical access network including passive optical networks (PONs).

[0002] Figure 1 shows a known unprotected passive optical network (PON) 2 that includes passive elements, such as fiber-optic cables 4A, 4B, passive optical splitters/couplers 6, connectors (not shown), and splices (not shown). Active network elements, such as an optical line termination (OLT) 8 and multiple optical network units (ONUs) 10, are located at end points of the PON 2. An ONU 10 that also provides a user port or network termination function is more commonly referred to as an optical network termination (ONT) 10. Optical signals traveling downstream across the PON 2, from the OLT 8 to the ONUs 10, are split onto multiple fiber-optic cables 4B by the optical splitters 6. Optical signals traveling upstream across the PON 2, from the ONUs 10 to the OLT 8, are combined onto the single fiber-optic cable 4A by the optical couplers 6. In the PON 2, there is effectively a single communication link between each ONU 10 and the OLT 8 because of the fiber-optic cable 4A. Such a point-to-multipoint configuration is typically used for residential applications.

[0003] With a single communication link connecting each ONU 10 to the OLT 8, communication between the two elements 8, 10 is lost once a fault develops in the communication link. Consequently, to increase access network survivability for the ONUs 10, each ONU 10 is connected to the OLT 8 via an additional communication link. Figure 2 shows such a protected system 12 for enhancing access network survivability, wherein each ONU 10 is connected to the OLT 8 via two links 14, 16. Two switching schemes, known as the 1 :1 and the 1 +1 switching scheme for switching traffic between the two links 14, 16 in this protected system 12 are described in the ITU-T G.983.5 and G.983.1 standards respectively. Figures 3 and 4 illustrate the 1 :1 and 1 +1 switching schemes respectively. One of the two links 14, is used as a working entity or PON while the other link 16, is used as a protection entity or PON as they are known in the standards.

[0004] According to the G983.5 standards for the 1 :1 switching scheme, traffic is classified as either high-priority/normal traffic or low-priority/extra traffic. High-priority traffic is required to be carried when one of the two PONs 14, 16 is faulty. Low-priority traffic is carried only when both PONs 14, 16 are operational. When both PONs 14, 16 are operational, the high-priority traffic is carried over a working PON while the lower-priority traffic is carried over the protection PON. In the event that the working PON 14 is found faulty, the low- priority traffic carried over the protection PON 16 is pre-empted and the high- priority traffic is switched to be carried by the remaining protection PON 16 instead. The high-priority traffic is thus said to be protected while the low- priority traffic is not. When carrying high-priority traffic, a PON 14, 16 does not carry any low-priority traffic even if spare or unused bandwidth on that PON is available. Bandwidth on the PON is thus not efficiently utilized.

[0005] In the downstream direction, both high-priority and low-priority traffic are broadcast to selectors in the OLT 8 for them to choose the type of traffic that is to be transmitted onto the working and protection PONs 14, 16. In the ONU 10, there is a simple function that selects the correct traffic flow from the OLT 8 using a PON selection trace (PST) message from the OLT. A similar survivability mechanism is used for transmission in the upstream direction. When switching transfer of high-priority traffic from over the working PON 14 to over the protection PON 16, ranging may or may not be required depending on implementation. If the protection PON 16 has been carrying low- priority/extra traffic, ranging would have been done between the OLT8 and ONU 10 for the protection PON 16. This ranging value may be different from the value for the primary PON 14. When the high priority/normal traffic needs to be switched over to the protection PON 16, a line card that terminates the primary PON 14 can carry out ranging over the protection PON 16, or get the information from the line card of the protection PON 16. Registration would have to be carried out for high priority/normal traffic over the protection PON 16 when it is switched over to be transferred on that PON 16. ) [0006] For the 1+1 switching scheme shown in Figure 4, traffic in the downstream direction is bridged to both the working and the protection PONs 14, 16. That is, the traffic is sent simultaneously on both PONs 14, 16 to be redundantly carried over the two PONs 14, 16, and there is thus less bandwidth for low-priority/extra traffic, as compared to that in the 1 :1 scheme. In the ONU 10, the traffic is selected based on the signal quality and/or a PON section trace (PST) message sent by the OLT 8. In the upstream direction the same functionality exists. In the upstream direction, traffic is bridged to both working and protection PONs 14, 16 and a selection mechanism at the OLT 8 selects traffic on one of the PONs 14, 16 based on the signal quality and provisioning. Switching traffic from one PON 14, 16 to the other PON 16, 14 is initiated by either the ONU 10 or the OLT 8, depending on which element 8, 10 detects a fault first. In the event of a failure that affects only downstream traffic, the ONU 10 will detect that a fault has occurred and notify the OLT 8 of the failure in order to complete the switching event.

[0007] Figure 5 shows a known generic X:N protection system that supports both protected ONUs 10 and unprotected ONUs 11 as illustrated in the ITU-T G983.5 standards. In this system, X number of protection PONs 16 are provided and configured for N number of working PONs 14 (with l≤X≤N). The N working PONs 14 can support a mixture of protected and unprotected ONUs 10, 11. The protected ONUs 10 can be connected to any of the X protection PONs 16. The protected ONUs 10 connected to the same working PON 14 can be connected to different protection PONs 16 and protected ONUs 10 connected to different working PONs 14 can be connected to the same protection PON 16. This connection scheme is compatible with both the 1 :1 and 1+1 switching schemes and is independent of protocol. [0008] The connection scheme provides protection against multiple failures in N working PONs 14 with less than N protection PONs 16. Other protection systems having different configurations are disclosed in U.S. Patent 5,896,474, entitled Optical network having protection configuration"; U.S. Patent 5,920,410, entitled "Access network"; U.S. Patent 6,327,400, entitled "Protection scheme for single fiber bi-directional passive optical point-to- multipoint network architectures"; and U.S. Patent No. 6,351 ,582, entitled "Passive optical network arrangement". In these systems, before traffic of an ONU 10 can be switched over to the protection PON 16, the ONU 10 has to be initialized on the protection PON 16, i.e. the ONU 10 has to establish a connection on the protection PON 16. Such an initialization includes ranging and registration of the ONU 10 with the OLT 8 of the protection PON 16. Ranging is a method of measuring the logical distance between each ONU 10, 11 and the OLT 8 and of determining the transmission timing for ensuring that upstream traffic from different ONUs 10, 11 on the same PON 14, 16 do not collide. The above-mentioned documents disclose protection switching but not bandwidth allocation in a PON 14, 16.

[0009] Figure 6 shows an example of a Tframe 18 frame format for the ITU-T G.983 155.52/155.52 Mbit/s broadband PON (B-PON) system. In the downstream direction, each Tframe 18 includes two Physical Layer Operation and Maintenance (PLOAM) cells 20, 22. The first PLOAM cell 20 contains twenty-seven grant fields while the second PLOAM cell 22 contains twenty-six grant fields. Both the PLOAM cells 20, 22 contain an additional 12-byte message each. This 12-byte message may contain information, such as the MESSAGE_PON_ID, MESSAGEJD, AND MESSAGE-FIELD TO MESSAGE-FIELD O. The grant field specifies the GRANTJD of the ONU 10 that is granted the permission to transmit in a corresponding upstream time slot. In the ITU-T Recommendation G.983.1 , grants and bandwidth are allocated in a static manner, which is also discussed in the article by D P Shea, J E Mitchell, R P Davey, "PONdering the access network," London Communications Symposium, London, UK, 9th-10th September 2002. Static bandwidth allocation refers to the allocation of a fixed amount of bandwidth to each ONU 10 in each frame. The OLT 8 does not adjust the bandwidth allocation based on the real-time demand of each ONU 10. For example, some ONUs 10 may need more bandwidth at a particular time than the fixed amount while others may not need that fixed amount but the OLT 8 is not able to adjust the bandwidth allocation to give more bandwidth to those ONUs 10 in need of the additional bandwidth.

[0010] G.983.4 specifies two dynamic bandwidth assignment (DBA) approaches which are improvements over the static bandwidth assignment approach defined in G.983.1. The first method is referred to as "idle cell adjustment". In this approach, the OLT 8 monitors the bandwidth used by each of the ONUs/ONTs and when the utilization exceeds a predefined threshold, additional bandwidth will be assigned if it is available. The second method is referred to as "buffer status reporting". According to this approach, the ONUs/ONTs 10 report the status of their buffers or queues by using minislots. These minislots are slots that are less than the size of an ATM cell of fifty-three bytes. Minislots within the boundary of an ATM cell can be assigned to different ONUs/ONTs 10 for transmission of reports. The OLT 8 reassigns the bandwidth according to the ONU/ONT reports. The enhancement also allows DBA to accommodate several Transmission Containers (T-CONTs) in one ONU/ONT 10 and each T-CONT in an ONU/ONT 10 can operate independently of other T-CONTs. T-CONTs are formats defined to separate traffic/information belonging to different streams, for example, a particular Asynchronous Transfer Mode Virtual Path or Virtual Channel. How the bandwidth is apportioned to different ONUs/ONTs 10 is not specified.

[0011] The above-mentioned prior art systems suffer a number of additional disadvantages. Another disadvantage is that bandwidth and services are protected only on a per-ONU/ONT basis. In other words, all the subscribers that are connected to a particular ONU 10 must have their service and bandwidth either protected or not protected at all. Different levels of reliability cannot be provided to individual subscribers that are connected to the same ONU 10. [0012] Yet another disadvantage is that as traffic from and for different subscribers are not managed and queued individually at the ONUs 10 and the OLT 8, quality of service (QoS) measurements, such as bandwidth and delay, can be provided at best on a statistical basis due to interference between traffic from different subscribers. This will lead to a complicated, not easily monitored and not easily enforceable service level agreement between the service provider and its subscribers. SUMMARY

[0013] According to an aspect of the invention, there is provided a method for transferring data in a communication system, which comprises a first communication node and at least one user unit. Each user unit is connected to the first communication node via a network of a number of communication paths such that, for each user unit, two distinct data communication connections, each on a different path, can be provided between the first communication node and the user unit. The path may be a wired or a wireless path. A wired path includes, but is not limited to, a path defined by copper cables and a path defined by fiber-optic cables. Two communication paths are considered to be different even if they share a common section. Such a common section shared by two paths may be a section that is less prone to damage. Ideally, the two communication paths are totally separate from each other, i.e. they share no common section, so that a break anywhere along one of the two paths would not affect the other. Such communication paths that do not share any common section are referred to herein as distinct communication paths.

[0014] The method includes, individually for each user unit, establishing and maintaining both data transfer connections simultaneously and assigning one of the data transfer connections as a primary connection and the other data transfer connection as a secondary connection between the first communication node and the user unit. The method further includes transferring the data over a primary communication path on which the primary connection is established and switching transfer of the data instantaneously from over the primary communication path to over a secondary communication path on which the secondary connection is established, when transfer of data over the primary communication path is no longer feasible, for example when there is a break in the path or when there is a malfunction of equipment associated therewith. [0015] For ease of description, transfer of data is referred to, hereinafter, simply as being transferred over the primary and secondary communication paths. Those skilled in the art should however appreciate that, for data transfer over each communication path to be meaningful, the data transfer should be over the logical connection established on the communication path according to a suitable communication protocol.

[0016] Advantageously, with the above-described method, a shorter delay is experienced for transfer of data on the secondary communication path as compared to that in the prior art, where it is necessary to first establish a connection before transfer of data over that connection on the secondary communication path is possible.

[0017] The number of distinct communication paths, i.e. paths not sharing any common section, in the network may be twice the number of user units as described above. Alternatively, the number of distinct communication paths in the network may be less than twice the number of user units such that, at least one communication section is shared by more than one communication path, and thus more than one user unit. The common communication path may be used as a primary and/or secondary communication path by the user units that are assigned that path. In other words, when the user units are individually assigned their respective data transfer connections, it is possible that more than one user unit may be assigned that path for establishing only their respective primary connections thereon. It is also possible that more than one user unit may be assigned that path for establishing only their respective secondary connections thereon. It is further possible that some user units may be assigned that path for establishing their respective primary connections thereon while other user units are assigned that same path for establishing their secondary connections thereon. [0018] The above-described method may further include assigning a protected information rate (PIR) and a committed information rate (CIR) to each user unit. The PIR and the CIR are indicative of a first maximum amount of data and a second maximum amount of data, respectively, that are transferable between that user unit and the first communication node. The assignment of PIR and CIR to the user units is carried out such that, for each path of the network, the bandwidth of the communication path is able to transfer the second maximum amount of data on that path for all those user units for which the path is assigned as a primary communication path, and to transfer the first maximum amount of data on that path for all those user units for which the path is assigned either as a primary communication path or as a secondary communication path. It should be noted that the bandwidth of a communication path is limited by the bandwidth of a critical section thereof, which might be a section shared by more than one communication path as discussed above.

[0019] In this manner, when only user units eligible for using the communication path as a primary communication path use the communication path, data amounting to their respective CIR are guaranteed to be transferable over the communication path. In a worst case scenario when it is not possible for all user units eligible for using or assigned the communication path as a secondary communication path, for transferring data over their respective primary communication paths, due to all being simultaneously out- of-order, transfer of data would be switched from over their respective primary communication paths to over that one single communication path. In such a scenario, the user units using the single communication path would be using it either as a primary communication path or as a secondary communication path and data amounting to their respective PIRs are guaranteed to be transferable over the communication path. [0020] When transferring data over a communication path, it is first determined if the communication path is required to carry data of any user unit using the communication path as a secondary communication path. If it is determined that the communication path is not required to carry data of any user unit using that communication path as a secondary communication path, bandwidth is allocated for the transfer of data for user units using that communication path as a primary communication path, subject to the CIR of those user units. In this way, the CIR of each user unit is met to ensure a guaranteed bandwidth for each eligible user unit. However, if it is determined that the communication path is required to carry data of any user unit using that communication path as a secondary communication path, bandwidth is allocated for the transfer of data for user units using that communication path either as a primary communication path or as a secondary communication path, subject to the PIR of those user units. In this way, the PIR of each user unit is met to ensure a protected bandwidth for each eligible user unit. [0021] After bandwidth of the communication path is allocated as described above, i.e. subjected to the CIR or PIR depending on the user units using that communication path, it is possible that there is bandwidth of the communication path remaining for further allocation. Such would be the case when not all eligible user units that are assigned the communication path as a primary communication path use that communication path, or even when all such user units use that communication path, not all use that communication path for transferring of the maximum amount of data as specified by their respective CIRs. Similarly, there would be remaining bandwidth when not all eligible user units that are assigned the communication path as either a primary communication path or secondary communication path use that communication path, or even when all such user units use that communication path, not all use that communication path for transferring of the maximum amount of data as specified by their respective PIRs. The remaining bandwidth can then be allocated for the transfer of any additional data for those eligible user units using the path that are in need of more bandwidth. The additional data is over and above those for which bandwidth is allocated earlier subjected to the CIR and PIR as described above. [0022] Optionally, the above-described method may include assigning a burst information rate (BIR), in addition to the PIR and CIR, to each user unit. This BIR is indicative of a third maximum amount of data that can be transferred for that user unit. The remaining bandwidth can then be allocated to each user unit, subject to its BIR. [0023] The remaining bandwidth may be allocated to each user unit for transfer of the additional data in proportion to the difference between the CIR and PIR, the difference between the BIR and PIR, or the difference between the BIR and CIR of the user unit. Such a method has the advantage that each eligible user unit gets a share of the remaining bandwidth. The method however results in fragmentation wherein allocated bandwidth does not fall at frame boundaries but rather in between frame boundaries. If only complete frames are transferable, a portion of the allocated bandwidth corresponding to a partial frame would be wasted. [0024] To prevent such wastage of allocated bandwidth, the bandwidth allocation may preferably take into account frame boundaries so that allocated bandwidth may fall thereat rather than therebetween so that only bandwidth corresponding to complete data frames are allocated to each user unit. [0025] Alternatively, the remaining bandwidth may be allocated in a round- robin manner to each user unit for transfer of the additional data, subject to its CIR or BIR. Bandwidth allocated in this manner would also suffer from fragmentation unless allocation is also frame-boundary based. Preferably, when allocating remaining bandwidth in a round robin manner, the remaining bandwidth may be allocated according to the actual bandwidth requirement of each user unit to avoid fragmentation. In other words, the bandwidth is allocated so that all data for the user unit is transferable. [0026] The above-described bandwidth allocation scheme has several advantages. Bandwidth is protected and guaranteed on a per-user unit basis. This means all the user units that are connected via the same communication path can have different levels of reliability and amount of protected and guaranteed bandwidth. Consequently, service level agreements with the user units can be more flexible. Unprotected or non-guaranteed bandwidth of individual user units could be degraded gracefully and not denied bandwidth completely when a fault occurs or when there is less unused bandwidth for non-bandwidth guaranteed user units, respectively. That is, the bandwidth allocated to such users is reduced to a level that is still sustainable but not zero.The scheme allows the bandwidth reserved for protection to be used by other user units during normal operation when no fault has occurred. The scheme allows unutilized bandwidth reserved for user units to meet their CIR to be re-allocated to other user units for them to burst beyond their CIRs. The service level agreement is also simple, pragmatic and enforceable. [0027] According to another aspect of the invention, there is provided a communication system in which the above-described method is implementable. An example of such a communication system is an optical access network, wherein the first communication node is an optical line termination (OLT) and the user units are optical network terminations (ONTs), each being connected to the OLT via two separate communication paths, which are generally known to those skilled in the art as passive optical networks (PONs). Alternatively, the user units may be network terminations (NTs), each being connected to the OLT via at least one optical network unit (ONU), wherein each ONU is connected to the OLT via a respective pair of PONs. The NTs may be connected to an ONU via any suitable access technologies, including but not limited to, xDSL, cable modems and switched Ethernet.

[0028] For such a communication system, establishing the data transfer connections includes ranging and registering each ONT and/or ONU with the OLT on the respective pair of PONs. Subsequently,. the OLT polls the ONT and/or ONU to find out if there is data thereat for upstream transfer over the PONs and allocates bandwidth of the PONs according to the above-described method. Each of the primary connection and the secondary connection may be maintained by the ONT and/or ONU indicating to the OLT, via a message sent thereto, that there is no data for transfer on the respective optical links. Alternatively, the ONT and/or ONU may simply ignore the poll by not responding thereto. Since downstream data is available for transfer at the OLT itself, the OLT is aware of the amount of data for transfer to each ONT and/or ONU and would allocate the bandwidth accordingly. The above- described method is particularly advantageous for such an optical access network since ranging and registration is time consuming and results in an unnecessary delay for transfer of data to be effected in the PON when transfer of data is switched thereover. [0029] The communication system may not be restricted to one configured in a multidrop topology as in the case of the optical access network described above. The communication system may also be other types of networks, such as networks connected using copper cables, which are configured according to any suitable topologies, such as but not limited to, a bus, a star or a ring topology. An application of a bus type network in which the invention can be used is one in which a host computer is connected to terminals or terminal clusters at several locations.

[0030] Furthermore, the communication system may be a wireless network wherein the above-described method can be used by a mobile station in an overlap region of two coverage areas of two respective base stations for establishing and maintaining two separate connections with the two base stations.

BRIEF DESCRIPTION OF DRAWINGS

[0031] The invention will be better understood with reference to the drawings, in which: Figure 1 is a schematic drawing of a prior art unprotected PON system, to which is connected an OLT and multiple ONUs; Figure 2 is a schematic drawing of a prior art protected PON system wherein each ONU is connected to an OLT via two PONs; Figure 3 is a schematic drawing illustrating the 1 :1 switching architecture for both upstream and downstream transmission between an ONU and the OLT in Figure 2; Figure 4 is a schematic drawing illustrating the 1+1 switching architecture for both upstream and downstream transmission between an ONU and the OLT in Figure 2; Figure 5 is a schematic drawing of a X:N protection PON system on which both the 1 :1 and the 1 +1 switching architectures in Figures 3 and 4 can be implemented; Figure 6 is a schematic drawing of a frame format for a 155.52/155.52 Mbits/s APON; Figure 7 is a schematic drawing of a protected PON system according to an embodiment of the present invention; Figure 8 is a flowchart of a sequence of steps for transferring data between a protected ONU and an OLT 8 in the protected PON system in Figure 7; Figure 9 is a schematic drawing of a polling cycle wherein each of a number of ONUs transmit data in its respective transmission window according to a cyclic upstream transmission sequence; Figure 10 is a schematic drawing of a single transmission window of an ONU in Figure 9, showing a transmission sequence of a number of subscribers connected to the ONU; Figure 11 is a schematic drawing of a PON frame structure that may be used for reporting queue statuses of the subscribers connected to the ONUs in Figure 9; Figure 12 is a schematic drawing of another protected PON system; and Figure 13 is a graph of the number of silver subscribers against the number of gold subscribers for the protected PON system in Figurel 2 under different conditions.

DETAILED DESCRIPTION

[0032] Figure 7 shows a protected communication system 30 having multiple PONs 14. Each PON 14 is configured in a multidrop topology to form a point-to-multipoint optical access network. This network includes passive optical components, such as couplers/splitters 6 with no active elements. [0033] Data transmissions in the system 30 are by an optical line termination (OLT) 8 at a headend and multiple optical network units (ONUs) 10 at the other endpoints of each PON 14. The system 30 may be used to implement a fiber-to-the-home (FTTH), fiber-to-the-building (FTTB), fiber-to- the-cabinet (FTTCab), or fiber-to-the-curb (FTTC) subscriber access network. The OLT 8 resides in a local exchange (not shown), acts as the central controller, and connects the subscriber access network to a backbone of a larger network (not shown). Each ONU 10 may reside at the curb or on subscriber premises to provide a combination of data, voice, video and other services to the subscribers. For the case in which an ONU 10 also provides the user port or network termination function, the ONU 10 is more commonly referred to as an optical network termination (ONT) 10. In general, multiple network terminations (NTs) 32 (Figure 12) will be connected to an ONU 10 via various access technologies, such as xDSL, cable modems and switched Ethernet. [0034] Each ONU 10 is connected to the OLT 8 via a PON 14A-14D serving as a communication link or path therebetween. If the PON 14 A-14D is the only PON A-14D between an ONU 10 and the OLT 8, the ONU 10 is considered as unprotected. However, if the ONU 10 is connected to the OLT 8 via at least another PON 14 A-14D, the ONU 10 is considered as protected. In the protected communication system 30 in Figure 7, ONU 1, ONU 2, and ONU X are protected while ONU 3 is unprotected. Each PON 14 A-14D, for example the tagged PON 14C, i.e. PON 14 under observation, which is shown in thicker lines in Figure 7, may serve as a primary or working PON 14C (for ONU 1, 3 and X) and as a secondary or protection PON 14C (for ONU 2). In Figure 7, the tagged PON 14C is the primary PON for ONU X and the other PON 14D to which ONU X is connected is the secondary PON for ONU X. With such an arrangement, an NT 32 connected to ONU X is connected to the OLT 8 via two separate PONs 14C, 14D.

[0035] Bidirectional transmission between the OLT 8 and each of the ONUs 10 can be implemented in a number of ways. One way is by having two parallel PONs, one for downstream transmission from the OLT 8 to the ONUs 10 and the other for upstream transmission from the ONUs 10 to the OLT 8. This configuration is known as a "simplex working" configuration according to ITU-T terminology. A more economical solution, known as a "diplex working" configuration, is to use different wavelengths for downstream and upstream transmissions on a single PON. Another solution, known as a "duplex working" configuration, provides bi-directional transmission using the same wavelength. The invention is applicable to any of the above configurations.

[0036] The ONUs 10 can receive all downstream transmissions by the OLT 8. An ONU 10 is able, with a suitable downstream frame format, to identify the data meant for it by the position of the data in a frame or by means of the ONU identification in header information in the frame. Due to the directional property of the passive splitters 6 in the PONs 14, transmission of an ONU 10 will not typically be received by other ONUs 10 but only by the OLT 8. However, multiple ONUs 10 transmitting at the same time may prevent the OLT 8 from receiving their transmission properly. A multi-access protocol is thus used to arbitrate different ONUs' upstream transmission.

[0037] There are several methods to facilitate communications between the OLT 8 and an ONU 10, and amongst the ONUs 10. These methods include wavelength division multiplexing (WDM), time-division multiple access (TDMA), subcarrier multiplexing and code division multiple access (CDMA). In WDM, a wavelength in the 1550nm window (1480-1580 nm) is used for the OLT 8 to broadcast its transmission to the ONUs 10 in the downstream direction and a wavelength in the 1310nm window (1260-1360 nm) is used for all the ONUs 10 to transmit their data to the OLT 8.

[0038] Although subcarrier multiplexing or code division multiple access (CDMA) may be used, time division multiple access (TDMA) is still the most efficient and cost effective multi-access solution for PON type of networks. This multi-access method is specified by ITU-T for the broadband PON (B- PON) system based on Asynchronous Transfer Mode (ATM) and advocated by the IEEE 802.3ah for the Ethernet Passive Optical Network (E-PON). [0039] A sequence 40 for transferring data between a protected ONU 10 (for example ONU X in Figure 7 and the OLT 8 to which ONU X is dual-homed to via the primary PON 14C and the secondary PON 14D is described next with the aid of the flowchart in Figure 8. The sequence 40 starts in a START step 42 and proceeds to an ESTABLISH AND MAINTAIN CONNECTIONS step 44. In this step 44, the ONU performs separate initialization processes with respective PON line terminators (LTs) 46 of the OLT 8. Each initialization process involves a registration and ranging operation known to those skilled in the art. The registration and ranging operations may be, for example, similar to those used in B-PON and E-PON systems. The initialization process allows ONU X to establish and maintain a first connection and a second connection on the primary PON 14C and the secondary PON 14D respectively for the transfer of data thereon between the OLT 8 and the ONU X During the initialization process, the respective PON LTs 46 in the OLT 8 assign the ONU X with respective identifiers by which ONU X uses to communicate with the OLT 8. It is also during this initialization process that the two PONs 14 used by an ONU X are designated as the primary PON 14C and the secondary PON 14D for ONU X [0040] The sequence 40 next proceeds to a TRANSFER DATA step 48, wherein data is exchanged between the OLT 8 and ONU X over the primary PON 14C. Data is transferred between the OLT 8 and ONU X using any suitable protocol. One such protocol involves the OLT 8 polling ONU X, and ONU X responding by indicating to the OLT 8, via a message, the amount of data of ONU X queued thereat for upstream transmission to the OLT 8. Other ONUs 10 connected to OLT 8 via the primary PON 14C will similarly indicate to the OLT 8 their respective queue statuses. Based on these ONU indications or reports, the OLT 8 allocates bandwidth of the primary PON 14C to the various ONUs 10. The bandwidth allocation scheme will be described in more details shortly. While exchanging data with the OLT 8 over the primary PON 14C, ONU X maintains the second connection on the secondary PON 14D by indicating to the OLT 8, when polled thereby via a message sent thereto, that no data is available for transmission over the secondary PON 14D. Such a report and grant scheme is described in more details in U.S. Patent 6,546,014, Kramer et al., entitled "Method and System for Dynamic Bandwidth Allocation in an Optical Access Network." Other report and grant mechanisms, for example those used in the broadband passive optical network (B-PON) and the Ethernet passive optical network (E-PON) may also be used.

[0041] The sequence 40 next proceeds to a PRIMARY PON DOWN? decision step 50, wherein the primary PON 4C is monitored to determine if it is feasible for transfer of data thereon. The primary PON 14C may be detected to be unavailable, for example, by monitoring the PON 14C for the presence or loss of signal thereon. Other means of detecting a fault condition of the primary PON 14C are also possible. Once a fault is detected, a transmission convergence function (not shown) in the OLT 8 governing bandwidth allocation for multi-access will be notified. If it is detected in this decision step 50 that the primary PON 14C is up or operational, the sequence 40 returns to the TRANSFER DATA step 48 to continue transferring of data over the primary PON 14C. Transfer of data over the primary PON 14C terminates when it is detected in the PRIMARY PON DOWN? decision step 50 that the transfer of data over the primary PON 14C is no longer feasible. [0042] When the primary PON 14C is detected to be down, the sequence 40 proceeds to a SWITCH TRANSFER OF DATA step 52, wherein transfer of data is switched from over the primary PON 14C to over the secondary PON 14D. Since connections over the primary and secondary PONs 14C, 14D are set up earlier in the ESTABLISH AND MAINTAIN CONNECTIONS step 44, transfer of data over the secondary PON 14D is effected instantaneously without having to first establish a connection thereon. ONU X has to merely indicate to the OLT 8 when polled that it now has data available for transmission over the secondary PON 14D. With this method, only the queue status information sent over the secondary PON needs to be changed; no registration and ranging of ONU X with the OLT 8 over the secondary PON 14 is required. The sequence 40 ends in an END step 54 when no data is available for transfer between ONU X and the OLT 8. [0043] The above steps are similarly performed for ONU 1 and ONU2 to allow an NTs connected to either one of ONU 1 and ONU 2 to be each connected to the OLT 8 via two separate PONs 14A-14D. The ONUs may share one or two PONs 14A-14D. In Figure 7, ONU 1 , ONU 2, ONU 3 and ONU X share a common PON 14C. This common PON 14 is assigned as a primary PON 14C for ONU 1 , ONU 3 and ONU X and as a secondary PON for ONU 2.

[0044] Advantageously, with the above-described method, a shorter delay for transfer of data on the secondary PON 14D is experienced by ONU X as compared to that in the prior art where it is necessary for a registration and ranging process to establish a new connection on the secondary PON 14D before transfer of data over the secondary PON 14D is possible. It should be noted that each ONU may be connected to the OLT 8 via more than two PONs 14A-14D to further increase the survivability of data transfer between the ONU 10 and the OLT 8. [0045] The manner in which the OLT 8 allocates bandwidth of a PON for transfer of data, such as the primary PON 14C in Figure 7, amongst several ONUs 10 connected thereto, or more specifically amongst subscribers 60 connected to these ONUs 10, is next described. The subscribers 60 share the primary PON 14C by multiplexing their data thereon. This same PON is also used as a secondary PON 14D as described above. For ease of description, a subscriber using the PON as a primary PON 14C and a secondary PON 14D is referred to as a primary subscriber and a secondary subscriber respectively of that PON. The primary subscribers using the PON 14C as a primary PON are shown in Figure 7 as being connected to the OLT 8 by solid unbroken lines throughout. All other subscribers connected to the OLT 8 via the same PON 14C but without such a connection are secondary subscribers of the PON 14. These secondary subscribers are shown in Figure 7 to be connected to the OLT 8 by a line that is at least partially broken. A subscriber 60 that is dual-homed to a single ONU 10 is a primary subscriber of the primary PON 14A-14D and a secondary subscriber of the secondary PON 14A-14D of the ONU 10 to which it is connected to. To provide greater flexibility and protection, i.e. to improve survivability, subscribers 60 who need additional protection could also be dual-homed or multi-homed to two or more ONUs 10. This configuration will be discussed later. A subscriber 60 that is connected to two ONUs 10 (such as a primary ONU 3 and a secondary ONU 2 for subscriber 3, 1) sharing a primary PON 14C of the primary ONU 3 remains a primary subscriber of that PON 14C. [0046] Bandwidth allocation is based on a service level agreement (SLA) between a service provider and each subscriber 60. The SLA includes a protected information rate (PIR) and a committed information rate (CIR) that are indicative of a first maximum amount of data and a second maximum amount of data, respectively, that are transferable between that subscriber 60 and the OLT 8.

[0047] The PIR is indicative of or specifies bandwidth that can be guaranteed or protected when all primary and secondary subscribers of a PON 14A-14D use the PON 14A-14D. The CIR specifies bandwidth that is guaranteed during normal PON operation when only primary subscribers of the PON use the PON 14. The difference between the CIR and the PIR of a subscriber is the amount of bandwidth of the subscriber 60 that is committed but not protected. The SLA with a subscriber 60 may further include a burst information rate (BIR), which is indicative of a third maximum amount of data that can be transferred for that subscriber 60 on the PON 14A-14D. The BIR can be implemented easily by limiting the data transmission rate on the ONU port of the ONU 10 to which a subscriber 60 is connected. Generally, the relationships between the BIR, CIR and PIR are as follows: BIR ≥ CIR ≥ PIR. [0048] By specifying the bandwidth requirement in this manner, a service provider can, for example, offer the following service classes with PIR, CIR and BIR assigned as follows: • Protected and guaranteed bandwidth service with PIR = CIR; BIR = port rate; • Unprotected guaranteed bandwidth service with PIR = 0; CIR allocated based on each subscriber's requirement; BIR = port rate; • Non-guaranteed (best-effort) bandwidth service with PI R = 0; CI R = 0; BIR = port rate.

[0049] A bandwidth allocation scheme for allocating bandwidth to a subscriber on the tagged PON 14C in Figure 7 is next described. The scheme determines a first bandwidth to allocate to each active primary subscriber, subject to the committed information rate (CIR) of the subscriber 60 when there is no active secondary subscriber on the PON 14C. An active subscriber is one with data at its ONU 10 for transfer over the PON 14C or to which data at the OLT 8 is to be delivered to. However, if there is at least one active secondary subscriber 60 on the PON 14C, the scheme determines a first bandwidth to allocate to each subscriber subject to a maximum of the protected information rate (PIR) of the subscriber. After determining the respective first bandwidths for allocation to the active subscribers, the scheme determines the remaining bandwidth available on the PON 14C. Subsequently, the scheme determines a second bandwidth from the remaining bandwidth to allocate to each active subscriber having additional data for transfer, i.e. data over and above that for which the first bandwidth is allocated. The respective first bandwidths and second bandwidths are then allocated to the active subscribers. The sum of CIRs of all primary subscribers 60 of the PON 14 and the sum of the PIRs of all primary and secondary subscribers 60 of the PON 14 are assigned such that they do not each exceed the bandwidth or data rate of the PON 14C.

[0050] To analyze bandwidth allocation using the above scheme mathematically, let subscriber (ij) denote a subscriber 60 that is connected to Port./ of an ONU / connected to the tagged PON 14C in Figure 7. The tagged PON 14C may be the sole PON 14C (in the case of an unprotected ONU, such as ONU 3), the primary PON 14 (ONU 1 , 3 and X) or the secondary PON 14 (ONU 2) for the ONU 10 to which a subscriber (/,/) is connected to. The following description will be based on transmission in the upstream direction. Transmission in the downstream direction is similar but simpler, and will be briefly described later. Let P_ltj be the PIR, Qj be the CIR and B _i be the BIR for the respective subscribers (i,j) in the upstream direction. A primary subscriber P\R,P_U , a secondary subscriber PIR, P_i a primary subscriber

CIR, c\ . , and a secondary subscriber CIR,C. . , may be defined as follows: [P_{t j} if the taggedPONis the primaryPON for subscriber^^', j) 0 if thetaggedPONisthesecondary,protectionPONforsubscriber(z, j)

(P_{( j} if the taggedPONis the secondary ^rotectionPON for subscriber^^', /) JO if the taggedPONis theprimaryPONforsubscriber(z, j)

[C_u if thetaggedPONistheprimaryPONforsubscriber(z, j)

C_{; j} = J ' (3) 0 if the tagged PON is thesecondar protectionPONfor subscriber (i, j)

[C_; if thetaggedPONisthesecondary;protectionPONforsubscriber(t, j) 10 if the tagged PONis the primaryPON for subscriber (i, j)

[0051] Let D_b be the one-way delay bound for delay-sensitive services. To ensure that the one-way delay bound is less than Db, each ONU needs to have a response time of, D_r, that fulfills the following condition: D_r ≤ D_b - D_p - D₀ - D_q (5) wherein D_p is the one-way propagation delay from an ONU to the OLT, D_o is the fixed processing delay within the ONU, and D_q is the queuing delay at the ONU.

[0052] Assuming that each ONU 10 assigns a dedicated buffer or queue

(not shown) to each subscriber 60 it supports, the upper bound of queuing delay, D_q, can be computed if the buffer size allocated to the subscriber is known a priori. The fixed processing delay, D₀, depends on the design of the ONU and several bit-time is achievable at a data rate of 1 Gbps. The ITU-T Recommendation G.114 specifies that one-way transmission time, D_b, should not exceed 1.5ms for an access network, which can be taken as the default value for D_b. For simplicity, it is assumed that the response time, D_r, is 1 ms although other values are also possible. Substituting these values of D_r and D_b in equation (5) above results in a total delay budget of 0.5ms for D_p, D₀ and D_q, which can be met easily. [0053] Figure 9 shows a cyclic upstream transmission sequence for m ONUs, each of which transmits data during its granted transmission window within a polling cycle. The cyclic sequence can simply be based on the ONU identifiers although other sequences are also possible. The sequence and duration for transmission by each subscriber 60 connected to an ONU 10, within the transmission window of the ONU 10, is shown in Figure 10. Similarly, the subscriber identifier or other suitable means may be used to determine the transmission sequence of subscribers. During each polling cycle (upstream), the queue status of individual subscribers connected to an ONU 10 will be reported by the ONU 10 to the OLT 8. The manner in which the queue status is reported will be described later. The queue status and other information will be used, by the OLT 8, to determine how much bandwidth or time the OLT 8 will allocate to each subscriber 60 for transmitting its data in the ONU queue in a polling cycle n. [0054] Additional notations that will be used to mathematically illustrate the bandwidth allocation scheme, are listed below: m: Total number of ONUs supported by the tagged PON (including ONUs that access the tagged PON as its secondary PON); f. : Number of subscribers supported by an ONU /; S: Data rate of tagged PON; w" : Total transmission time allocated to ONU / in transmission cycle n; W_t : Maximum transmission time for ONU / for each transmission cycle; t" : Size or duration of polling/transmission cycle n; u" : Transmission time allocated to subscriber (/, j) in polling cycle n; v" : Additional transmission time, over and above «"_;. , which may be allocated to subscriber (/,/) in polling cycle n to meet its CIR requirements; y" : Additional transmission time, over and above v"_; , which may be allocated to subscriber (i,j) in polling cycle n to burst beyond its CIR; q _j : Time needed to transmit, in polling cycle n, all the frames of subscriber (/,/) already queuing at ONU / at the time ONU / reports its queue status to the OLT in cycle n-1 ; U Transmission time corresponding to the PIR of subscriber (/^',/) for each polling cycle; and g: Guard time between ONU's transmission in a polling cycle.

[0055] Information on the PIR, CIR and BIR of individual subscribers can be obtained by the OLT 8 from predetermined values stored in a central database or individually configured. That is, some operators may decide to offer different grades of services, for example, Platinium, Gold and Silver, each having a set of fixed PIR, CIR and BIR. In this case, the PIR, CIR and BIR values will be obtained from a table in the database. The operators may also choose to allow each customer, such as a major customer, to request for a desired set of PIR, CIR and BIR according its need and/or budget. In this case, the information needs to be configured for that particular customer. [0056] For an arbitrary subscriber (/^', j), transmission time, £7^. , corresponding to the above-described first bandwidth when there is at least one active secondary subscriber, is set based on the PIR and the tagged PON's data rate, S, using the following equation:

U_t = ^-(D_r -G) (6) wherein G = mg (7) is the total overhead in each polling cycle. [0057] The total allocated transmission time, uⁿ , referred to above as the first bandwidth and the second bandwidth, can be set to either £/_u+ "_; + y"_; or a", , whichever is smaller, i.e., u^* _} = min(l/, . + v'_} + y_u' ,q_u" ) (8)

[0058] q- is obtainable by the OLT 8 from the queue status information sent by ONU / to the OLT using a report message. It represents the bandwidth that is required by the subscriber for transmission of all of its queue data. The manner in which v"_;. and y"_} is computed will be described later.

[0059] To avoid fragmenting a frame (also known as a cell) for the case of U^+v. + y _j) < #". wherein the total allocated bandwidth is insufficient for transfer of all queued data of the subscriber (/, j), the allocated transmission time, u. , may be set to a value just enough for the subscriber (/, J) to send complete frames whose total transmission time when added up does not exceed U^+v. + y. ). Let z" _>kbe the transmission time of frame k

(numbered in terms of their transmission order) of the subscriber (/, j). For the case of (Ui + v- _j + y- ) < q" _t let frame K be the frame such that K K+l

22zz"" _{JC ≤}≤ L U_{ii +}+ v^«",_j ++ ^ y " << 22 z" _ιk , Equation (8) above can be modified as 1 1 follows: u ■ = K JC+l ^■ (9) A ^{lf Q}u U_t,j + v" + y and | _≤ U_UJ + ^ + y"_, . < J i.j,k

[0060] To allocate u" based on the exact amount of time required in accordance with Equation (9) results in additional overheads that are required to describe the length of each frame in the queue at the ONU for subscriber (i,j). The overheads can be reduced by defining the length as a number of blocks, each of which is some integer multiple of bytes. [0061] w" , w, and t"can be represented in terms of w". and u_u using the following equations:

[0062] If the tagged PON is the primary PON for subscriber (/ ), the actual value of ^ will be conveyed to the OLT 8 in the report message or a similar reporting mechanism. If the tagged PON is the secondary PON for subscriber (i,j), an ONU 10 either does not send any report message to the OLT 8 or report a zero value for q"_} for subscriber (ij). Note that only one report message/PLOAM cell is required in each cycle for an ONU to report the queue status of all the subscribers connected to it.

[0063] By allocating bandwidth for each PON in the system 30 in this way, there will be sufficient bandwidth in the system 30 to meet all the subscribers' PIR requirement even if a PON 14 fails completely. If the primary PON 14 of a subscriber 60 fails, the protected data of the subscriber 60 is carried over a secondary PON 14 of the subscriber 60.

[0064] The manner in which v"_yandy"_y in Equations (8) and (9) can be computed is next described. According to the definition above, v"_y is the extra time or bandwidth, over and above w". , allocated to subscriber (ij) to meet its

CIR requirement. Any unallocated or remaining bandwidth may be allocated to the subscriber for this purpose. Additionally, bandwidth utilization can be increased further by also making use of the reserved protection bandwidth that is not utilized during normal operation. [0065] During normal operation when there is no active secondary

subscriber 60 on the PON 14, the total reserved bandwidth, , and

m f, unallocated bandwidth, S - P_{l }} , could be allocated to the active primary i-l j-1 subscribers 60 on the tagged PON to transmit up to their C_I . Hence,

^m f< since Y J° ₇ is zero during normal operation. «-l -1

[0066] The above result is not surprising; it merely suggests that the sum of the CIR for all primary subscribers, i.e. subscribers using the tagged PON as the primary PON, must be less than the data rate, S, of the PON as described above. The CIR for each subscriber may be set depending on the types of services and charging model the service provider adopts. Only simple arithmetic is required to determine if it satisfies Equation (13). Since "₇ is the extra time or bandwidth reserved in cycle n to account for the additional bandwidth "_; - J°" required by the subscriber (/, j), it can be computed using the following equation:

[0067] For the computation of y"_; , which is the extra bandwidth that may be allocated to a subscriber (ij) due to other subscribers not requiring their CIR on the primary PON, the extra or excess time, e" , or remaining bandwidth of the PON 14, in cycle n that is still available after taking into consideration the CIR requirement of all active primary subscribers in cycle n has to be determined. This excess time, e" , can be expressed as follows: m fi eⁿ = D_r - G - ^ u?. (15) ^j-1 J-1 wherein ^ = min(rJ_{i +} ^,^) (16)

[0068] For the case of (U_t + v"_y) < q"_}. , i.e. there is more queued data than is transferable based on the CIR, the following equation can be used to replace Equation (16) if the amount of time allocated is required for transmitting an integer number of complete Ethernet frames or ATM cells etc.

Note that ii". is computed in a similar way as u"_}. in Equation (9).

[0069] There are many different ways to allocate the extra time, e" , available in cycle n to the active subscribers. The following categories of subscribers do not need any more bandwidth and thus do not require any of the extra time, e" : • Subscribers who use the tagged PON as their secondary PON, i.e. inactive secondary subscribers; and • Subscribers whose q"_>}. ≤ U_t + vⁿ _j and hence u_t"_}. = q_t"_j according to Equation (16). [0070] The extra time could be distributed to those active subscribers who do not belong to the above categories in, but not limited to, one of the following ways: 1. Proportional to the difference between their BIR and CIR, i.e., B"_}. - C"_; ; 2. Proportional to the difference between their PIR and CIR, i.e., C". ~P"_} ; 3. Proportional to the difference between their BIR and PIR, i.e., B"_t] -P _} ; 4. Given all the extra time each active subscriber needs, i.e., min(0, <?,". -U_u -v".)in a predetermined order until the time available for allocation is completely allocated. For example, the time may be allocated in a round robin manner or sequence based on the subscriber identifier or the ONU identifier. The time available in a transmission cycle is allocated to as many subscribers as possible until it runs out. The last subscriber, for example a subscriber x, who is allocated time in a transmission cycle is noted. In a next transmission cycle, any excess time available for allocation is allocated to a subscriber y next in the sequence. In this manner, all subscribers get to share extra time that is available and no subscriber is deprived of an opportunity to be allocated extra time that is available. [0071] Options 1 to 3 above will in most cases result in more wastage as compared to option 4 because the extra time allocated to each subscriber is unlikely to fall at boundaries of the frames to be transmitted by the subscriber but in between frame boundaries. Hence, option 4 is preferred to options 1 to

3 for achieving higher bandwidth utilization. [0072] The bandwidth allocation method according to the above-described embodiment of the invention has several advantages. Bandwidth is protected and guaranteed on a per-subscriber basis. This means all the subscribers that are connected via the same ONU can have different levels of reliability and amount of guaranteed bandwidth. Consequently, service level agreements can be more flexible. Unprotected or non-guaranteed bandwidth of individual subscribers could be degraded gracefully and not denied bandwidth completely when a fault occurs or when there is less unused bandwidth for non-bandwidth guaranteed subscribers, respectively. The method allows the bandwidth reserved for protection to be used by other users during normal operation when no fault has occurred. The method allows unutilized bandwidth reserved for subscribers to meet their CIR to be re-allocated to other users for them to burst beyond their CIRs. The service level agreement between the service providers and subscribers used is also simple, pragmatic and enforceable. [0073] Next, the overhead of reporting queue statuses in the bandwidth allocation scheme is analyzed. Figure 11 shows a PON frame structure that may be used for reporting the queue status of subscribers connected to an

ONU. It should be noted that other PON frame structures may also be used. A number of bits, say 8, could be used to report the number of subscribers supported by the ONU. The use of 8 bits for such a purpose limits the number of subscribers that can be supported per ONU to 256. Two bytes may be used for reporting the queue status for each subscriber. 13 of the 16 bits in these two bytes are used to indicate the number of bytes in the queue while the remaining 3 bits can be used for other purposes. This allows a maximum buffer occupancy of 8,192 bytes to be represented.

[0074] G in Equation (9) may be rewritten to cater to the bandwidth that is required for implementation of the frame format in Figure 11 as follows:

= m(gS ₊ 2) ₊ lβγ_iJ_i

= m(gS +32) + 16M bits (17) wherein

is the total number of subscribers supported by the tagged PON.

[0075] Guard time, g, is primarily determined by a clock and data acquisition (CDA) function of a burst mode receiver (not shown) at the OLT. For a 1 Gb/s PON, a guard time in the order of 1 μs or 1000 bits has sufficient margin for a transceiver with a slow CDA time. According to some OLT vendors CDAs requiring a guard time, g, in the order of only a few bits are available. However, for analysis purposes, the following assumptions are made: • the total number of subscribers, M, supported by a PON is 1000; • the total number of ONUs, m, in the PON is 32; • guard time, g, is 1 μs; and • D_r is 1 ms or 10⁶ bits.

[0076] The overhead due to guard time and frame header can then be computed as follows:

% overhead due to guard time, g =100%*32*10^"6*10⁹/10⁶ = 3.2% % overhead due to PON frame header =100% * (32*32+16*1000) /10⁶ = 1.7%

[0077] The results show that overhead due to the PON frame header is negligible and overhead due to guard time, g, of 1 μs is also insignificant. [0078] The effect of re-allocation of reserved but unused protection bandwidth is next analyzed. The case of a service provider offering the following three types of services to subscribers in a particular area served by two PONs, as shown in Figure 12, is considered. • Gold service - Protected and guaranteed bandwidth service wherein PIR = CIR = 5Mb/s; BIR = 5Mb/s • Silver service ~ Unprotected guaranteed bandwidth service wherein PIR = 0; CIR = 5Mb/s; BIR = 5Mb/s • Bronze service ~ Unprotected guaranteed bandwidth service wherein PIR = CIR = 0 Mb/s; BIR = 1 Mb/s

[0079] The guard time, g, between two ONUs' transmissions is 1 μs. For traffic engineering purposes, the service provider assumes that all the subscribers transmit at their BIR all the time. The following are also assumed: • Number of ONUs connected to the tagged PONs, m, is 32 (half of which regard a first PON as the primary PON and a second PON as the secondary PON, while the other half of the ONUs regard the first PON as their secondary PON and the second PON as their primary PON as shown in Figure 12); • Number of Gold service subscribers supported by each ONU: varies from zero to twelve, in steps of one; • Number of Bronze service subscribers supported by each ONU: none; and • Fault condition: One of the PONs has a fibre cut near the OLT. [0080] The total number of Silver service subscribers that can be supported for the following 3 cases are determined: A. When protection bandwidth is not allocated to other subscribers during normal operation, even though the protection bandwidth is unused (as practiced in the prior art); B. When the unused protection bandwidth is allocated to other subscribers during normal operation; and C. When fault has, occurred and all PIR data of Gold service subscribers has to be supported.

[0081] Let a_g, a_s and a_b be the number of Gold, Silver and Bronze service users per ONU, respectively. Let _g, p_s and p_b be the PIR, c_g, c_s and c_b be the

CIR and b_g, b_s and b_b be the BIR of the Gold, Silver and Bronze service subscribers, respectively. The overhead in each cycle due to guard time, g (for example, in seconds), and the PON frame overhead is:

G = m(gS+32) + 16M = = m(gS+32) + 16m(a_g+a_s+a_b) (19) where S is the data rate in bits/s. [0082] The bandwidth that is available after deducting the overhead due to guard time, g, and the PON frame overhead is: S' = S(DrS- G)/(DrS) = S - GIDr (20)

[0083] For case A above where bandwidth reserved for protection for those ONUs that regards a PON as the secondary PON cannot be used by Silver subscribers for meeting their CIR, the bandwidth available for supporting the Silver subscribers for each PON is S" = S'- ma_gP_g = S - G/Dr-magPg (21)

[0084] For 2 PONs and m ONUs, the number of Silver service subscribers that can be supported per ONU using the left-over bandwidth is α_s(CαseA) = max 0, .(22)

Note that [* j means the largest integer smaller or equal to *.

[0085] For case B above where bandwidth reserved for protection can be used for CIRs, during normal operation, the number of ONUs that see each PON as the primary PON is m/2; hence, the bandwidth left over for supporting the Silver subscribers supported by these m/2 ONUs is: S" = S' - ma_gP_g/2 = S - G/D_r- maφ_g/2 (23) [0086] The number of Silver subscribers that can be supported per ONU is a_s(CaseE) = min 0, .(24)

[0087] After a fiber cut near the OLT of one PON, all the m ONUs have to be supported by the surviving PON and the bandwidth left over for supporting Silver subscribers is: S" = S(Dr- G)/ D_r - mag g (25)

[0088] The number of Silver subscribers that can be supported per ONU is a_s(CaseC) = max 0, .(26)

[0089] For all the three cases, a_s is tabulated in Table A below. Table 1 : Silver subscribers that supported in different scenarios

[0090] Figure 13 shows a plot of a_s against a_g, for the results tabulated in Table A. As can be seen from the results, higher bandwidth utilization is achievable in case B.

[0091] Although the invention is described as implemented in the above- described embodiment, it is not to be construed to be limited as such. [0092] For example, the bandwidth allocation and protection scheme described above for a PON between an ONU and its OLT can also be implemented independently for a link between an NT and its primary ONU, i.e. the NT is connected to two separate ports of the primary ONU. It is not necessary that the scheme be restricted to the type of protocol used between an NT and an ONU; any suitable protocol may be used. [0093] For certain implementation scenarios in which there are many low- bandwidth subscribers that are connected to an ONU, the huge overhead of maintaining a queue for each subscriber and reporting the queue status to OLT may make the above-described bandwidth allocation scheme less attractive. To reduce these overheads, only a few subscribers who need large bandwidth needs to be given a dedicated queue each. Several types of shared queues could be provided to different categories of subscribers. In this case, all subscribers that share the same queue would be treated as if they are a single subscriber in the above-described bandwidth allocation scheme. This is similar to the forward equivalent class (FEC) concept of multiprotocol label switching (MPLS). [0094] As discussed above, to improve survivability for certain subscribers, their NTs could be dual-homed or multi-homed to two or more ONUs. For the case of a subscriber dual-home to two ONUs, say to Port j of ONU / as a primary access and Port y of ONU x as a secondary access, the subscriber is said to be subscriber (ij) with a non-zero PIR, CIR and BIR as well as subscriber (x,y) with a zero PIR, CIR and BIR during normal operation. Upon detecting a fault developed at the NT's primary access to its primary ONU, subscriber (x\y)'s PIR, CIR and BIR will assume the values of subscriber (ij) while the PIR, CIR and BIR of subscriber (ij) will be set to zero. If the two ONUs, to which the subscriber is dual-homed to, are attached to two disjoint primary and secondary PONs, the dual-homed subscriber can survive up to failures of three separate PONs.

Claims

CLAIMSWhat is claimed is:

1. A method for transferring data in a communication system, which comprises a first communication node and at least one user unit, each user unit being connected to the first communication node via a network of a number of communication paths such that, for each user unit, two distinct data communication connections, each on a different path, can be provided between the first communication node and the user unit, the method comprising, individually for each user unit: establishing and maintaining both data transfer connections simultaneously; assigning one of the data transfer connections as a primary connection and the other data transfer connection as a secondary connection between the first communication node and the user unit; transferring the data over a primary communication path on which the primary connection is established; and switching transfer of the data instantaneously from over the primary communication path to over a secondary communication path on which the secondary connection is established, when transfer of data over the primary communication path is no longer feasible.

2. The method according to Claim 1 , wherein the number of communication paths in the network is less than twice the number of user units such that, at least one communication path is shared by more than one user unit as a primary and/or secondary communication path.

3. The method according to Claim 2, further comprising: assigning a protected information rate (PIR) and a committed information rate (CIR) to each user unit that are indicative of a first maximum amount of data and a second maximum amount of data, respectively, that are transferable between that user unit and the first communication node, wherein said assignment is carried out such that, for each path of the network, the bandwidth of the communication path is able to transfer the second maximum amount of data on that path for all those user units for which the path is assigned as a primary communication path, and to transfer the first maximum amount of data on that path for all those user units for which the path is assigned either as a primary communication path or as a secondary communication path.

4. The method according to Claim 3, wherein transferring data comprises: determining, for each communication path, if the communication path is required to carry data of any user unit using the communication path as a secondary communication path; allocating bandwidth for the transfer of data for user units using the communication path as a primary communication path, subject to the CIR of those user units, if it is determined that the communication path is not required to carry data of any user unit using the communication path as a secondary communication path; allocating bandwidth for the transfer of data for user units using the communication path either as a primary communication path or as a secondary communication path, subject to the PIR of those user units, if it is determined that the communication path is required to carry data of any user unit using the communication path as a secondary communication path; and allocating any remaining bandwidth of the communication path for the transfer of any additional data, over and above those for which bandwidth is allocated, for those user units.

5. The method according to Claim 3 or 4, further comprising assigning a burst information rate (BIR) to each user unit that is indicative of a third maximum amount of data that can be transferred for that user unit, wherein the remaining bandwidth is allocated to each user unit, subject to its BIR.

6. The method according to Claim 4 or 5, wherein the remaining bandwidth is allocated in a round-robin manner to the user units for transfer of the additional data.

7. The method according to Claim 4 or 5, wherein the remaining bandwidth is allocated to each user unit for transfer of the additional data in proportion to the difference between the PIR and CIR, the difference between the PIR and BIR, or the difference between the CIR and BIR of the user unit.

8. The method according to any of Claims 4 to 7, wherein the bandwidth is allocated to a user unit for transferring only complete data frames.

9. The method according to any of Claims 1 to 8, wherein the communication system is an optical access network, the first communication node is an optical line termination (OLT) and the user units are optical network terminations (ONTs), each being connected to the OLT via two passive optical networks (PONs), and wherein establishing the data transfer connections comprises ranging and registering each ONT with the OLT on the respective PONs.

10. The method according to Claim 9, wherein maintaining the primary connection or the secondary connection comprises the ONT indicating to the OLT, via a message sent thereto, that there is no data for transfer on the respective PONs.

11. A communication system comprising: a first communication node; and at least one user unit, each user unit being connected to the first communication node via a network of a number of communication paths such that, for each user unit, two different paths are provided between the first communication node and the user unit for respective data communication connections, wherein, for each user unit, the respective data transfer connections on both paths can be established and maintained simultaneously wherein one of the data transfer connections can be assigned as a primary connection and the other data transfer connection as a secondary connection between the first communication node and the user unit, data can be transferred over one of the two paths on which the primary connection is established and which is designated as a a primary communication path; and transfer of the data is instantaneously switchable from over the primary communication path to over the other one of the two paths on which the secondary connection is established and which is designated as a secondary communication path, when transfer of data over the primary communication path is no longer feasible.

12. The communication system according to Claim 11 , wherein the number of communication paths in the network is less than twice the number of user units such that, at least one communication path is shared by more than one user unit as a primary and/or secondary communication path.

13. The communication system according to Claim 12, wherein each user unit can be assigned a protected information rate (PIR) and a committed information rate (CIR) that are indicative of a first maximum amount of data and a second maximum amount of data, respectively, that are transferable between that user unit and the first communication node, wherein said assignment can be carried out such that, for each path of the network, the bandwidth of the communication path is able to transfer the second maximum amount of data on that path for all those user units for which the path is assigned as a primary communication path, and to transfer the first maximum amount of data on that path for all those user units for which the path is assigned either as a primary communication path or as a secondary communication path.

14. The communication system according to Claim 13, wherein the first communication node is able to: determine if the communication path is required to carry data of any user unit using the communication path as a secondary communication path; allocate bandwidth for the transfer of data for user units using the communication path as a primary communication path, subject to the CIR of those user units, if it is determined that the communication path is not required to carry data of any user unit using the communication path as a secondary communication path; allocate sufficient bandwidth for the transfer of any data for user units using the communication path either as a primary communication path or as a secondary communication path, subject to the PIR of those user units, if it is determined that the communication path is required to carry data of any user unit using the communication path as a secondary communication path; and allocate any remaining bandwidth of the communication path for the transfer of any additional data, over and above those for which bandwidth is allocated, for those user units.

15. The communication system according to Claim 13 or 14, wherein each user unit can be further assigned a burst information rate (BIR) indicative of a third maximum amount of data that can be transferred for the user unit, wherein the remaining bandwidth is allocated to each user unit, subject to its BIR.

16. The communication system according to Claim 14 or 15, wherein the first communication node is able to allocate remaining bandwidth in a round- robin manner to the user units for transfer of the additional data.

17. The communication system according to Claim 14 or 15, wherein the first communication node is able to allocate remaining bandwidth to each user unit for transfer of the additional data in proportion to the difference between the PIR and CIR, the difference between the PIR and BIR, or the difference between the CIR and BIR of the user unit.

18. The communication system according to any of Claims 14 to 17, wherein the first communication node is able to allocate bandwidth to a user unit for the transfer of only complete data frames.

19. The communication system according to any of Claims 11 to 18 comprises a optical access network, wherein the first communication node is an optical line termination (OLT) and the user units are optical network terminations (ONTs), each being connected to the OLT via a pair of passive optical networks (PONs).

20. The communication system according to any of Claims 11 to 18 comprises a optical access network, wherein the first communication node is an optical line termination (OLT) and the user units are network termination (NTs), each being connected to the OLT via at least one optical network unit (ONU), wherein each ONU is connected to the OLT via a respective pair of passive optical networks (PONs).

21. The communication system according to Claim 19 or 20, wherein the ONT or the ONU is able to establish the data transfer connections by ranging and registering with the OLT on the respective PONs, and to maintain the primary connection or the secondary connection by indicating to the OLT, via a message sent thereto, that there is no data for transfer on the respective PONs.