Path-based restoration mesh networks

[0002] This invention relates to a technique for restoring service in a mesh network upon the failure of path.

[0003] Manufacturers of optical transport systems now offer an ever-increasing amount of available capacity on individual optical fibers. For this reason, achieving fast self-healing in a network from an accidental failure is of paramount importance. Presently, SONET/SDH ring networks comprise the most common type of self-healing network design. While SONET/SDH ring networks offer quick restoration and are easy to implement, such ring networks are not cost effective for many applications. For example, a highly connected network with most of the nodes each having at least three links requires more spare capacity when designed in a ring configuration than a comparable design with mesh configuration in which spare capacity is shared by the entire network more efficiently.

[0004] While mesh networks achieve more efficient use of spare capacity, restoration in a mesh network is more complex and slower than in a ring network. One type of mesh network restoration scheme is Link-Based Restoration (LBR) in which failed traffic is restored between the two nodes adjacent to the failure location regardless of the originating and terminating nodes of the failed signals. Another mesh restoration scheme is Path-Based Restoration (PBR) in which failed traffic is restored between the pairs of nodes where the failed signals originate and terminate regardless of the failure location. PBR requires less spare capacity to effect restoration as compared to LBR. On the other hand, PBR algorithms for establishing alternate paths when a failure occurs are more complex and therefore PBR may be slower, as compared to LBR.

[0005] Utilizing pre-computed routes can increase the PBR speed. One proposed restoration path computation approach employs node and link disjoint restoration paths. link- and node-disjoint restoration path is a path (i.e., a collection of individual channels in a series of links, each coupling a pair of nodes) that does not share any link or any intermediate node with the original path. However, accomplishing PBR using node-and link-disjoint restoration paths is inefficient in terms of maintaining sufficient spare capacity. On the other hand, it is necessary to determine which of the links on the signal path failed in order to implement a PBR method using unrestricted restoration paths. It is often problematic to make such a determination at the endpoint nodes. Further, computing a restoration path that is distinct from other restoration paths when multiple paths fail simultaneously is often not fast.

[0006] Thus, there is a need for a restoration technique that is not restricted to link-and node-disjoint paths, thus making the most efficient use of spare capacity. Additionally, there is a need for a mesh network restoration technique that affords restoration speeds comparable to those achieved in SONET/SDH rings. Further, there is a need for a mesh network restoration technique that is self-healing and self-managing.

[0007] Briefly, the invention comprises a method for restoring service in a mesh network formed of a plurality of nodes, at least two of which serve as end points for originating and/or terminating traffic. A plurality of links, each having one or more working channels and one or more restoration channels, selectively couple pairs of the nodes. The collection of connected channels in a series of links carrying traffic between a pair of end-point nodes comprises a path. In accordance with the invention, at least one of a pair of connected nodes monitor the channels in connected links to detect a link in failure, that is, a link having at least one failed channel carrying traffic. Upon detecting a link in failure, one node determines whether the link in failure includes at least one available channel (typically, a restoration channel) to carry the traffic unable to pass on the failed channel. If so, the one node signals the node at the other end of the link in failure to route traffic on the available channel in the link, thus achieving “localized” restoration in a timely fashion.

[0008] Unfortunately, not all links may possess an available channel. (Indeed, a multiple-channel link, when severed, will suffer a failure of all its channels.) In the event that a link in failure lacks available restoration capacity (i.e., one or more available channels), then the nodes, which detected the failure, alert the path end-point nodes to restore the failed paths. Each end-point node accesses a database storing a plurality of pre-computed paths corresponding to the failed path. Each pre-computed path corresponds to a pre-selected link and identifies, for an associated pair of end-point nodes coupled by a path having a link in failure, an alternate path that bypasses the link in failure. After accessing the database, each end-point node selects the alternate path corresponding to the link in failure. The end-point nodes then route traffic in accordance with the pre-computed alternate path information to restore the network.

[0009] FIG. 1 depicts a block schematic diagram of an exemplary mesh network;

[0010] FIG. 2 depicts localized restoration of a single channel within the network of FIG. 1 in accordance with an aspect of the present invention;

[0011] FIG. 3 depicts path restoration on a link-disjoint path within the network of FIG. 1 in accordance with another aspect of the invention;

[0012] FIG. 4 depicts path restoration on a non-disjoint path within the network of FIG. 1 in accordance with the invention; and

[0013] FIG. 5 is a block diagram of coding block for accomplishing automatic restoration in accordance with the invention.

[0014] FIG. 1 depicts an exemplary mesh network 10 of the type commonly employed for transporting telecommunications traffic. The network 10 comprises a plurality of nodes 12A, 12B, 12C, 12D, 12E, 12F and 12G, pairs of which are coupled linked by optical fiber links 14₁-14₁₀as indicated in Table I below.

[0015] Each of links 14₁-14₁₀comprises at least one, and preferably, a plurality of bi-directional working channels 16 and at least one bi-directional restoration channel 18. In practice, each of the channels 16 and 18 carries traffic at an OC 48 rate, although the channels 16 and 18 could have higher or lower rates without departing from the teachings of the invention. Each of the nodes 12A-12G typically comprises a Digital Cross-Connect System (DCS) that possesses the ability to cross connect traffic received on one link to one or more other links. Such DCSs are well known and are available from a variety of manufacturers, including Lucent Technologies, and Alcatel. The working channels 16 may include multiple channels multiplexed by Wavelength Division Multiplex (WDM) Systems to carry all bi-directional channels over a single pair of fibers. Since the WDM systems are transparent to the transport of signals and are not involved in the restoration, they are not shown.

[0016] In the exemplary embodiment illustrated in FIG. 1, nodes 12D and 12E serve as end-point nodes at which particular traffic enters and/or exits the network 10 at each node ports 19D and 19E, respectively. Thus, for example, traffic entering the network at node 12E at port 19E can pass via channel bc in link 14₃to node 12C for routing therefrom via channel de in link 14₁₀to node 12A and then pass via channel fg in link 14₉to node 12D to exit the network at port 19D. The set of channels in links 14₃, 14₁₀and 14₉coupling the end-point node ports 19E and 19D in nodes 12E and 12D, respectively, comprises a “path.” This working path is shown in FIG. 1 as Path 30₁. Another exemplary path comprises channel βx in link 14₁, channel δε in link 14₉, channel φγ in link 14₆coupling the end-point node ports 19B and 19G. This working path is shown as Path 30₂.

[0017] Alternatively, traffic received at node 12E at port 19E and destined for node 12D for exit at port 19D passes between such nodes on an alternate path comprising a channel in each of the links 14₆, 14₇and 14₈when connected on demand.

[0018] FIG. 2 illustrates the network 10 of FIG. 1 configured with a Restoration Path Computing System (RPCS) 20 in accordance with the invention for effecting the computation of restoration paths in the network. The RPCS 20 includes a computing platform 22 in the form of a computer or the like Each of the nodes 12A-12G monitors the links 14₁-14₁₀to detect a link in failure. A link in failure is a link having one or more failed channels 16 and 18. The RPCS 20 creates and maintains a Link database 24 and a working Path database 26. The Link database 24 contains information about each of the links 14₁-14₁₀, including: (1) the Link Identification (LID) that typically includes the identity of the nodes to which the link is coupled, (2) Working channel within the link requiring restoration upon failure (WCR), (3) Working channel within the link not requiring restoration upon failure (WCN), (4) Spare restoration channel (SPC) within the link for use in restoring a failed working channel and (5) Not used channel (NUC) within the link. A working channel assigned either as a WCR or WCN must be connected to another OC-48 channel dropping at the same node or to another working channel running towards another node. A channel assigned as a SPC is typically not connected to another channel via the node. Normally, an idle signal is inserted in a SPC. A NUC is usable in the same way as a WCN but may not be connected to any other node. Neither a WCN nor NUJC participates in restoration.

[0019] The Path database 26 includes records containing information about each path, that is, the collection of channels in links that couple ports in one end-point node to another, such as nodes 19E in 12E and 19D in 12D for Path 30₁in FIG. 2. In practice, each path record contains a Path Identification (PID) that can include an identifying port number on each node (not shown) coupled to a link end. In addition, each path record in the database 26 includes a priority for each path that can be computed if no service priority exists, or can be assigned in accordance with a service priority. The RPCS 20 typically updates the path database 26 following any change in the network 10,such as set up or removal of a path or unavailability of a restoration channel.

[0020] The RPCS 20 utilizes the information in the Link database 24 and the path database 26 to compute alternate restoration paths between end-point nodes for downloading to such nodes as described hereinafter. Optionally, the RPCS 20 can store information about such restoration paths in a restoration path database 28 so that when any update is made, the RPCS need only send the incremental update information to the nodes. For each working path and for each link that could fail in the working path 30₁between end-point nodes, the RPCS 20 computes a restoration path. Thus, for example, the RPCS 20 computes a restoration path between the end-point nodes 12E and 12D assuming a failure in the link 14₃. Similarly, the RPCS 20 also computes a restoration path between the end-point nodes 12E and 12D assuming a failure of the link 14₁₀and another restoration path assuming a failure in of the Link 14₉. These restoration paths may have common channels, and in some cases may be identical. As discussed hereinafter the RPCS 20 will download the pre-computed restoration path information to the end-point nodes whenever the RPCS updates its restoration path database 28 to facilitate restoration.

[0021] The RPCS 20 computes the restoration paths in the following manner. First, the RPCS 20 creates a temporary list of all available restoration channels 18 in the network 10. For each of the links 14_i(where i in an integer ≦10), the RPCS 20 establishes a working path (P_i) for each of the n (where n is an integer) working path channels in a link 14_ithat requires restoration according to the restoration priority. Suppose that the restoration priority decreases with increasing channel number. To that end, the RPCS 20 creates a set (R_i) of channels. The channel set R_iconsists of all available restoration channels in the entire network except the ones in link 14_iand all the channels in path P_iexcept the one in link 14_i.

[0022] The RPCS 20 can use a number of computation methods to compute a restoration path from the channels in the set R_i. For example, restoration path computation could be based on the shortest path. For example, the RPCS 20 would consider a channel (say channel 1) in the link 14_iand then determine the path (P_i) in which the selected channel belongs. The path terminating nodes for P_iare A_iand Z_I. The RPCS 20 then computes the shortest path (P_ir) between the terminating nodes A_iand Z_Ifrom the restoration channel set R_iA number of shortest path algorithms exist and any such algorithm can be used to compute the shortest path. For example, Dijkstra's shortest path algorithm for a sparsely connected or an all pair shortest path algorithm using dynamic programming for a densely connected network can be used to compute shortest paths efficiently. The shortest path criteria can be defined by mileage of the links on the path, number of nodes traversed by the path, a weighted sum of these two, etc. The path information (P_ir) is then sent to the corresponding terminating nodes A_iand Z_i. Note that restoration path (P_ir) may have some common channels with the path P_i.

[0023] If no path is found, the RPCS 20 computes a shortest path, assuming that there are sufficient channels in all links as needed. The RPCS 20 then compares the actual available channels in the links and informs an appropriate Operations Systems (OS) (not shown) that no restoration path is found and which links need to be augmented with the necessary number of channels so a restoration path can be found. This information can be used to add restoration channels in appropriate links.

[0024] Next, the RPCS 20 proceeds to the second step and then begins the restoration path computation for the next highest priority path P_jthat includes a second channel of the WCR type (channel 2) in link 14_i. When the restoration path for channel i in link 14_iis completed, the restoration channel set R_iis updated by: (a) removing the channels already used by path P_ir, (b) removing all the channels of Path i and (c) adding all working channels of Path j except the one in link 14_i. The RPCS 20 then computes the shortest restoration path P_jrfrom the updated path set R_iand stores that information in the Restoration Path database 28. The information about path P_jris downloaded under the command of the RPCS 20 to the terminating Nodes A_jand Z_jcorresponding to Path P_j.

[0025] The process of computing the restoration path for the second channel in the link 14_iis repeated for the each of the remaining channels of WCR type in the same link. Thereafter, the RPCS 20 considers the next link and completes the previously discussed procedure for each link. The RPCS 20 computes the restoration paths for each link and restoration paths are obtained for all links. When a new channel is established for service or for restoration, restoration paths are re-computed and downloaded to the appropriate nodes.

[0026] Since the procedure described above determines additional channels in specific links needed for restoration of each path, the procedure can also be used to compute restoration capacity required for restoration of all paths in the network. A specific approach to determine restoration capacity would be the following. First, assume that there is an infinite number of SPCs available for each link. Then, the RPCS 20 computes restoration paths for each link failure using either a shortest path algorithm or another suitable algorithm. The set of used SPCs then constitute the spare capacity needed for the network. This set, of course, may not be sufficient to restore traffic for a node failure. The set can be extended by including the restoration paths for node failures as well.

[0027] The network 10 accomplishes restoration of the affected traffic upon a link in failure in three phases. First, the network attempts “localized” restoration by seeking to route traffic on a restoration channel 18 within the link in failure comprising part of the existing path when such a channel is available in the failed link. If no restoration channel is available, then the network attempts to route the traffic on restoration path that excludes the link in failure in the second phase using “Path based Automatic Restoration Switching (ARS-P).” If restoration is unsuccessful for some channels after the second phase because of multiple failures or node failure then RPCS 20 takes control to restore remaining channels in the third phase

[0028] The manner in which localized restoration is effected may best be understood by reference to FIG. 2. Upon the failure of a working channel in link 14₉between the nodes 12A and 12D, the nodes 12A and 12D detect the failed channel because of a Loss of Signal (LOS), Loss of Frame (LOF) or Signal Degrade (SD). Upon detecting a failed channel, the nodes 12A and 12D communicate this event, typically via overhead bytes in the traffic signals as will be discussed below. The communication, however, can be accomplished via external communication channels established between nodes. The transmitted information includes the identity of the failed link, type of failure, the Path Identification (PID), and the target Node Identification (NID), that is, the identification of the end-point node.

[0029] Upon receipt at the node 12A of an indication of a failure in the link 14₉, the node 12A inserts a standard path alarm indication signal (AIS-P) on a still-working segment of the path for receipt at the end point node 12E at which the traffic terminates. Then, Node 12A checks for the availability of a restoration channel 18 on the link 14₉. Should an available restoration channel exist, then the Node 12A accomplishes automatic restoration on that existing channel and node 12D does likewise. In the event that no available capacity exists on the link 14₉, then nodes 12A inform node 12E which, in turn utilizes a pre-computed restoration path, as does node 12D, to restore service.

[0030] Table II below lists the localized restoration protocol in case of a failure of a channel in the link 14₉.

[0031] While the above-described localized restoration process is similar to the conventional 1 :n bi-directional SONET APS protocol, there are several major differences. First, the localized restoration process of the invention dynamically creates a protection boundary between the egress points of signals in an OC-48 system. Further, the localized restoration process of the invention requires no dedicated channels as are required in the SONET APS protocol. Additionally, the localized restoration scheme of the invention accomplishes selection at a node exit point, not at the receiving point closest the failure location. Thirdly, the present technique accommodates more than fourteen channels, the maximum limit of the SONET APS protocol.

[0032] As discussed, the failure detecting nodes in Network 10 attempt localized restoration first when a link is in failure. However, in some instances, the failure detecting nodes cannot effect localized restoration. Referring to FIG. 3, should the entire link 14₉become severed, then no restoration channel 18 is available. Under such circumstances, the node 12A sends a restoration code for Path 30₁towards node 12C for ultimate receipt at node 12E, which, in response assigns itself as the restoration switching node. Likewise, node 12D being the end point of the failed path 30₁assigns itself as the other restoration switching node. The nodes 12E and 12D then effect restoration by a previously computed and downloaded path from RPCS 20.

[0033] In the event of an inability to locally restore traffic on the link 14₉, the nodes 12E and 12D must switch traffic. Thus, for example, if the node 12E and 12D had previously routed a traffic via a path comprising the links 14₃, 14₁₀and 14₉, now node 12E and 12D must route traffic over a restoration path that, in the illustrated embodiment, comprises restoration channel pq in link 14₈, restoration channel rs in link 14₇and restoration channel tu in link 14₆. In particular, the node 12E bridges the traffic that previously passed on the link 14₃to the restoration channel pq in the link 14₈while the node 12F connects that traffic received in channel pq on the link 14₈onto the restoration channel rs in the link 14₇. The node 12G likewise connects the traffic now received from the restoration channel rs on the link 14₇onto the restoration channel tu on the link 14₆. Lastly, the node 12D switches it's output port 19D from the now-failed link 14₉to traffic-carrying restoration channel tu in the link 14₆. The Node 12D completes a like path in the reverse direction.

[0034] The restoration path depicted in FIG. 3 is link and path disjoint. In other words, the original traffic-bearing path comprised of the links 14₃, 14₁₀and 14₉share no common nodes or links with the restoration path comprised of the links 14₈, 14₇and 14₆, other than the end-point nodes 12E and 12D. If restoration were restricted to only to link and disjoint paths, the potential capacity savings achievable in the mesh network 10 would disappear. Advantageously, the restoration method of the invention is not restricted to link and disjoint paths. As seen in FIG. 4, the pre-computed restoration path downloaded from the RPCS 20 to the nodes 12E and 12D could comprise the working path channel bc on link 14₃, restoration channel vw on the link 14₅, and restoration channel tu on link 14₆. Note that link 14₃comprises part of the original path, thus obviating the need for any bridging at node 12E. However, node 12C must bridge the traffic received on the channel bc on link 14₃to the restoration channel vw of the link 14₅to effect restoration.

[0035] To effectuate automatic restoration switch in accordance with the method of the invention, the following connection rules apply

[0036] Rule-1. A node originating the Automatic Restoration Switching-Path (ARS-P) bridges the path to the restoration channel if that channel is different from the original working channel and sends the ARS-P code over the working channel if that is on the restoration path without any bridging action.

[0037] Rule-2. An intermediate node receiving ARS-P request from a working channel passes through the request when the outgoing restoration channel is also a part of the original working path (WCR to WCR).

[0038] Rule-3. An intermediate node receiving ARS-P request from a working channel bridges the incoming channel to the outgoing channel when the outgoing channel is not part of the original working path (WCR to SPC).

[0039] Rule-4. An intermediate node receiving the ARS-P request from a restoration channel connects the incoming channel to another outgoing restoration channel when none of the channels are part of the working path (SPC to SPC).

[0040] Using the above rules, restoration of two or more paths can occur simultaneously without conflict. In this way, the restoration technique of the invention can achieve fast restoration, comparable to speeds achieved in SONET/SDH rings.

[0041] Upon repair of a failed channel in a link, the nodes coupled to that link now receive valid signals because either the path terminating node or the bridging node did not remove the original connection. When a node receives a valid signal, it checks the Path ID to determine if the signal is valid. Upon receipt of a valid signal, the switching node inserts a Wait-to-revert (WTR) code in the channel of the restored path. After a prescribed WTR time expires after the code is received by the node at the other end of the path, the node at the other end of the path then reverts to the original path.

[0042] Implementation of the restoration scheme of the invention requires not only pre-computation of the restoration paths as discussed previously but also require a communications protocol that enables simultaneous communication of restoration path information to the nodes. A particular implementation of the communication method in the restoration method of the invention employs eighty-seven bytes for the Automatic Restoration Switching (ARS) protocol. In a typical OC-48 signal, there are forty-seven bytes next to the K1 byte and another forty-seven bytes next to the K2 byte, thus providing sufficient capacity for the eight-seven bytes needed for restoration. The specific coding of the eight-seven bytes is depicted in FIG. 5 and listed in Table III below.

[0043] 1 An alternative LC ID coding using 2 bytes: First 8 bits for NID at one end of the link and the next 8 bits for the NID for the node at the other end of the link.

[0044] The following steps demonstrate the execution of the above-protocol using the ARS bytes and the failure of a working channel in the link 14₉in both directions due to Loss Of Signal (LOS) as depicted in FIG. 2.

[0045] 1. Both Nodes 12A and 12D detect LOS in a working channel fg of Path 30₁, in the link 14₉, Node 12A decides after examining the priority of other failed channels (if any) that it can use a restoration channel 18 in the same link to restore the failed working path 30₁. Node 12A inserts a Path Alarm Indication Signal (AIS-P) in the outgoing direction of channel ed of Path 30₁, on link 14₁₀and the node 12D inserts such a signal at its output port 19D in the outgoing direction. (The AIS-P signal comprises of “all 1s” in the entire payload including pointer bytes.) Other SOH (Section Overhead) bytes are normal i.e., the ARS bytes are coded for no request. Nodes 12A and 12D become the switching nodes and initiate Automatic Restoration Switching-Local (ARS-L) using the available restoration channel xy in the link 14₉.

[0046] 2. Node 12A sends bridge request code to the node 12D. Bridge request code in the switching state of the node is the failure indication coded in bits 5-8 of ARS Byte 1. The ARS-L code carried in the restoration channel xy in link 14₉includes a value 0-1 in bits 3 and 4 of ARS Byte 1 in channel xy, 1100 (for LOS Local) in bits 5-8 of Byte 1, link and channel Identifications (LC) for the failed channel gf in the link 14₉in the first 12 and last 8 bits of LC # 1 of the ARS bytes, respectively. LC #1 is used to carry the switch requesting link and channel identification in ARS-L. LC #2 is used to indicate the bridged link and channel IDs of the failed channel in the link 14₉. The remaining ARS bytes are set to NULL. The same action is taken by node 12D in the opposite direction on the restoration channel yx of the link 14₉.

[0047] 3. Upon receiving the ARS-L code described in Step 2 on the restoration channel in the link 14₉, node 12D, being the switching node, removes connection between the failed channel in the link 14₉to its output port 19D and makes connection between its output port to the restoration channel xy in the link 14₉. Similarly, node 12A switches from the working channel in the link 14₁₀to the restoration channel in the link 14₉to complete the bi-directional switch. Note that if the failure is only in one direction, say from 12D to 12A, then 12D receives the same ARS-L code as in Step 2. However, Node 12D, not detecting any failure, does not yet become a switching node. Node 12D, upon receipt of the ARS-L code from Node 12A, bridges 19D to the restoration channel yx and sends Reverse Request in bits 1-4 of ARS byte 1 to Node 12A. Node 12A, upon receipt of the Reverse Request from Node 12D, executes a bridge of the channel de in link 14₁₀to the channel xy in link 14₉and a switch from the gf-ed channel connection to the yx-ed channel connection. Node 12A then sends bridged LC ID in LC #2 to Node 12D. Node 12D, upon receipt of the bridged channel code, executes a switch from the fg-19D connection to the xy-19D connection. Thus, a Bi-directional switch is completed by using Reverse Request code in case of a unidirectional failure.

[0048] 4. Suppose that failed link 14₉does not have any SPC on the same link either because other higher priority channels captured available SPCs or no SPC available due to the failure of the entire link. Nodes 12A and 12D begin the ARS-P phase of the PBR protocol of the invention. Node 12A inserts an 10 code (Switch Control Request ARS-C) in Restoration Mode bits 3 and 4 of the ARS Byte 1 to indicate to node 12E that it needs to take control to switch the failed path. Request type bits (5-8 of ARS byte 1) indicate the failure type such as LOS with 1011 (Signal Fail-Link). LC # 1 (20 bits) indicate the link number of the failed channel (12 bits) and the channel number (8 bits) in the link. Remaining 31 LCs and other bytes are set to NULL.

[0049] 5. Node 12C receives the ARS-C code and determines that it is not the path terminating point for the failed path 30₁, and passes the ARS-C code to Node 12E.

[0050] 6. Node 12E receives the ARS-C code from the working channel cb of the failed path 30₁, on the link 14₃and determines that the node must take control for switching the failed path 30₁. Looking at LC #1, node 12E determines which link failed. Then node 12E examines the restoration path for the path 30₁, corresponding to the link 14₉in failure. Suppose that the restoration path is 30₃as shown in FIG. 3 comprising restoration channels pq, rs, and ut in the links 14₈, 14₇and 14₆, respectively. The node 12E now becomes a switching node. The node 12E inserts “11 ” in restoration Mode bits of ARS Byte 1 of the restoration channel pq in link 14₈to indicate path switch mode to the other end Node 12D. The request Type bits are set to the received value from working channel in the link 14₃to indicate the request type. LC #1 is coded with the Link and Channel ID for the restoration channel in the link 14₈, LC # 2 for the restoration channel in the link 14₇and LC #3 for the restoration channel in the link 14₆to indicate to the nodes on the restoration path to make appropriate connections. Node 12E inserts the Node ID for node 12D in the Target Node ID byte, its own ID in the originating Node ID byte, the failed Path ID in the requesting Path ID and the bridged path ID in Bridging Path ID locations of ARS bytes. Note that Switch requesting Path ID and bridging Path ID bytes are identical in this case.

[0051] 7. Node 12F receives the code from the restoration channel pq in the link 14₈and the node determines from the Restoration Mode that it needs to look at the first two channels and connect them from LC #1 to LC #2. In this case, it connects the restoration channel pq in the link 14₈to the restoration channel rs in the link 14₇. Then, the node 12F shifts all the LC codes one unit to the left. In other words, LC #1 code is removed, LC #2 is written in LC #1, LC #3 code is written in LC #2 and LC # 3 is set to NULL.

[0052] 8. The node 12G receives the ARS-P switch request code on the restoration channel rs in the link 14₇and the node repeats the step in 7.

[0053] 9. The node 12D receives the switch request code from the restoration channel tu in the link 14₆and now examines at LC #1 and 2 as in Step 7,and determines that the restoration path 30₃ends at the node. The Node 12D can further verify that it is the end node by looking at the Target Node ID. The node 12D after verifying that the working path 30₁is the requesting path to be restored path 30₃from the other end 12E executes a bridge and switch function. In the switch function, the node 12D removes the connection to its output 19D to the working channel in the failed link 14₉and makes the connection to the restoration channel tu in the link 14₆, thus completing the switch in the direction of node 12E to 12D. In the bridge function, the node 12D connects port 19D to channel ut in link 14₆without removing 19D-gf connection. If the failure is in both directions on the working channel of the link 14₉, then node 12E would similarly switch the connection between its output port 19E from the working channel in the link 14₃to the restoration channel qp in the link 14₈in response to a request by node 12D. On the other hand, if the failure was only in the direction from node 12D to node 12A, the node 12E would insert bridge request code in that restoration channel as described in Step 3. The node 12D, upon receipt of the bridge request from Node 12E, becomes a switching node. Node 12D then executes a bridge of 19D to restoration channel ut and inserts Reverse Request code in the restoration channel ut and bridged signal indication in LC #2. Node 12E, upon receipt of the Reverse Request from node 12D, executes a bridge of 19E to the restoration channel pq and a switch from cb-19D connection to qp-19D connection. This completes the bi-directional switch whether the failure is unidirectional or bi-directional.

[0054] 10. Steps 1 through 9 are followed to restore other paths failed at the same time without any conflict because there is no common channel between the restoration paths as computed beforehand.

[0055] 11. When failed working channel in the link 14₉is repaired, Node 12A and Node 12C receive valid signal with No Request (NR) code in ARS. Node 12C passes through the NR code to Node 12E. After receiving NR from WCR cb in link 14₃, the node 12E then inserts Wait-to-restore (WTR) in bits 5-8 of the ARS Byte 1 which is passed through by Nodes 12F and 12G to Node 12D. Node 12D, upon receipt of the WTR waits for a predetermined WIR period and then reverts to the original connection, i.e., releases the tu-19D connection and makes fg-19D connection. Similarly Node 12E, upon receipt of WTR and waiting for a predetermined WTR period reverts to the original path. If the original failure was unidirectional, say 12D-12E direction, then the node 12E sends WTR to node 12D when the failure clears. Node 12D, upon receipt of the WTR code from node 12E, inserts WTR in the reverse direction, waits for a pre-determined WTR period and then executes a switch back to the original path 30,by disconnecting tu-19D and connecting fg-19D. Node 12E, upon receipt of WTR from node 12D, waits for a pre-determined WTR period and then executes a switch back to the original path 30₁by disconnecting qp-19E and connecting cb-19D. That completes the bi-directional switch after a unidirectional failure is repaired. Nodes 12D and 12E may check PID before switching for further verification. Upon execution of the switch back function, nodes 12D and 12E insert No Request code in the channels of Path 30₃. The node 12F, upon receiving the NR code from node 12E releases the connections of the restoration channels from the links 14₇to 14₈and pass through the NR code to the node 12G. Node 12G, likewise, releases the connection from restoration channel in 14₆to the restoration channel in 14₇and then passes through the NR code to Node 12D. Node 12D, upon receipt of NR code from Node 12E, releases the bridge 19D-ut. Similar action is taken in the other direction. All channels then return to the normal state.

[0056] 12. After the steps 1-10 are executed, some or all of the channels may not be restored if the failure in the network 10 was due to a node failure or there was more than one link failure. The paths that are not restored or a predetermined time expires before the restoration can be completed, the path end-point nodes communicate to the RPCS 20 that one or more paths could not be restored. The RPCS 20 then initiates the third phase of the restoration. All nodes, when they detect a failure, send the failure information to the RPCS 20. From this information, the RPCS 20 determines that there is a node failure or there is more than one link failure. The RPCS 20 computes yet another set of restoration paths based on the unavailability of restoration channels in multiple links or all links through the failed node. The RPCS 20 then downloads the new restoration path information to the end-point nodes of the failed paths and commands these nodes to initiate restoration. The steps 1 through 10 are re-executed with these new restoration paths to enable the end-point nodes to restore the remaining paths that were not restored in the first two phases. Return to normal state is accomplished in the same manner as described in Step 11.

[0057] The above-described restoration method of the invention affords rapid restoration. By using SONET-like layer protocols, the nodes can detect failures in less that 10 milliseconds, the same as in a SONET/SDH ring network. For localized restoration, the total time required for the nodes to determine the available restoration channels and communicate that information amongst themselves is of the order of 20-30 milliseconds and the complete switching time including that of detection is less than 60 milliseconds. Alternate Restoration Switching-Path (ARS-P), in accordance with the second phase of the restoration method of the invention takes somewhat longer, due to the need to first determine the inability to achieve localized restoration. Also, a small amount of time will be consumed to determine the appropriate restoration paths at the end-point nodes, to communicate between the end-point nodes through intermediate nodes on the restoration path using the ARS-P protocol and to execute the switching at the intermediate nodes. Even for the ARS-P, a restoration of 100-300 milliseconds is possible.

[0058] The foregoing describes a technique for achieving rapid path-based restoration in a mesh network.

[0059] It is to be understood that the above-described embodiments are merely illustrative of the principles of the invention. Various modifications and changes may be made thereto by those skilled in the art, which will embody the principles of the invention and fall within the spirit and scope thereof.