Shortest path first restoration routing in a fiberoptic network

Abstract

Interconnection failures in a communication network require the interconnections between the nodes of the network to be restored. SPF allows for the determination of the paths between the nodes in some optimal way. An improved implementation of SPF avoids or reduces the enormous amount of calculations required by SPF at each restoration operation.

Description

CROSS-REFERENCES TO RELATED APPLICATIONS

[0001] This patent application claims priority from Provisional Patent Application Nos. 60/215,399 and 60/215,182, both filed Jun. 29, 2000, and are hereby incorporated by reference.

BACKGROUND OF THE INVENTION

[0002] The present patent application is related to fiberoptic networks, and, in particular, to switches for WDM and DWDM network systems.

[0003] In WDM (Wavelength Division Multiplexing) fiberoptic networks, optical signals are sent at predetermined wavelengths over optical fibers. Each predetermined wavelength forms a communication channel in the network and the wavelength (or frequency) of the optical signal is used to control the destination of the signal through the network. An advanced version of WDM networks is the DWDM (Dense Wavelength Division Multiplexing) network in which the number of wavelength channel is increased by reducing the channel wavelength separation to 100 GHz, as set by the ITU (International Telecommunications Union). Hence the term, DWDM, is used herein to refer to both WDM and DWDM networks and other fiberoptic networks which rely upon wavelength to define communication channels, unless indicated otherwise.

[0004] In networks, including such fiberoptic networks described above, switches or routers are used to select paths for signals through the networks. In fiberoptic networks switches and routers not only direct optical signals from one optical fiber to another but also from one wavelength channel to another. The availability of light paths is critical to the users of a network. One way to provide reliability for a light path within the network is to explicitly provide for a redundant path beforehand. However, this approach does not utilize the bandwidth of the network efficiently, i.e., some of the available network capacity is removed for the backup reserve.

[0005] The present invention, on the other hand, is directed toward on-the-fly light path restoration to achieve efficient bandwidth usage and availability at the same time. New paths are quickly rerouted through the network in place of the lost light paths. For the rerouting, the present invention provides for an efficient implementation of so-called “Shortest Path First” (SPF) algorithm developed by F. W. Djikstra.

SUMMARY OF THE INVENTION

[0006] The present invention provides for a method of operation in a communication network having a plurality of nodes interconnected by links between pairs of nodes, particularly a fiberoptic network. The method optimally selects paths between the nodes of the network. Each source node and a destination node forms a path of concatenated links selected by an extreme value of a predetermined metric for the path between the source node and the destination node after one or more failed links of the paths. The method comprises maintaining for each node information of selected paths from the node to remaining nodes of the communication network in an order of the predetermined metric value; eliminating the information of each selected path having a failed link in the path; selecting a path between the node and each one of the remaining nodes from the information of the determined paths provided that the information of the path has not been eliminated; and selecting a path between the node and each one of the remaining nodes by concatenating links having an extreme value of the predetermined metric for the path if the information of the path has been eliminated.

[0007] The information includes a first database set of network paths between nodes of the network, the paths ordered by the predetermined metric values; a second database set of evaluated network paths between nodes of the network, the paths also ordered by the predetermined metric values; a database set of failed links; a database set of destination nodes evaluated for a path from the node to each remaining node in the network by the value of the predetermined metric, the database set including pointers from each destination node to the second database of evaluated network paths so as to provide a selected path from the node to the destination node if the predetermined metric has an extreme value.

BRIEF DESCRIPTION OF THE DRAWINGS

[0008] FIG. 1 is an exemplary DWDM network of a plurality of switch nodes operating according to the present invention;

[0009] FIG. 2 is a flow chart of steps of conventional SPF operations for selecting paths in a network; and

[0010] FIGS. 3A-3C illustrate a flow chart of network restoration steps according to the present invention.

DESCRIPTION OF THE SPECIFIC EMBODIMENTS

[0011] FIG. 1 illustrates an exemplary fiberoptic network with a plurality of switch nodes. Only five switch nodes 10-14 are shown for the purposes of ease of explanation; more or less switch nodes could be used. Each of the switch nodes 10-14 is connected to external data fiberoptic lines 20, 22, 24, 26 and 28 respectively. For example, the switch node 10 is connected to a plurality of fiberoptic lines connected to sources and receivers (not shown) external to the fiberoptic network, which lines are represented by the line 20. Likewise, the switch node 11 is connected to a plurality of externally-connected fiberoptic lines represented by the line 22, and so on. Within the fiberoptic network, the switch nodes 10-14 are interconnected by fiberoptic lines 21, 23, 25, 27, 29, 30 and 31. The fiberoptic lines carry optical signals to, from, and between the nodes 10-14 at distinct and preselected wavelengths which form network communication channels.

[0012] Details of the architecture and operation of the switch nodes 10-14 are described in U.S. application Ser. No. 09/648,518, entitled, “Scalable DWDM Network Switch Architecture With Wavelength Tunable Sources,” filed Aug. 25, 2000 by Chien-Yu Kuo, Niraj Gupta and Ronald Garrison, which patent application is assigned to the present assignee and which is hereby incorporated by reference. The memory and computational capabilities at each network for the distributed provisioning and restoration operations of the network are disclosed in U.S. application Ser. No. ______, entitled “Method For Wavelength Switch Network Restoration,” which patent application is also assigned to the present assignee and is hereby incorporated by reference. Provisioning refers to the operations by which the nodes of a network receive information to begin operations and restoration refer to operations. Both operations entail control and signaling operations by which the network nodes are each provided information about the network environment, e.g., the status of the other nodes and the paths along which signals can pass from one node to another. The present invention is concerned with the determination of these signal paths.

[0013] When a failure occurs in the connection between the nodes of the network, the network connections must be restored. Various algorithms have been proposed to make the “best” connections between the nodes of a network. The present invention uses an efficient implementation of the “Shortest Path First,” (sometimes termed SPF) developed by F. W. Djikstra. Descriptions of this algorithm can be found in many sources, such as the textbook, Routing in the Internet, Second Edition, by Christian Huitema, Prentice Hall, Upper Saddle River, N.J., p. 125. In network terminology, the one-segment connection between two nodes is a communication link. The connection between two nodes, which may be a link or a concatenation of links, is termed a path. It should be noted that in the exemplary fiberoptic network illustrated in FIG. 1, the physical link illustrated by a fiberoptic line between two nodes, say line 21, is representative of many communication channel links. In current fiberoptic networks, there are typically 64 or 65 wavelength channels on each fiberoptic line, most of which are used to carry data between the nodes and some of which are used as substitute links when a regular communication link is faulty and the communication link is broken. Hence the description below refers to communication links, unless explicitly stated otherwise.

[0014] FIG. 2 shows a flow chart of steps required by SPF to determine the optimum connections after a network failure for one node S in the network. After start step 40, database sets E and R are initialized so that set E contains only the node S and set R contains all of the remaining nodes of the network. Then step 42 initializes a list O, an ordered database of links, i.e., the one-segment paths, from the node S. In step 42 only the links from node S are placed in list O. By step 43, the list O (at this point containing only links from node S) is ordered by a predetermined metric. The particular order of the database is set by the “cost” of the path. Each link and, therefore, each concatenated path through the network, has an associated metric, or “cost.” For example, a metric might be the physical distance of the segments, the transmission time, or the actual expense of transmitting messages through the links. The list O provides that the links and paths are ordered so that the “shortest” links and paths come first for the most efficiency.

[0015] Decision step 44 asks if the list O is empty. If so, i.e., there are no more paths from node S, the remaining nodes in set R is marked as unreachable from node S by step 46 and the process is terminated at step 47. Even if the list O is not empty, similar decision step 45 checks whether the first path in list O have an infinite metric, i.e., whether the “shortest” path in list O is open, indicating that there remains no real path from node S. If the metric is infinite, step 46 marks the remaining nodes in set R as unreachable and the process is terminated.

[0016] If the metric is not infinite, then step 48 removes the shortest path P from list O. Step 49 checks whether the last node, termed V, in the removed path P is already in set E. In other words, has a path from node S to node V already been found? If yes, the process is returned back to step 44. If not, the node V is moved from set R (of remaining nodes) to set E (of evaluated nodes) by step 50. In step 51, new paths from node S to nodes connected to node V by links are determined by concatenating path P to each of the links connected to node V. The new paths are placed into, and ordered in, list O by step 52. The cost of the new paths is the sum of the cost of path P and the metric of each link from node V to the nodes connected to node V.

[0017] The process continues until all the paths from node S are determined. In the case of the exemplary fiberoptic network of FIG. 1, the process continues for all the wavelength channels of the network. This SPF process is invoked every time a new failure, such as a link going down, in the network occurs and new paths must be calculated for each node in the network. It should be noted that with any network of consequence, the calculations required to determine the optimum paths by SPF are enormous. For example, with a network of 100 nodes and each optical fiber link carrying 64 wavelength channels, the break in one fiber link requires the calculations described above, a very large number of calculations indeed.

[0018] In contrast, the present invention reduces the calculations required by SPF to determine the optimal paths in a network restored after a failure. Hence the speed of restoration of the connections after a failure is greatly enhanced.

[0019] For each node of the network, the present invention stores information about the links and connecting paths of the network channel from that node for each wavelength. Some of the databases are the same as described above, others are expanded with more information, while still others contain different. The databases include the list O of ordered paths, a list Q of ordered paths which have been already considered; the database set E of evaluated nodes and a database set U of links unavailable since the last calculation of paths for interconnecting the network nodes. The database set U of unavailable links permits some flexibility in the path restoration operations, which must be performed immediately if the database set U did not exist. Each of the nodes of the database set E contains information includes identification of the destination node from the source node along a path, a list L of pointers to the paths in list Q with each pointer associated with a destination node. Paths are stored with information of their source node, their destination node and the set PL of ordered links from the source node to the destination node.

[0020] Upon invocation of the process, the determination of the connecting paths are carried as follows according to the present invention. FIGS. 3A-3C illustrate a flow chart of steps required according to the present invention. After start step 60, decision step 61 asks if the databases sets E and U, and lists O and Q, as described above exist, are nonempty, or in other words, if a routing procedure has been performed previously. If no, then the process moves to step 71 on FIG. 3B. If yes, then the process eliminates paths and pointers to the paths with links which are now unavailable. Step 62 deletes paths which have links now unavailable (links in database set U) from list Q and corresponding pointers in list L to these paths in list Q. Step 63 likewise deletes paths which have unavailable links from list O and step 64 then reinitializes the database set U to empty. For any remaining destination nodes in database set E, step 65 follows the corresponding path in list Q as indicated by the corresponding pointer in list L, i.e., a connection has been found from the source node to the destination node.

[0021] The process then continues to the operations described in FIGS. 3B and 3C, which have many steps similar to the those described previously with respect to the conventional SPF operations. By step 71 database sets E and R, augmented as described above, are initialized so that set E contains only the source node S and set R contains all of the remaining nodes of the network. Then step 72 initializes the list 0, the ordered database of links, i.e., one-segment paths, from the node S. By step 73, the list 0 (at this point containing only links from node S) is ordered by the predetermined metric. As described previously, the list 0 provides that the links and paths are ordered so that the “shortest” links and paths come first.

[0022] Decision step 74 asks if the list O is empty. If so, i.e., there are no more paths from node S, the remaining nodes in set R is marked as unreachable from node S by step 76 and the process is terminated at step 77. Even if the list O is not empty, similar decision step 75 checks whether the first path in list O have an infinite metric, i.e., whether the “shortest” path in list O is open, indicating that there remains no real path from node S. If the metric is infinite, step 76 marks the remaining nodes in set R as unreachable and the process is terminated.

[0023] If the metric is not infinite, decision step 83 determines whether the last link in the path under consideration is a substitute link which exhausts the spare capacity of the link. The optical fiber between the two nodes has no more spare channels available to substitute for nonfunctioning channels. If not, the process continues to step 78 to build up the network paths from the source node S. If so, step 84 places the exhausted link in list U of unavailable links and the information is broadcast to the other network nodes. Then the process returns to step 78 and the process of finding the network paths from source node S continues. The shortest path P is removed from list O and step 79 checks whether the last node, termed V, in the removed path P is already in set E. In other words, has a path from node S to node V already been found? If yes, the process returns back to step 74. If not, the node V is moved from set R (of remaining nodes) to set E (of evaluated nodes) by step 80. In step 81, new paths from node S to nodes connected to node V by links are determined by concatenating path P to each of the links connected to node V. The new paths are placed into, and ordered in, list O by step 82. The cost of the new paths is the sum of the cost of path P and the metric of each link from node V to the nodes connected to node V.

[0024] The process loops back to step 73 and continues until all the paths from node S are determined.

[0025] The present invention reduces the amount of network restoration calculations tremendously for networks of any size. If M equals the number of physical links, i.e., optical fibers between nodes, and C is the number of circuits of a failed link, i.e., the number of connection paths using the failed link, in a network, conventional SPF calculation require an order of calculations of C*M*(log M). On the other hand, network restoration calculations according to the present invention as described above are an order of(N−1)*M*(log M) calculations with N being the number of nodes in the network. As the complexity of the network increases, it is evident that the restoration calculations in accordance with the present invention become increasingly more efficient.

[0026] Further improvements to the network restoration are possible. For example, in the determination of the network paths, there is an implicit assumption that the communication paths are unidirectional. Communication is from source node S to a destination node D. In the fiberoptic network described, communication is bidirectional so that a communication path once determined from node S to node D can be reversed so that a communication path from node D to node S is also determined. This creates another savings in network restoration calculations.

[0027] Therefore, while the description above provides a full and complete disclosure of the preferred embodiments of the present invention, various modifications, alternate constructions, and equivalents will be obvious to those with skill in the art. For example, while the present invention has been described in terms of fiberoptic networks, it should be applicable to other types of networks. Thus, the scope of the present invention is limited solely by the metes and bounds of the appended claims.

Claims

1. A method of selecting paths between nodes of a communication network, each path between a source node and a destination node formed by concatenated links selected by an extreme value of a predetermined metric for said path between said source node and said destination node after one or more failed links of said paths, said method comprising

maintaining for each node information of selected paths from said node to remaining nodes of said communication network in an order of said value of said predetermined metric;

eliminating said information of each selected path having a failed link in said path;

selecting a path between said node and each one of said remaining nodes from said information of said determined paths provided that said information of said path has not been eliminated; and

selecting a path between said node and each one of said remaining nodes by concatenating links having an extreme value of said predetermined metric for said path if said information of said path has been eliminated.

2. The method of claim 1 wherein said information comprises

a first database set of network paths between nodes of said network, said paths ordered by values of said predetermined metric;

a second database set of evaluated network paths between nodes of said network, said paths ordered by values of said predetermined metric;

a database set of failed links;

a database set of destination nodes evaluated for a path from said node to each remaining node in said network by said value of said predetermined metric, said database set including pointers from each destination node to said second database of evaluated network paths so as to provide a selected path from said node to said destination node if said predetermined metric has an extreme value.

3. The method of claim 2 further comprising setting said database set of failed links to empty after said eliminating step.

4. In a communication network having a plurality of nodes interconnected by links between pairs of said nodes, a method of selecting paths between each node and a destination node formed by concatenated links selected by an extreme value of a predetermined metric for said path between said source node and said destination node after one or more failed links of said paths, said method comprising

maintaining for each node information of selected paths from said node to remaining nodes of said communication network in an order of said value of said predetermined metric;

eliminating said information of each selected path having a failed link in said path;

selecting a path between said node and each one of said remaining nodes from said information of said determined paths provided that said information of said path has not been eliminated; and

selecting a path between said node and each one of said remaining nodes by concatenating links having an extreme value of said predetermined metric for said path if said information of said path has been eliminated.

5. The method of claim 4 wherein said communication network comprises a fiberoptic network, each link comprising a wavelength channel between a pair of nodes.

6. The method of claim 5 wherein said information comprises

a first database set of network paths between nodes of said network, said paths ordered by values of said predetermined metric;

a second database set of evaluated network paths between nodes of said network, said paths ordered by values of said predetermined metric;

a database set of failed links;

a database set of destination nodes evaluated for a path from said node to each remaining node in said network by said value of said predetermined metric, said database set including pointers from each destination node to said second database of evaluated network paths so as to provide a selected path from said node to said destination node if said predetermined metric has an extreme value.

7. The method of claim 2 further comprising setting said database set of failed links to empty after said eliminating step.