SE512055C2 - Detection of loop - Google Patents

Abstract

The invention relates to a loop-detection means produced by a source switch AS when establishing a route to a destination switch ES wherein the loop-detection means is a loop back cell (20) having the source switch AS in both source (21) and turning point (22) fields.

Description

15 (fl - .. x FO (D (TI (fl 2) This identifier changes when needed by the exchange when the cell passes through the node and is transferred to the next link. The payload contains information from the flow or message transmitted on the connection.

For connections that are connected due to initiation signals from the end systems, loops cannot normally occur, as this is prevented by functions in the network that handle the signaling.

To ensure that the road is operating correctly, special maintenance packages, called feedback cells, can be sent out on the road. A feedback cell contains information with which the source and destination of the feedback cell can be identified.

There is also an indicator, called a feedback indicator, which shows whether the cell is traveling to or from the feedback point / destination. If node A wants to check that the path to node B works correctly, it therefore transmits a feedback cell with A as source and B as destination. The cell contains information which means that the cell does not belong to any user and in particular that it is a maintenance cell. The maintenance cell has a field that indicates that it is a feedback cell.

This field informs the destination switch or node, in this case node B, that it must invert the cell and send it back to the source. Therefore, when a node receives a feedback cell, it examines the feedback point / destination field to see if it should send the cell back. If it sends the cell back, it changes the feedback indicator to show that the cell has been sent back. It does not change source identity or feedback identity. There may also be another field in the feedback cell, called the correlation tag, which remains unchanged. If the path between A and B works correctly, node A will receive the feedback cell back from B with the feedback indicator showing that the cell has been sent back. This event occurs some time after the feedback cell was loaded. This time depends on the length of the road and any queues or other delays on it.

The connection connection procedure takes some time. on the order of a few tens to a few hundred milliseconds per node passed. and therefore introduces a certain latency for the end systems. However, once the connection is established, however, the information can be transmitted through the nodes / switches in an enactive manner by the switching hardware. new connection must be connected, the new connection can then be routed so that it avoids the part of the network that has problems.

In large unconnected networks (by unconnected is meant that no fixed communication is connected between communicating units, but that the information is sent in the best available way at all times), e.g. Internet, where information is sent as discrete information packets between nodes, e.g. A, B, the packages comprise a starting field, i.e. a part that contains information about, i.a. its identification, where it came from (the source), where it is going to (destination) the length and other useful information. Each node uses a routing protocol to exchange routing information on how other nodes in the network can best be reached. A description of a routing protocol is described in I-IUITEMA, CHRISTIAN, “Routing in the Internet”, ISBN 0-13- 132192-7, Chap. 4.4 to 4.2.6, the contents of which are incorporated into this application by reference. When these information packets are routed between nodes in a network, there is always a possibility that a loop can be formed. This can happen when a link or a node in a path is shut down or is temporarily or permanently unable to receive a new information packet from a previous node. In this case, the previous need forwards information packets via an alternative path that it has stored in its routing information protocol. However, it is possible that this alternative path at some stage contains a link that has routing information protocols that lead back to this previous node. In this case, the packets are routed from the previous node via the alternative path back to the previous node. In this way a loop is formed. If a loop is formed, system resources are wasted on unnecessarily sending packets around the loop. Information packets captured in a loop can be seen as useless, as it is unlikely that they can be delivered to their destinations on time even if the loop is broken. 10 15 25 b) CD 512 055 4 However, loops are usually transient problems, because the routing information protocols are updated with information about the status of the network so that the primary and alternative paths for each node are therefore also continuously updated. This update can be performed, for example, by exchanging distance vectors or information about the status of links between nodes.

Distance vectors contain information for each destination in the network, usually one element per destination, about the time, distance or “cost” of sending a packet from the specified node to that destination. Each node, according to a routing protocol, exchanges distance vectors with all its neighboring nodes. These exchanges take place periodically, and, possibly, also when events occur that lead to changes in costs. The distance vectors received from neighbors and knowledge of the cost of the links to the neighbors are used by nodes to recalculate their distance vectors.

When a packet is received to be forwarded to a particular destination, the corresponding distance vector determines which link to use to forward the packet at the lowest cost. The path with the lowest cost is then used to direct the package. If a node or a link to a node stops working in a certain path, its neighboring nodes register that the cost of using this node or link is infinite and that an alternative path must therefore be chosen. The information about the failing node is sent to other neighboring nodes in an updated distance vector. These nodes then update their own spacer vectors and send them to other grammar modes, and so on. In this way, the information about the failing link is spread longitudinally through the network. Each updated distance vector affects all the neighboring nodes and a number of exchanges of distance vectors are required between neighboring nodes before the distance vectors converge and a new stable state is reached. During this convergence period, it is possible that packets sent on the original path are affected by changes in the distance vectors and ends in a loop. This loop terminates naturally when the distance vectors have converged. During the convergence period, which can be up to one minute, however, scarce processor and bandwidth resources are wasted. 10 15 (TI 'IQ' YEAR 5 ih Ûvv Another reason for the transient nature of loops is that in most non-connection networks there are protocols according to which each information packet has a counter, usually in the start field, called "lifetime" ("Time to Live"). ") Which ensures that the packet is discarded, ie destroyed and removed from the network, after a certain time has elapsed or, more commonly, after the packet has passed through a predetermined number of nodes. Depending on the value of the" lifetime "counter when loops occur and on the delays in the loop, a loop can exist for a period of about one minute, during which time scarce routing processor resources are wasted handling the packets in the loop which, as mentioned above, are considered useless and should be discarded. lifetime “rälmare fi nns in HUITEMA, CHRISTIAN,” Routing in the Intemet ”, ISBN 0-13-132192-7, Chapter 3.3.1,“ The Internet Header ”which is included in this document by reference.

Unconnected networks are flexible and messages can be sent quickly as there is no need to establish a reserved connection between the source and destination before sending a message. However, these networks require considerable processing capacity to determine the destination of each packet being transmitted, and loops can run into systems, wasting resources.

To improve the performance of the networks, it has been proposed to form so-called "label switching networks", also known as multi-protocol label switching networks, in which a connectionless network is laid on top of a connection-oriented network. . In other words, a disconnected device such as a path selector is associated with a connection-oriented device such as a switch, for example in an ATM network. Each route selector uses its routing protocol to build its routing table with destination routing that indicates the best route to each specific destination, ie. to which node, called the downstream node, it must forward traffic to reach the desired destination.

It will then request a logical channel number, commonly called a label. from this downstream neighbor, for the traffic / package to each destination. When the node receives a corresponding request from its upstream neighbor, it can order its switching component to interconnect these channels. Since this process is performed in the nodes of the network, link storage connections are established from each source to each destination. These connections will follow the same paths as the best paths determined by the routing protocols and distance vectors.

A problem with such label switching networks is that route selection measurements can cause a loop which is then laid on top of the connection-oriented network. Since there is no special mechanism in a connection-oriented network for detecting loops, it is possible for such loops to exist for a relatively long time before the switching network detects, in the usual way, that there is a loop.

Summary A problem with unconnected networks, e.g. route selection measures that are placed on top of connection-oriented, e.g. exchange and / or ATM networks, is that the presence of a loop causes a waste of system resources. Routing protocols according to known technology developed for route selection measurements provide no guarantees against loops. They only provide conditions under which the protocols ensure that the maximum life of a loop is limited.

This means that if a loop occurs, system resources are wasted until the loop reaches its maximum life or the system is subjected to any change that happens to break the loop.

The present invention seeks to reduce the waste of system resources when a loop occurs by providing active means for detecting loops and for breaking loops in a connection-inoculated network, in particular an ATM network, and more specifically a network comprising a connectionless network laid on top. a connection-oriented network. This is achieved by a method according to which a source exchange, when establishing a connection to a destination exchange or node in a connection-oriented network, inserts a loop detection means, e.g. a loop detection cell, in the path. This loop detection cell contains identification means that can be recognized by the source node if the loop detection cell returns to the node. If there is no loop, this cell will continue to the destination point of the road where it will be removed. If there is a loop, the loop detecting cell will return to the source node. It will be recognized by the source node and this will make us aware of the source node that there is a loop, and measures can thus be taken to avoid the loop, e.g. the source node can inform an associated network control unit, e.g. a path selector that there is a loop. This has the advantage that a loop can be detected within a few milliseconds instead of about one minute.

In a preferred embodiment of the invention, the loop detection means is a feedback cell in which the source node or the switch in the cell's information field is both the source and destination (ie turning point) of the feedback cell. If there is no loop, the feedback cell can not return to the source node or switch without (actively) sending it back because it is transmitted on a link that only leads to the destination point of the road. If there is a loop, a feedback cell that has the source identity in the fields for both source and destination / feedback will return to the source node or exchange. The feedback indicator shows that the cell is traveling towards the turning point / feedback point. The node or switch detects the loop by realizing that it itself is both the source and destination of the feedback cell. It examines whether it is the destination of the feedback cell and if it is not the destination, the cell is transmitted on its logical channel. To identify that it is the source node of the feedback cell, it can use the source identity field or the correlation tag, or both. It must also check that the feedback indicator shows that the cell is moving in the direction of the feedback point.

In one embodiment of the invention, the source node may queue incoming traffic until a certain time has elapsed after a loop detection cell has been sent out, to ensure that information packets are not sent into a potential loop. This nd is selected so that there is a significant probability that if there really is a loop, the loop detection cell will return to the source node before time has elapsed. However, the elapsed time must not be too long as this leads to unacceptable delays in the handling of packages and requires large buffets. The time is therefore chosen depending on the type of information sent and the expected length of the road. In another embodiment of the invention, in order to avoid delaying packets unnecessarily, the node starts sending packets without delay on the way after the loop detection cell has been sent. This means that an assumption is made that the new path will not cause a loop. This assumption is acceptable in cases where the presence of new loops is unusual and / or it is unacceptable to delay packets.

Brief Description of the Drawings Figure 1 shows a schematic diagram of an embodiment of a simplified network according to the invention.

Figure 2 schematically shows an embodiment of a loop detection cell according to the invention.

Detailed description of modes of embodiment Figure 1, which will be used to show the basic idea of the invention, shows an example of a hypothetically simple unconnected network laid on top of a connection-oriented network consisting of nodes A, B, C, D and E. In this embodiment, each node has a disconnected unit, for the sake of example shown as route selectors AR, BR, CK DK resp. YOUR. It is of course possible to apply the present invention in hybrid networks in which non-connected units and / or connection-oriented units and / or combinations of non-connected and connection-oriented units are present.

The switches and switches are connected by links L1, L2, L3, L4 and L5.

Each node A-E has routing means, e.g. in the form of a dinging table, RTA, RTB, RTC, RTD, resp. RTE, which contains road information means, e.g. in the form of path information EA_B, Elg END, EA _;.;, etc. to EX.Y (where X is the source and y is the destination node), which shows how packets are to be routed from each node to the other nodes in the network. This information in the road selection tables is made up of road food identification bodies in a known manner, e.g. in the form of distance vectors, which each node transmits to its neighboring nodes in the network and which are not described in detail here. As an example, the route selection table for node A. RT ~ contains information on how information packets 10 15 9 512 are to be sent from node A to node B, from node A to node C, from node A to node D and from node A to node E Note that there are ways in which node A can send packets to node E, ie. via links L1, L3, L4, L5 or via links L2, L4, L5. Normally, node A would send packets to node E via links L2, L4, L5 because this provides the shortest and thus "cheapest" way (assuming that the cost of using a link is the same for all links in the network) so that the distance EM from the source node A to the destination node E would have the cost 3. If the link L2 is broken, node A will use the links L1, L3, L4, L5 instead, and the distance from node A to node E would increase to 4 and the path selection orientation would change to reflect this.

In the same way, node B has a distance to node E, EH which is 3, if node 3 is intact and which rises to 4 if node 3 stops working.

For node D, the distance to node E is EM; 1 unless link L5 is broken. In this case, the distance becomes infinite, ie. it becomes impossible to reach node E from node D.

The AR-ER selectors use their routing tables to establish connections through their switches as decided above. In a normal situation, therefore, the path selector AR requests, in a normal situation where it wishes to send tra fi k to E, a label from the node CR, ie. in the ATM case a logical channel number, CLEa-c for the traffic to E. Packets received for node E received in node A will then be forwarded to C through this channel.

The path selector CR asks node D for a label CLEc-d for traffic to node E and can then order its switch CS to connect CLEa-c with CLEc-d, so that the traffic from A to E is allowed to be handled by the switch CS and thus bypassed, and save resources in, the path selector CR.

In the same way, node X upstream of A, node A requests a label C LEx-a for tra- fi k towards E, which enables AR to command its switch AS to connect CLEx-a with CLEa-c. 10 IQ Un 10 These connections established as described above will follow the paths determined by the digxing tables RTx in the path selectors. If any event causes these path selectors to form a loop, this will lead to a connection leak.

If the link L4 is broken, e.g. following. It is discovered by the path selector CR who tries to inform his neighbors, wave wave arries AR and BR, that its distance to D and E is infinite. If, for any reason, this message to B is lost, AR will believe that it can reach D via B at cost 3 and AR will inform C about this. This leads to a routing loop.

Prior to this event, A has a channel CLEa-c to C for the traffic towards E and B has a corresponding channel CLEb-c for its traffic towards E.

Due to this event, the node A releases its channel to C and instead asks B for a channel to B, CLEa-b for its trajectory towards E. BS connects it to the existing channel CLEb-c. C asks A for a channel CLEc-a for its trajectory towards E. The switch AR commands the switch AS to connect CLEc-a with CLEa-b and the switch CR commands the switch CS to connect CLEb-c with CLEc-a so that a loop formed.

As long as the loop remains, all packets towards E will circulate in the loop and cause a waste of links and gear resources and eventually lead to problems such as overloaded gear buffers and the like.

The present invention reduces the problem of waste of resources by providing a loop detection means. In the preferred embodiment of the arrangement shown in Figure 2, the loop detection means comprises a specially modified maintenance feedback cell 20 which is generated and transmitted by a exchange when it establishes a communication channel to another exchange. The cell 20 has a starting field 18 and a payload 19. This loop detection cell 20 has the original gear indicated both in the source field 21 as the source and in the feedback field 22. This cell 20 may also have your feedback indicator field 23 which shows if the cell has been sent back. If the exchange AS 10 b) C> 512 055 ll is to establish communication with exchange ES, it therefore generates a loop detection cell 20 in which the cell contains AS as source and AS as feedback field. This cell 20 is then sent on the road that would lead to the exchange ES and normally it would never come back to the exchange AS. If the exchange AS later receives a cell containing AS in the fields for both source and feedback point (21, respectively 22) and if the feedback indicator 23, if present, shows that the cell has not been sent back, it means that a loop has formed. The source can also be identified in another source field, e.g. a correlation tag 24 as specified in the ITU-T 610 specification. Växel AS can then take appropriate measures to prevent the loop from being used and / or break the loop and / or send an alarm out of the network. In this way, a loop can be detected in the time it takes for the loop detection cell to reach the loop in the newly created path. In general, it can be assumed that the loops contain only a few nodes and links. The time it takes for a loop detection cell to recover from the loop is therefore generally short and a loop will be detected quickly with a method according to the invention.

If there is no loop, the loop detection cell will continue to the destination node and be discarded, or possibly sent back with the feedback indicator changed to “feedback”.

If there is a loop but the loop detection cell is dropped or stops before it returns to the gear that generated it, the loop will be removed when the time is normal.

In this preferred embodiment of the method according to the invention, the switch which generates the loop detection cell assumes that there are no loops along the path and forwards information packets on the path after the loop detection cell has been sent. This means that the information packets are not delayed, but can lead to some packets being lost if there really is a loop. However, the number of packets lost will be less than the number that would have been lost in any case if no loop detection cell had been sent. since the loop is detected earlier when the loop rejection cell is used. This embodiment therefore leads to improved transmission properties and no signal delay.

In a second embodiment of the method according to the invention, the gear which generates the loop detection cell assumes that there is a loop on the road. In this case, it has a timer that is set to go out a predetermined time interval after the loop detection cell has been sent on the alternative path. The time interval is selected to correspond to a mandatory maximum length of the loop. Such a maximum length can, e.g. selected to correspond to the time it would take for a packet to travel around a loop of any number of links, or may be a fixed number of transmission periods. Packages with a destination along the road are stored until the end of the timer period. If the loop detection cell has not yet returned to the exchange, it is assumed that no loop is firmed and the stored information packets are lost in the loop at the expense of introducing a delay in the transmission of one or more information packets.

The choice of which embodiment of the invention to use depends, of course, on the type of information in the data packets. A delay may be acceptable in a data transmission, where, however, it is not possible to lose any information packets without the quality deteriorating markedly. In a telephone call, on the other hand, it may be acceptable with reduced quality but not with a transmission delay. It is possible for a node or switch to determine which embodiment to use for each received packet: some are transmitted immediately while others are delayed.

In a third embodiment of the invention, which is not shown, the loop detection cell is generated at pre-programmed or predetermined intervals.

In a fourth embodiment of the invention, which is not shown, the loop detection cell is generated each time a path is established, but after a path has been established a pre-programmed or predetermined number of times. 10 13. 13 -12 us The network is of course not limited to networks with only five nodes and five links, but can be adapted to networks of any size.

While the invention idea has been shown with a disconnected network laid on top of a connection-oriented network as an example, it is of course possible to use the present invention in purely connection-oriented networks as well as in networks comprising a connection-oriented network or a switch.

The invention is not limited to loop detection means based on a feedback cell but may use any other cell that may be suitably designed or modified to inform a source unit that a loop exists.

Claims (6)

10 15 w i 'E12 G55 14 Patent claim A method for detecting loops in a network comprising nodes (A-E) and links (LI-LS) interconnecting said nodes, wherein at least one node comprises a connection-oriented network unit, e.g. a switch (AS-ES), characterized by the following steps: a) a connection-oriented source unit, e.g. an exchange (AS) generates a loop detection means (20) when it connects a communication path to a connection-oriented destination unit, e.g. with a gear (ES); b) said connection-oriented source unit (AS) sends said loop detection means to said connection-oriented destination unit (ES) over said communication path; and c) said connection-oriented source unit (AS) determines that a loop exists if said loop detection means returns to said connection-oriented source unit (AS). Method according to claim 1, characterized by the steps that d) said connection-oriented source unit (AS) starts a timer for loop detection at or after step b) above. e) said connection-oriented source unit (AS) stores packets for said connection-oriented destination unit (ES) until said timer expires; and f) said connection-oriented source unit (AS) sends said packet via said communication path if said loop detection means does not return to said connection-oriented source unit (AS) before the timer has expired. A method according to claim 1 or 2, characterized in that said loop detection means is a feedback cell generated by said connection-oriented source unit (AS), said feedback cell having fields identifying said connection-oriented network unit (AS) as both source and turning point. 10 15 512 055 Method according to any one of the preceding claims, characterized in that said network comprises at least one node (AE) comprising a connection-oriented network unit (AS, BS, CD, DS, ES) and a connectionless network unit (AR, BR, CR, DR, YOUR). A loop detection cell (20) comprising source identification means, e.g. source field (21) and / or correlation tag (24) and a turning point identifier, e.g. a turning point field (22), characterized in that the unit (AS-ES) identified in said source identification means (21-24) is the same as the unit (AS-ES) identified in said turning point identifier (22). Loop detection cell (20) according to claim 5, characterized in that it comprises feedback indicator part as a feedback field (23) to indicate whether said cell has been sent back.