SE514430C2 - Method and system for determining network topology - Google Patents

Abstract

The present invention relates to a method and a system for determining the topology of a network of nodes that are interconnected via unidirectional connections. According to the invention, the existence of a network loop within said network is determined using message forwarding among said nodes, and, as a result, information related to the existence of said network loop is distributed to nodes within said network.

Description

514 430 the local topology of the network is exchanged between the nodes of the network.

Based on received messages, each node will create and save its own map over the network, or at least part of it. An example of a since solution has been described in US 5,682,479 node in the network is arranged to send vector-routed packets (Newhall et al.), Where each over the network in different specified directions, where each packet collects information about the topology of the network along its path. The packets are then returned to the original node with the collected information.

As another example of the prior art, US 5,506,838 transmits from one or more source nodes to others (Flanagan) describes a solution where so-called explorer ~ packet nodes in the network and thus inform the network node about the topology of the network. OBJECTS OF THE INVENTION A disadvantage of the latter two examples of the prior art is that the methods consider the network in terms of bidirectional connections, each typically comprising two unidirectional point-to-point connections. For example, if any point-to-point connection between two nodes in the network lacks bidirectional connectivity, parts of the network topology may become impossible to determine. The calculation of routing tables and the like can also result in sub-optimized solutions.

An object of the invention is therefore to create a distributed method for determining network topology.

Another object of the invention is to create a method which does not have to rely solely on bidirectional point-to-point connectivity.

A further object of the invention is to create a method in which the amount of messages sent in the network for the purpose of determining its topology is kept down. Disclosure of the Invention The above and other objects of the invention are achieved by a method and a system in accordance with the appended claims.

The invention thus provides a method and a system for determining the topology of a network of nodes which are interconnected via unidirectional connections. According to the invention, the presence of a network loop within said network is determined by utilizing forwarding of messages among the nodes of the network.

Information related to the existence of said network loop is then distributed to nodes within the network.

The invention is thus based on the idea of considering the network in terms of network loops (at least in determining the topology of the network), and of determining and distributing information related to the existence of such loops.

According to a preferred embodiment, a node in the network sends a message from an output port, which is sometimes referred to hereinafter as a topology explorer message, and later determines receipt of a forwarded version of said message at an input port, thereby marking the presence of a network loop. In addition, each node in the network that receives a topology explorer message is preferably arranged to forward the message to at least one, typically all, of its ports.

Each, or at least a plurality, of the nodes of the network are preferably arranged to send out and detect messages of this kind. In addition, preferably all nodes in the network are arranged to forward such messages.

After the existence of such a loop is determined, information related thereto is distributed to nodes in the network, typically to nodes which form at least a part of said network loop and / or to other nodes in the network, i.e. nodes which do not form part of said loop. Such information typically includes information relating to which nodes, and which ports thereof, form part of said network loop. Generation and distribution of such information preferably also takes place through the use of forwarding of messages.

Advantageously, the transmission and retransmission of a single and relatively small message will be sufficient to determine the existence of a network loop. In addition, such a loop will be detected even if the network, and more specifically the loop as such, may include a number of point-to-point connections that lack bidirectional connectivity, which in some cases of prior art would be impossible. No. 5,506,838 discloses a method utilizing forwarding of messages, wherein For comparison, the above-mentioned US describes each node receiving so-called explorer packets will determine whether the information in these has already been received previously. If the information is new, the receiving node will record the packet information, modify the packet to identify itself, and forward the packet to other nodes. However, if the information has already been received, forwarding of the packet is terminated to prevent unnecessary transmission of messages. Thus, when a packet contains information previously received, no determination is made of the existence of a network loop or distribution of information relating to said network loop to nodes forming part of the loop, as the invention proposes. No use is made of the information obtained, except for the purpose of simply preventing further forwarding of the message.

A node which has two or more outputs, and which has not yet succeeded in determining which of its outputs forms part of a specific network loop, is according to an embodiment of the invention arranged to send two or more respective messages from the respective outputs, where each message identifies the respective output port from which it is transmitted and thus enables later determination of which of the two or more of the node's outputs that form part of a network loop. For example, transmission of such two or more different messages may be initiated by the reception of a topology explorer message or initiated by the transmitting node itself. These messages can then be used later, for example at the transmitting node or at another node which forms part of the network loop and which has received said message, or a forwarded version thereof, in order to identify the port which forms part of the current network loop. The advantage of this procedure is that with a simple and efficient mechanism it eliminates uncertainties regarding which gate forms part of which loop.

Expressed in terms of steps performed by the nodes of the network, a specific example of the above-mentioned embodiment comprises the steps of: sending a message from an output port on a first node; receiving said message, as such or in a forwarded version, at an input on a second node; sending two or more modified versions of said message from two or more respective ports on said second node, each modified version identifying the respective port used for transmission thereof; receiving one of said modified versions of said message at said first node, thereby identifying which of said two or more outputs forms part of said network loop; and sending a message from said first node to said second node, identifying the output port on said second node which forms part of said network loop.

Information regarding which nodes, and preferably also which ports thereof, form part of a determined network loop is generated in a preferred manner using forwarding of messages. Each node receiving a topology explorer message, or equivalent, will forward said message and will at the same time include in the message information related to the identity of the forwarding node, and typically also the node's output port. The loop information generated in this way can then be distributed to the nodes forming said loop, preferably also using message forwarding.

In summary, the solution of considering the network in terms of network loops, of determining the existence of such loops, and of distributing information relating to such loops, at least to nodes forming part of such loops, forms a clear inventive idea with a clear inventive step.

The above-mentioned and other aspects, features and details with respect to the invention will become further apparent from the following description of a preferred embodiment thereof.

Brief Description of the Drawings An exemplary preferred embodiment of the invention will now be described with reference to the accompanying drawings, in which: Figures 1a and 1b schematically show the respective network topologies; Figure 2 schematically shows a flow chart of an algorithm for detecting network topology according to a preferred embodiment of the invention; and Figures 3-7 schematically show an exemplary network and the exchange of messages between nodes in the network according to the exemplary algorithm for detecting network topology shown in Figure 2.

Detailed description of a preferred embodiment Figure 1a shows a simple network comprising three nodes 1, 2 and 3 connected via unidirectional connections 4, 5, 6 and 7. The connections 4 and 7 form a network loop comprising the nodes 1 and 2, and the connections 5 and 6 form another network loop comprising nodes 2 and 3, as marked by semicircular, dashed arrows.

The connections 4, 5, 6 and 7 can also be said to form an overall network loop, ie the loop from the emergency 1 via the connection 4 to the node 2, via the connection 5 to the emergency 3, via the connection 6 to the node 2, and via the connection 7 back to the node 1, which, however, is not specifically marked in figure 1a. Although it is possible to define such a loop, the description below will primarily focus on the discovery of the smallest possible network loops that each node forms part of. In addition, as can be seen from Figure 1a, each of the marked network loops in fact forms a bidirectional connection which interconnects the respective nodes.

Figure 1b shows another simple network comprising four nodes 10, 20, 30 and 40 which are connected via unidirectional connections to form a ring topology.

In this case, the ring topology forms a network loop that includes all four nodes, as marked by the semicircular dashed arrow in the middle of the figure.

A flow chart of an example of a topology detection algorithm according to a preferred embodiment of the invention will now be described with reference to Figure 2. The main object of this algorithm is to determine the presence of network loops of the type marked in Figures 1a and 1b and that provide the nodes that form part of each network loop with corresponding information.

Referring to Figure 2, the network topology detection algorithm includes a loop detection step S10, a master exclamation step S20, a crossroads reduction step S30, a loop list building step S40, a loop list distribution step S50, and a routing table calculation step S60.

Each of the steps in Figure 2 will now be described with reference to an exemplary network shown in Figures 3-7, which network comprises six nodes 10, 20, 30, 40, 50 and 60. The node 10 has two outputs 11 and l3 and two inputs l2 and 14. Each of the other nodes also has two outputs and two inputs designated in a corresponding manner. l5 l5 20 25 30 35 514 430 The node 10 outlet of the node 10 is connected via a unidirectional connection to the inlet 22 of the node 20, the outlet 21 of the node 20 is connected via an inlet 32 of the node 30, the outlet 31 of the node 30 is connected via a single directional connection to the input port 42 of the node 40, and the output port 41 of the node 40 is connected via a unidirectional connection to the input port 12 of the node 10, as shown in Figures 3-7. In addition, the node 43 output port 43 is connected via a unidirectional connection to the node 50 input 52, the node 50 output port 51 is connected via a unidirectional connection to the node 60 input port 22, and the node 60 output port 61 is connected via a unidirectional connection to the node 40 input port. 44. It is assumed in this example that the other inputs and outputs are not connected to other nodes.

An example of the loop detection step S10 of the topology detection algorithm in Figure 2 according to the preferred embodiment of the invention will now be described with reference to Figure 3. During the loop detection step S10, the network nodes will send out so-called probe messages used to detect the presence of loops in the network. A node sends probe messages from all outputs that do not form part of loops whose existence has already been determined. When a node creates probe messages, each message is provided with a unique identification that identifies the origin of the probe message, ie that identifies the output port from which the message is sent, for example the output's unique MAC address. The nodes of the network are arranged to forward received probe messages to all output ports. When a node forwards a probe message, the contents of the received message are mapped to the message being sent. The contents of the forwarded probe message will thus be essentially a copy of the contents of the received probe message, and will thus identify the output port of the node that originally sent the probe message. The distribution of probe messages is limited by a hop counting mechanism which limits the number of hops a probe message can be forwarded. For the sake of simplicity, it is assumed in the illustrated example that the number of jumps that a message is allowed to travel is set to four.

When a node receives one of its own probe messages from another node, it will determine that a loop exists, and it will then know which input forms part of this new loop. The node then transitions to work as a so-called build-up master for the new loop and continues to the master call stage for the new loop.

In the example shown in Figure 3, the node 10 sends a probe message PR (11) 20, which probe message identifies the probe message from its output port 11 to the node origin. The node 20 then forwards the probe message from its output port 21 to the node 30, after having incremented the hop count included in the probe message by one (1).

The node 30 forwards the probe message from its output port 31 to the node 40. it forwards the message both on the output port 41 to node 10, since the node 40 has two outputs, as on the output port 43 to the node 50. The node 50 will then forward the probe message from its output port 51 to the node 60. Since the message has now traveled the maximum number of jumps allowed, the node 60 will not forward the probe message. At the same time, however, the node 10 will have received its own probe message from the node 40 via the input port 12 and will therefore determine that a network loop exists and that the output port 11 and the input port 12 form part of this new loop. Node 10 will then assume the role of master node and proceed to the master exclamation stage.

An example of the master exclamation step S20 of the network topology detection algorithm according to a preferred embodiment of the invention will now be described with reference to Figure 4. When a node has determined the occurrence of a new loop using probe messages as described by referring to Figure 3, it will assume the role of master node for the new loop. In order to inform other nodes of the existence of the new loop, the master node sends out a so-called master call message on the output port that was previously identified as forming part of the new loop.

The master call message is forwarded according to the same rules as probe messages and serves two purposes. The first purpose is to inform other nodes of the existence of a new loop and to specify an identifier for the new loop, which is typically the MAC address mentioned above.

This loop identifier is included in all subsequent messages regarding the new loop and enables the handling of several loops at the same time without the risk of confusion of messages relating to different loops. Nodes that receive a master call become so-called build-up slaves for the new loop and automatically know which of their inputs forms part of the new loop. The second purpose of the master call is to provide a master node selection mechanism. If two nodes simultaneously receive their own probe messages, they will both try to assume the role of master node. In the preferred embodiment, this is solved by a selection mechanism based on the MAC addresses of the nodes. If a master node receives a master call (on the input for which it is currently trying to become the master node) from a node with a higher MAC address, it backs up, at least temporarily, to work as a slave node instead. If a mace node receives a master call from a node with a lower MAC address, the master call is not forwarded.

When the master node finally receives its own master exclamation, it knows that all other nodes on the new loop are now working as slave nodes and that it itself is the only master node for the new loop. The master node then proceeds to the crossroads reduction step. 20 25 30 35 514 450 ll loop identifier. This message is then forwarded in a similar manner to the probe message described with reference to Figure 3. Since the node 40 has two output ports, it will forward the master exclamation on both of its output ports. As each of the nodes 20, 30, 40, 50 and 60 receives the master exclamation MA (II), they will assume the role of slave nodes. When the node 10 receives its own master call from the node 40 via the input port 12, it knows that all the other nodes on the network loop are now operating as slave nodes, and it will therefore proceed to the crossroads reduction step.

An example of the cross-sectional reduction step S30 in Fig. 2 of the network topology detection algorithm according to the preferred embodiment of the invention will now be described with reference to Figs. Sa-5c. A so-called crossroads node is a node that has two or more connected outputs. When a node forwards a probe message or a master call message, it does so from all of its output ports, because it does not know which port forms part of the new loop. The goal of the crossroads reduction step is to determine which of the crossroads node's outputs forms part of the new loop.

The cross-country reduction step is initiated by the master node, which sends out a so-called cross-country call from the exit from which the node previously sent the probe message and the master call message. The crossroads call is provided with an identifier that identifies the exit port from which it is sent.

Crossroads exclamation is forwarded according to the following rule: If the forwarding node has only a single exit, or if it already knows which exit forms part of the new loop (ie it has already "solved" the crossroads issue), the node forwards the crossroads exclamation in unmodified form via the correct (or only) exit. Otherwise, the node is considered a crossroads node. It then sends out its own crossroads calls, instead of the received crossroads call, from all its ports. Each new junction call contains an identifier that identifies the output port from which it is transmitted.

When the master node receives a crossroads call with an identifier identifying an output port on another emergency, it knows that this output port forms part of the new loop.

The master node therefore informs the crossroads node about this by sending out a so-called crossroads reduction message which identifies this output port. The cross-country reduction message is forwarded in the same way as probe and master call messages. When a crossroads node receives a crossroads reduction message identifying one of the node's outputs, it knows that this port forms part of the new loop. The crossroads has now been "solved", and the crossroads reduction message does not need to be forwarded. When the next crossroads call is received by the emergency, it therefore forwards this in unmodified form from the now established output for the new loop.

When a crossroads node has been informed of which of its outputs forms part of the new loop, it sends out a so-called free message on all other outputs. This is done to inform slave nodes downstream of the gates that do not form part of the loop that they can now remove all protocol states ("states") for the loop and instead try to detect other loops.

Immediately after sending a crossroads reduction message, the master node sends out a new crossroads call to find the next crossroads node. The crossroads reduction step continues in this way to reduce, or solve, crossroads one at a time starting at the crossroads closest to the master node's input.

When the master node finally receives one of its own more unresolved crossroads on the loop. All nodes that form part of the crossroads exclamation, it knows that none of the loops exist now knows which of its outputs forms part of the loop, and all messages regarding the new race are in future sent only to nodes included in the loop. So far, however, 10 | _ | (j) 20 25 30 35 514 430 13 each node only information about its own ports, ie no node has complete knowledge of all nodes on the new loop (except in the case of very simple topologies). The master node therefore continues to the so-called loop list building step.

In the example shown in Figures 5a-5c, the master node 10 begins by transmitting a cross-country call SPA (11) from the output port 11 to the node 20. The cross-country call SPA (11) is provided with an identifier identifying the output port 11 from which it is transmitted. The exclamation point SPA (II) the nodes 20 and 30 to the node 40. Since the node 40 has been forwarded by two connected outputs, the node 40 sends out its own new crossroads exclamations SPA (41) and SPA (43) from its own respective outputs, where each exclamation point identifies the exit port from which it is sent. When the master node 10 receives the exclamation point SPA (41), the output port 41 forms part of the new loop. As from the node 40 in Figure 5a, it knows that it is identified in Figure 5b. in the same way as the previous messages. When the node 40 receives the crossroads reduction message SPR (41) identifying the output port 41, it knows that the port 41 forms part of the new loop.

Node 40 then sends a release message RB to the remaining output port 43. When nodes 50 and 60 then receive the release message RB, they stop working as build-up slaves and can start searching for other loops. Immediately after sending the cross-country reduction message SPR (41), the master node 10 sends out a new cross-country call SPA (11), as shown in Fig. 5c. Since the crossroads at node 40 has now been "solved", the new crossroads call SPA (11) will be forwarded all the way back to the master node 10.

The master node 10 will therefore determine that there are no more crossroads at the loop.

An example of the loop list building step S40 of the network topology detection algorithm according to the preferred embodiment of the invention will now be described with reference to Figure 6. During the loop list building phase, the master node collects enter information about which nodes, and which ports thereof, form part of the new loop by using message forwarding. The master node initiates the loop list building step (LB in the output port belonging to the new loop. Each slave node by sending out an empty loop list figure 6) from along the loop path to the master node input adds itself to the loop list and forwards the new loop list to the next node. When the loop list reaches the master node, the master adds itself to the end of the list.

The loop list building step is now completed and the master node has complete knowledge of the loop topology. It can then proceed to the step of distributing the loop list.

An example of the step of distributing the loop list of the network topology detection algorithm according to the preferred embodiment of the invention will now be described with reference to Figure 7. During this step, the master node informs all nodes forming part of the new loop about the loop topology using message forwarding. The same message format is used to distribute the loop list as the format used during the list's construction phase. The master node sends the list from the output port belonging to the new loop. Each slave node that receives a loop list during distribution forwards the list via the output port that belongs to the same loop as the input port on which the list was received. The master node, which originally sent out the loop list, monitors that the list returns via the input port that belongs to the same loop as the output port on which the list was sent. If the loop list is not received within a predetermined time interval, or if an error is detected by the node during the distribution of the loop list, the list is sent out again.

When the master node has received the complete and correct loop list in the same condition in which it was sent out, it will stop working as a master node and consider the new loop up and the condition of the loop as clarified. When a slave node has received the complete and correct loop list, it will correspondingly stop working as a slave node and consider the loop as up and the condition of the loop as cleared. The previous master and slave nodes then proceed to the step of calculating the routing table.

Once a node has received and accepted a new loop list, it will use the new loop list in the routing table calculation step to calculate an updated routing table based on the available loops that are considered valid. The routing table will contain an object for each reachable node. Each object identifies the output port that must be used to reach a destination and the MAC address of the destination input port.

As will be appreciated by those skilled in the art, the above steps may be altered, modified and / or integrated with each other.

In addition, steps may be added or removed based on the desired functionality within the scope of the invention, which is defined by the appended claims.

For example, the loop detection step, the crossroads reduction step and the loop list building step according to an alternative embodiment can be integrated into a single step, whereby the rules for handling loop detection messages may include features for reducing crossroads and building a loop list. An advantage of such an approach is that it reduces the amount of messages sent between nodes in the network. On the other hand, it can increase the size of the messages and the amount of message processing at the nodes.

According to another example, other means of restricting forwarding of messages than using a simple hop counting mechanism can be used.

Based on the concept of the invention, many different rules for handling topology messages can be used, for example to determine how and when a node is to become the structure master, how and when probe messages are to be sent, how to 1014 430}. CW and when new loops are to be sought, and so on, and the scope of the invention is of course not limited to the specific embodiment described in detail above.

Decisions regarding how the invention is to be realized and implemented will thus typically depend on how the specific network type is positively or negatively affected by aspects such as the amount of messages sent in the network, the size of the messages, the amount of processing of messages, states and the preservation of state ( in each node, further.

Claims (15)

10 15 20 25 30 35 514 450 17 CLAIMS A method of determining the topology of a network of nodes interconnected via unidirectional connections, the method comprising the steps of: determining the presence of a network loop within said network using forwarding of messages among said nodes; and distributing information related to the existence of said network loop within said network. A method according to claim 1, wherein the step of determining the presence of a network loop comprises the steps of: sending a message from an output port on a first node; and receiving a forwarded version of the message at an input on said first node. A method according to claim 1 or 2, wherein the step of determining the presence of a network loop comprises the step of receiving a message at an input port on a node and transmitting the message from one or more of the node outputs, A method according to any preceding claim, wherein the step of distributing information related to the existence of said network loop comprises using forwarding of messages to distribute said information. A method according to any preceding claim, wherein said information comprises information about which nodes form part of said network loop. A method according to any preceding claim, wherein said information comprises information about which ports form part of said network loop. 10, _ | UI 20 35 514 430 l8 A method according to any preceding claim, comprising the steps of: sending two or more messages from the respective two or more outputs on a node, each of which identifies the respective output port used for transmission thereof; and receiving a message relating to one of said two or more messages, thus identifying which of said two or more ports on said node forms part of said network loop. A method according to any preceding claim, comprising the steps of: sending a message from an output port on a first node; receiving said message, as such or in a forwarded version, at an input on a second node; sending two or more modified versions of said message from two or more respective ports on said second node, each modified version identifying the respective port used for transmission thereof; receiving one of said modified versions of said message at said first node, thus identifying which of said two or more outputs forms part of said network loop; and sending a message from said first node to said second node, identifying the output port on said second node which forms part of said network loop. A method according to any preceding claim, wherein forwarding a message comprises the step of including in the message information related to the identity of the forwarding emergency. 10 15 20 25 30 35 5-14 450 19 A method according to any preceding claim, wherein forwarding a message comprises the step of including in the message information related to the identity of the port from which said message is sent. A system for determining the topology of a network of nodes interconnected via unidirectional connections, comprising a first node arranged to send a message from one of its ports, to determine the presence of a network loop within said network by determining reception of a forwarded version of said message at one of its inputs, and as a result distributing information related to the existence of said network loop to nodes within said network. A system according to claim 11, comprising one or more other nodes arranged to forward said message on one or more of their outputs upon receipt of said message on one of their inputs. A system according to claim 11 or 12, wherein said nodes are arranged to distribute said information related to the existence of said network loop by utilizing forwarding of messages. A system according to any one of claims 11-13, wherein a node having two or more ports is arranged to send two or more respective messages from respective ports, each message identifying the respective port to which the message is sent. from and in this way enables later determination of which of said two or more outputs on said distress forms part of said network loop. 10 15 514 450 20 A system according to any one of claims 11-14, comprising: the first node arranged to send a first message; a second node arranged to receive said message, as such or in a forwarded version, and to send two or more modified versions of said message from two or more respective output ports, each modified version identifying the respective output port used for transmitting thereof, said first node being arranged to identify which of said two or more outputs on said second node forms part of said network loop by determining reception of one of said modified versions of said message at said first node and as a consequence send a message to said second node identifying the output port of said second node which forms part of said network loop.