FI91690B - Hierarchical synchronization procedure and a data traffic system that uses message-based synchronization - Google Patents

Description

91690

The invention relates to a hierarchical synchronization method according to the preamble of appended claim 1, which is used in a communication system using message-based synchronization. The invention also relates to a communication system according to the preamble of appended claim 6.

In this representation, the intersections of the transmission links of the system are referred to as a node. The node can be any device or hardware that is capable of interfering with the clock rate, e.g., a branch or cross-connect device.

15 The nodes of a system using message-based synchronization are interconnected by the transmission connections they use to transmit data. The connections used also forward the sender's clock frequency to the recipient. Each node selects as the source of its own clock frequency either the frequency of a signal from one of the 20 neighboring nodes or the frequency of its own internal clock source. In order to make all the nodes of the system work at the same clock frequency, the aim is usually to get the system to synchronize to one clock source, the so-called the master source. In this case, all nodes in the system directly connected to the selected main source are synchronized to this main source, and nodes connected to these but not directly connected to the main source are synchronized to these nodes adjacent to the main source. Correspondingly, nodes that are always farther away from the main source 30 are always synchronized to those nodes that are one connection distance closer to the main source.

In order to build a synchronization hierarchy as described above inside the system, the nodes of the system transmit synchronization messages to each other. These 35 messages contain information that allows individual 91690 2 nodes to select the timing source. The nodes in the system are prioritized and the system tends to synchronize to the clock frequency of the node that is at the highest priority level. At the same priority level, nor-5 has only one node in the system painted. Normally, the synchronization messages contain information on who the clock frequency of the node sending the message originates from, which is the node priority, and a value that describes the quality of the clock signal. This allows a single node to select the clock frequency of the neighboring node that originates from the desired node and is of the best quality as the source of its own clock speed.

In the system start-up phase, each node selects its own internal clock-15 source as its clock frequency source because no incoming synchronization message has been processed. When the first incoming synchronization messages have been processed, the clock frequency of the neighbor with the highest priority is selected as the source of its own clock frequency. Once the system has reached a synchronization-stable state after all messages have spread to the system, the system is hierarchically synchronized to the clock frequency of the main source.

Figure 1 shows a system using message-based synchronization in a stabilized situation. The priorities assigned to the others of Sol-25 are marked with numbers inside the circles describing the nodes. The lower the number, the higher the priority of the node. The synchronization messages sent by node n (n = 1 ... 6) are marked with the reference number MSGn. The synchronized message sent by each node is usually different and depends on the message-based synchronization method used. The propagation of the clock frequency from the master clock (Node 1) to the other nodes of the system is shown by solid lines. Dashed connections between nodes are not normally used to synchronize the system, but they are

II

91690 3 are available in case of change.

The simple principle in message-based synchronization is that the user defines a node synchronization hierarchy, giving each node its own identifier indicating the node level in the hierarchy, and the system synchronizes to the defined master clock independently using all existing connections between nodes if necessary. ). If the connection to the master clock is lost and no alternate connection 10 exists, or the master clock fails, the system synchronizes to the next highest level node. Fig. 2 shows a situation when the main clock fails in the system according to Fig. 1. The response to the change in synchronization takes place by means of an inter-node messaging 15. When the incoming timer is interrupted, the synchronization hierarchy is rebuilt from the point of interruption onwards (away from the system master). This is done, for example, in such a way that the node that first notices the interruption first goes to the internal timing mode for a certain period of time and forwards 20 information about the change. When the next node notices the changed situation, it also goes to the internal timing mode for a period of time and sends information about the change, etc. When the deadlines of the individual nodes expire, the formation of a new synchronization hierarchy begins.

25 Usually, the end result is a hierarchical structure similar to the original, in which the failed connection is replaced by a functioning connection, while the rest of the structure remains almost unchanged.

A network using message-based synchronization is described, e.g., in U.S. Patents 2,986,723 and 4,837,850. Both patents disclose methods that use time limits depending on the size and shape of the system in system failure situations, during which the nodes are forced to enter a predetermined constant state to prevent incorrect synchronization in connection with failure conditions 91690 4. In the event of a fault, information about the fault is transmitted as described above by means of system messages. Once the information about the changed situation has spread throughout the system or over a sufficiently large area, 5 synchronizations are rebuilt at the change point and possibly further afield, if necessary. Deadlines are used to ensure that information about the change is spread over a sufficiently wide area. Upon detecting a change / failure, the node forwards this information and starts its own timer. When the time limit has elapsed, the node restarts its normal operations to obtain the timing, and the system begins to resynchronize those parts that were affected by the change / failure. The aforementioned U.S. Patent 2,986,723 specifically discloses a system in which said constant state is an internal timing state in which a node uses its own internal clock as the source of its timing. The method disclosed in this patent is hereinafter referred to as Self-Organizing Master-Slave Synchronization (SOMS) and will be used as an example in the further description of the invention.

It should also be noted that when we refer to a deadline in this context, it means the predetermined time by which the system is intended to prevent the acceptance of erroneous / outdated synchronization messages.

When using the SOMS method referred to above, the aim is to remove erroneous synchronization identifiers from the network instead of the fixed internal timing of the nodes.

20 When the node returns to the normal state, the synchronization identifier of any incoming message does not contain at least too positive erroneous information about the original source of the synchronization, i.e. the synchronization is in no case claimed to come from a node with a higher hierarchy level than it actually does.

II

91690 5 currently has. U.S. Patent No. 2,986,723 discloses a time for the duration of a forced internal timing during which a significant portion of the nodes in the system have detected the change. This significant part is not defined in more detail, 5 but the absolute number of nodes required for a significant part is found to depend only on the total number of nodes in the system. In this case, in each error situation, most of the part of the system that gets its synchronization via the failed connection or node 10 goes into the internal timing state. Only then will synchronization be completely rebuilt in this part of the system.

Thus, in different synchronization methods, the deadlines depend on the size and shape of the system in order to ensure that the information about the change / failure has spread over a sufficiently wide area before resynchronization is started. As a result, setting deadlines is difficult. In large systems, deadlines easily become too long, making it impossible to maintain the required level of quality in the timing. This is because when a node waits for a timeout, it usually does not have the system master clock frequency available to it. As the system grows, deadlines must be redefined and reported separately for each of the 25 nodes, making it difficult to grow the system.

It is an object of the present invention to provide a method and system which do not have the disadvantages described above. This is achieved by the method according to the invention 30, which is characterized by what is described in the characterizing part of the appended claim 1. The telecommunication system according to the invention is in turn characterized by what is described in the characterizing part of the appended claim 6.

The idea of the invention is to select the minimum time in which the (nearest) environment of the node has reacted to the change and in addition the node has received information about it. The most preferred embodiment is, of course, the one in which the time limit is exactly 5 minimum times.

Thanks to the solution according to the invention, the synchronization of the system is speeded up in the event of a fault, and it is not necessary to use low-quality clock sources as long as before. Also, the deadline no longer depends on the size or shape of the system 10, so it can be defined as a constant constant already during the construction phase of the system. There is therefore no need to change the deadline as the system expands, which makes the extension even simpler.

The invention and its preferred embodiments will now be described in more detail with reference to the examples of Figures 3-9f in the accompanying drawings, in which Figure 1 shows a message-based synchronization system in general form synchronized to the main source clock frequency, Figure 2 shows the network of Figure 1 when its master node is failed, Fig. 3 shows the network using self-directed slave synchronization (SOMS) in its initial state, Fig. 4 shows the network of Fig. 3 in a stable state, Fig. 5 shows the resynchronization of the network of Fig. 4 after its main node fails, Fig. 6 shows the resynchronization of the network of Fig. 4 to two nodes Fig. 7 shows the state of a system using message-based synchronization during a change situation, Fig. 8 shows the elements of a single node in which the method according to the invention is implemented, and Figs. 9a-9f show as a series of events of the invention

II

91690 7 in the SOMS system.

Figure 3 shows a system using Self-Organizing Master-Slave synchronization (SOMS) according to the aforementioned U.S. Patent 2,986,723, comprising in this case five nodes (or devices) denoted by reference numerals 1 ... 5 according to their hierarchical level. (The main node of the network has the smallest SOMS address.) The nodes forward messages containing the above SOMS addresses. This allows the nodes to identify each other using these address numbers and build a synchronization hierarchy, allowing the entire network to synchronize with the master node.

As mentioned above, the messages transmitted continuously in the network depend on the message-based synchronization method used. In addition, the messages are unique for each sending node. In a SOMS network, a synchronization message comprises three different parts: a frame structure, an identifier, and a checksum. The SOMS identifier is the most important part of the SOMS message. It consists of three consecutive numbers D1 ... D3:

D1 is the origin of the synchronization frequency of the sending node of the SOMS message, i.e. the SOMS address of the node visible to the sending node as the master node.

D2 is the distance to the node expressed by D1. This distance is expressed as the number of nodes in between.

D3 is the SOMS address of the sending node.

Each node (or device) continuously compares incoming SOMS tags with each other and selects the smallest of these. In the identifier, the parts D1, D2 and D3 are combined directly into one number by putting them in sequence (D1D2D3) (for the sake of clarity, a dash will be written below to separate the different parts; D1-D2-D3). Thus, the primary selection criterion for the smallest address becomes the SOMS address (D1) of the node visible to the previous nodes as the main node, i.e. 35 nodes tend to synchronize to a signal with a frequency of 91,690 8 originally coming from the node with the smallest address. In this case, in a stable situation, the entire network is synchronized to the same master node (because the master node of the entire network has the smallest SOMS address).

5 If two or more of the incoming signals are synchronized to the same main node, the one with the shortest route (D2) is selected. The last selection criterion remains the SOMS address (D3) of the node sending the SOMS message, on the basis of which a selection is made if otherwise no difference is obtained between the incoming signals.

Once a node has accepted one of the neighboring nodes as its new synchronization source based on the incoming SOMS identifier, the node has to re-establish its own SOMS identifier. The new SOMS identifier can be derived from the 15 selected smallest SOMS identifiers as follows: the first part (D1) is left untouched, the second part (D2) is incremented by one and the third part (D3) is replaced by the node's own SOMS address.

Each node also has its own internal SOMS identifier X-O-X, where X is the SOMS address of the node. If none of the incoming SOMS messages contain an identifier that is less than the internal identifier, the node uses its own internal oscillator or possibly a separate synchronization input as the source of its clock frequency. The outgoing 2.5 outgoing SOMS message naturally uses the internal SOMS identifier.

The nodes continuously send SOMS messages in each direction so that the changed information in the SOMS identifiers spreads as quickly as possible and the operation of the neighboring nodes is constantly clear. Before SOMS tags can be compared, incoming SOMS messages must be accepted and SOMS tags must be separated from them.

When a SOMS message is first received from a particular transmission link, its SOMS identifier is immediately accepted for comparison if the message was error-free.

Il 91690 9

When the incoming transmission link has an accepted SOMS identifier and an error-free message containing the same identifier arrives continuously, the situation remains unchanged. If an SOMS message is found to be invalid, the old SOMS identifier is retained until three consecutive SOMS messages are received as invalid. In that case, the SOMS tag for comparison. Waiting for three consecutive SOMS messages is intended to eliminate momentary interference.

10 If no SOMS message arrives from the connection, even if the connection is otherwise working, wait three consecutive SOMS messages until the current SOMS identifier is discarded. If the connection is completely lost, the SOMS identifier is rejected immediately. If, due to interference in the incoming signal, a valid SOMS identifier is not obtained for comparison, the the SOMS identifier of the transport connection. In this case, the as a SOMS identifier of the incoming transmission link, a standard identifier in which all parts (D1, D2 and D3) receive their maximum value -20 (MAX-MAX-MAX).

When a new changed SOMS identifier is detected in an incoming SOMS message, it is immediately accepted for comparison if the message was error-free. This does not cause unnecessary delays for network changes.

25 In the initial situation, each node uses its own internal synchronization source, in which case it sends its own internal SOMS identifier X-O-X to the other nodes. This tag is also compared to incoming SOMS tags.

If none of the incoming tags is less than the internal 30 tags, the node to use its own internal timing.

• In Figure 3, the SOMS network is shown in the initial state, when no node (or device) has had time to process incoming SOMS messages. For all nodes, the highest priority is given to the node's internal SOMS identifier, because others have not yet been processed. In Figure 3, the incoming SOMS tags are marked next to each node, and the selected tag is written inside the frame (in the initial situation 5 according to Figure 3, all nodes use their internal timing source). The connections in use for synchronization are drawn with a solid line, the spare connections with a broken line (in the initial situation according to Figure 3, all connections are spare).

10 When the nodes have time to process the incoming SOMS messages, the node 1 remains in the internal timing, the nodes 2 and 4 are synchronized to the node 1 based on the identifier 1-0-1, the node 3 is synchronized to the node 2 (2-0-2) and the node 5 is synchronized to the node 3 (3-0-3). At the same time, the nodes generate their own new SOMS identifiers as described above and exchange the new identifier for their outgoing SOMS message.

The state of the network after its stabilization is shown in Figure 4.

All nodes are synchronized to the main node 1 by the shortest possible route.

If, in the case of Figure 4, the main node fails, the nodes 2 and 4 immediately forcibly enter the internal timing state after losing the incoming SOMS identifier 1-0-1. When nodes 3 and 5 detect a change in nodes 2 and 4, they also forcibly go to internal timing. When Node 2 returns to normal, it receives from its nodes 3 and 4 their internal SOMS tags (3-0-3 and 4-0-4) and stays in the internal timing because there is no smaller SOMS tag from the outside than its own internal tag (2- 0-2). Node 4 in turn synchronizes to node 2. After stabilizing, the network is in the state shown in Figure 5, where Node 2 has become the new main node of the network.

If, for example, only the connection between nodes 1 and 2 is lost (Fig. 6), only node 2 is forced to enter the internal timing state. When it returns to normal mode, it synchronizes 35 to node 4, which is connected to the main node of the network. Once the entire network li 91690 11 has stabilized, the synchronization is still from node 1 despite one interruption. This situation is shown in Figure 6.

When the time limit according to the invention is used to prevent the occurrence of outdated and erroneous synchronization messages, the situation in the system is as shown in Fig. 7. From the point of change (the common center of the circles), we move away to zone 70, where the nodes are forcibly within a certain constant time limit of 10 (this does not apply to nodes between the point of change and the main body of the system, which operate normally all the time because they are not affected by the fault / change). synchronization). At some point, the situation is e.g. as shown in Figure 7; the nodes 15 within the innermost zone have already returned to the normal state, the nodes in the forced time zone are in the region of the next zone (70), and the nodes operating on the basis of the old data are still in the region of the outermost zone. The synchronization of the system is re-established when the zone 70 20 has reached the edge of the area and exited, whereby all the nodes again operate according to the valid data.

However, an individual node does not have to wait for the entire system to detect the change and go to the 25 defined steady state. It is sufficient that the environment of the node knows the changed situation and has reacted to it. In this case, the time period to be selected must be so long that all neighboring nodes synchronized to a particular node have time to enter the defined state, and not the the node no longer receives viral synchronization messages when it returns to normal mode and begins to select the best synchronization identifier again. The minimum duration of the time period K can be calculated using the following formula: (1) K = 2 x (S + H + V), 35 where S is the maximum duration for generating a synchronization message and its transmission between two nodes, H is the time for accepting the synchronization message. the maximum duration at the node and V is the maximum duration at the node of the comparison of the incoming synchronization identifiers. The time period according to Equation (1) does not depend at all on the size of the network, but is a constant depending on the node implementation, the transmission rate used in the system, the synchronization method and message length, and the system transmission delays, which can be determined during the system construction phase.

10 Equation (1) takes into account, first, the time from neighboring devices of a node that has entered a forced steady state to detect that that node has entered that state, and secondly, the time that a node 15 that has forcibly entered said state has reacted to before other nodes respond. to the changed situation and have time to go through the changed inbound sync tags themselves.

Figure 8 shows the elements of the node in which the method according to the invention is implemented. The figure shows two connections to the system node from neighboring nodes, A and B. The transmission line of each connection is connected to the signal transmission and reception section 13a and 13b, respectively, which perform the processing of the physical signal. Parts 13a and 13b forward the synchronization message 25 to the synchronization message transmission and reception section 16a and 16b, respectively, connected thereto. The sending and receiving part of the synchronization message performs e.g. checking the error of the message and forwarding the message to the central synchronization decision-making section 30 of the node 23, the input of which is connected to the output of the respective receiving section 16a, 16b. The signal transmitting and receiving sections 13a and 13b also monitor the quality of the received signal and store the information thereon in the connection-specific fault databases 24a and 24b, respectively. The sending and receiving section 16a of the Sync-35

II

91 6? ϋ 13 data from the database 24a and the transmission and reception section 16b from the fault database 24b, respectively. The detection of a connection fault / change in the signal transmission and reception parts takes place as is known per se.

The decision-making section 23 compares the messages and stores them in the memory 21, e.g. in order of priority, so that the synchronization identifier is always selected at the top. The decision-making section also receives the fault information of the corresponding signal from the connection-specific transmission and reception block 11a or 11b, either in the form of a synchronization message or as separate fault information. When the decision-making part detects from the information received that the node has to go to a fixed constant state for a certain time, it selects its source in the synchronization method used for the situation 15, gives a corresponding for parts 16a and 16b and starts its timer member 25. With the new identifier 20, the node indicates the change that has taken place to its neighboring nodes.

After the timer members 25 have notified the expiration of the period K, the decision section 23 may again select the timing source according to the normal procedure.

Figures 9a-9f show as a series of images a solution according to the invention 25 in a SOMS system. The nodes in the forced internal timing at each moment are marked by drawing a line below their number. In the initial situation (Fig. 9a), the part of the system shown has a connection to the rest of the system and thus also to the main node of the system 30 via the node 17. The node receives from the other system a synchronization identifier 1-6-16 originating from the main node of the system (Node 1) and transmitted by the node 16. Upon detecting the loss of synchronization, the node 17 forcibly enters the internal timing state (Figure 9b), 35 starting to send the internal synchronization identifier. 14 916 * 0 17-0-17. The other nodes have not yet detected the change. In the next step (Fig. 9c), the nodes 18 and 19 have detected the change and are themselves forced to enter the internal timing mode, starting to send out their own internal timing identifiers 18-0-18 and 19-0-19, respectively. Thereafter (Fig. 9d), after the nearest nodes of the node 17 react to the change and the node 17 is informed about it in the form of these changed identifiers, the time limit (K) of the node 17 expires, and the node is again free to choose the source of its timing. However, since neither of the incoming tags is better than the node's own internal tag, the node 17 stays in the internal timing, free to choose a better tag as soon as it arrives. At this point, the node 20 has also detected the change and entered the internal timing mode for the time period described above.

In the step shown in Figure 9e, the deadlines of nodes 18 and 19 have expired and both select their new timing source based on the clock frequency identifier 17-0-20 17 of node 17. Eventually (Fig. 9f), the amount of time of node 20 also expires and the part separated from the rest of the rest of the system is synchronized to the clock frequency of node 17 when no alternative route to the main node of the whole system is found.

When in a predetermined standard state, such as an internal timing state, in the case of a SOMS network as described above, it is preferable for the node to receive and process incoming synchronization identifiers at all times. In this case, after the expiration of the time limit, it already has a pre-formed synchronization priority list (memory 21, Fig. 8), from which a new timing source can be selected immediately.

From the point of view of system synchronization, it is also advantageous that all nodes have the same deadline. In this case, 35 unnecessary changes and fluctuations in the synchronization are omitted li 91690 15, because the nodes return to the normal state in order. Thus, no islands are formed for synchronization either, because different parts of the system do not react to changes at different rates. In this case, the time limit in question 5 is determined by the node with the longest time determined according to Equation 1.

From the point of view of system synchronization, it is also advantageous to ensure the bidirectionality of the connection used before it is used for synchronization. If the connection cannot be reliably determined to be bidirectional, its use for synchronization is prohibited. In this way, it can be ensured that the synchronization cannot be confused under any circumstances, despite the reduced time limit. This could be the case, for example, when when a node returns to its normal state after an even shorter period of time, it receives, for example, an outdated synchronization identifier via a one-way connection (information about the change has not reached neighboring nodes because the connection to them is one-way).

In order for a connection between two nodes to be found to be one-way, communication between the nodes is required. The practical implementation of communication can range from the transmission of one bit to a handshake by means of messages. The latter option is advantageous in that it makes it possible to determine with certainty the two-way orientation of the connection without any presuppositions about the system. When each connection is turned on, both parties to the connection shake the connection two-way before sending the first actual synchronization message. This handshake is performed again after each interruption and is performed in the transmission and reception parts of the synchronization message (cf. Fig. 8). Ensuring bidirectionality is described in more detail in co-pending patent application FI-92xxxx, to which reference is made for a more detailed description.

Although the invention has been described above with reference to the examples according to the accompanying drawings 91690 16, it is clear that the invention is not limited thereto, but can be modified within the scope of the inventive idea set forth above and in the appended claims. Thus, although the SOMS system has been used as an example above, the solution according to the invention is applicable to all such systems using message-based synchronization, in which the aim is to prevent the illumination of erroneous / outdated synchronization identifiers in a fixed period mode. Although an example has been described above in which the timer members measure a predetermined time limit K, it is also possible to implement the invention so that the end of the time limit is tied to the detection of changed identifiers from neighboring nodes.

Il

Claims (6)

A hierarchical synchronization method for a data communication system using message-based synchronization, said data communication system comprising a plurality of nodes interconnected with connections (A, B), in which the nodes transmit each other signals containing synchronization messages which include information about a corresponding signal's priority internal synchronization hierarchy, and in which method a node in a change situation, such as an error situation, transfers to a predetermined default state for a predetermined period of time to prevent any incorrect synchronization messages, characterized as being used as said fixed time period as It is preferable that a neighbor's nodes receive information that the node has given over to said standard state, reacted to the change and the node itself has received information about their response from the neighbors. 20 2. A method according to claim 1, characterized in that for all nodes in the system the same fixed period of time is used, which is determined by the node whose time period determined in this way is the longest. 3. A method according to claim 1, characterized in that the node receives and processes incoming synchronization messages even during said fixed period of time, but selects the source for its timing due to these having expired only after said fixed period of time. 30 4. A method according to claim 1, characterized in that the bi-directionality of the connection between two nodes is detected, and as soon as the connection is not bi-directional, the use of this connection for synchronization is prohibited. 91690 20 5. A method according to claim 4, characterized in that the connection is bi-directional by ensuring that a bi-directional handshake is performed between the nodes at least always after such a change which has led to the transition to said standard condition. 6. Data communication systems using message-based synchronization, which system comprises a plurality of nodes interconnected with connections (A, B), which nodes transmit each other signals containing synchronization messages, which include information on the priority of a corresponding signal in the system's internal synchronization hierarchy, and in which system a node in a change situation, such as a failure situation, assigns to a predetermined default state for a predetermined period of time to prevent the occurrence of incorrect synchronization messages, characterized in that a node in the system comprises timer means (25), which measure the predetermined time period (K), which corresponds to the time that eats for a node's neighbor nodes to receive information that the node has passed to said standard state, reacted to the change and the node itself from the neighbor nodes fetch information about their reaction. IN!