KR101192896B1 - Distributed synchronization method and apparatus for fault tolerance - Google Patents

Abstract

The present invention relates to a fault-tolerant distributed synchronization method, wherein step S1 in which any one of a plurality of time masters is designated as a starter node, step S2 in which a sync time master is synchronized with the starter node, a starter node and the sink time master The sync frame of S3 may be periodically transmitted to the time slave and the S4 step of synchronizing with the plurality of time masters using the sync frame transmitted by the time slave.
The present invention provides a distributed synchronization method and apparatus for synchronizing a slave based on an average value of offset values calculated from sync frames transmitted from a plurality of time masters, so that even if a defect occurs in some masters.

Description

Fault-tolerant distributed synchronization method and device {DISTRIBUTED SYNCHRONIZATION METHOD AND APPARATUS FOR FAULT TOLERANCE}

The present invention relates to a distributed synchronization method and apparatus used for synchronization in an Ethernet based system. The present invention relates in particular to a distributed synchronization method and apparatus capable of fault tolerance using a plurality of time masters in an Ethernet based system.

Motion control networks based on Ethernet communication are mainly used to transmit high speed real time data such as synchronous motion control. In this system, IEEE 1588, a clock synchronization standard, is widely used for synchronization. IEEE 1588 is used in EtherCAT, PowerLINK, etc.

In the IEEE 1588 synchronization method, one time master synchronizes a plurality of time slaves. Thus, if the time master fails, there is a problem with the entire synchronization. In this case, in a safety-critical application such as an automated robot, a great life or property problem may occur.

To prevent this, if a problem occurs in the IEEE 1588 Best Master Clock Algorithm (BMCA) or the existing master, a master redundancy technique is widely used where the backup master takes over the role of the existing master.

In the case of BMCA, the best one is selected among several masters and the network is synchronized based on the master. At this time, if a fault occurs in the elected master, the election algorithm is performed again to select a new best master. In case of BMCA technique, the problem of existing one master can be solved, but it is difficult to provide high performance synchronization because the time required for re-election process is long.

On the other hand, the master redundancy technique takes a long time to replace because the backup master has to perform operations such as initializing to replace the existing master.

The time required to run the backup master also requires synchronization of the entire network, which can not be solved.

In addition, since a single time master is used, if a time master fails and a wrong time stamp is transmitted, the entire network synchronization may not be performed properly.

The fault-tolerant distributed synchronization method and apparatus according to the present invention aims to solve the following problems.

First, even if the master fails, the entire network will be synchronized normally.

Second, even in the case of a failure of the master, synchronization is to be performed immediately without a short time.

Third, even if a defect occurs in one or more of the plurality of masters, synchronization of the entire network is normally performed.

Fourth, it is intended to reduce the load of the entire system by placing a plurality of time masters that are involved only in synchronization among the masters.

Fifth, the defect-free synchronization ensures that data is transmitted reliably and efficiently over the network.

Sixth, applications such as automation robots and shipbuilding robots, in which stability and synchronization performance are important, operate stably.

The solution to the problem of the present invention is not limited to those mentioned above, and other solutions not mentioned can be clearly understood by those skilled in the art from the following description.

The present invention relates to a distributed synchronization method capable of fault tolerance.

According to the method of the present invention, a step S1 in which one of a plurality of time masters is designated as a starter node, a step S2 in which a sink time master, which is a time master other than the starter node among the plurality of time masters, is synchronized to the starter node, and a starter node and a sink And a step S3 in which the sync frame of the time master is periodically transmitted to the time slave and a step S4 in which the sync frame of the time master is synchronized with the plurality of time masters using the sync frame in which the time slave is transmitted.

The step S1 of the present invention includes a step of checking whether a plurality of time masters are normally operated and a step of specifying one of the normally operating time masters.

 Step S2 of the present invention includes the sync time master being synchronized to the starter node using the sync frame transmitted from the starter node.

In the step S4 of the present invention, the time slave uses a plurality of clock frames to receive a plurality of sync frames, a step S4-1 to calculate a plurality of clock offset values from the plurality of sync frames, and a plurality of clock offset values. S4-3, wherein the slave is synchronized to the plurality of time masters.

Step S4-3 of the present invention includes synchronizing the time slave based on the average value of the plurality of clock offset values calculated in step S4-2.

The step S4-3 of the present invention includes synchronizing the time slave based on the calculated average value except for the maximum value or the minimum value among the plurality of clock offset values calculated in the step S4-2.

The step S4-3 of the present invention includes synchronizing the time slave based on the calculated average value except for the maximum value and the minimum value among the plurality of clock offset values calculated in the step S4-2.

Step S4-3 of the present invention includes performing a plurality of steps of excluding a maximum value and a minimum value among the plurality of clock offset values calculated in step S4-2 and synchronizing the time slave based on the average value of the remaining offset values. do.

Excluding the maximum value and the minimum value that are performed a plurality of times in the present invention includes the execution of a specific number of times input as a limit condition that the remaining clock offset value except for the maximum value and the minimum value is one or more.

The method of the present invention uses a first mode or a distributed synchronization method according to any one of claims 1 to 9, wherein only one of the plurality of time masters is used as a master to synchronize a time slave by an IEEE 1588 synchronization method. The time master includes a second mode for synchronizing the time slave, and includes a mode change between the first mode and the second mode.

The present invention relates to a distributed synchronization device capable of fault tolerance.

The apparatus of the present invention is networked with a designated starter node among a plurality of time masters, a sync time master which is a time master other than the starter node among the plurality of time masters, and a plurality of time masters, and a plurality of sync frames from the plurality of time masters. It includes a time slave synchronized with the received.

The starter node of the present invention includes designating among the time masters that operate normally after checking whether the plurality of time masters operate normally.

The starter node of the present invention includes transmitting a sync frame to the sync time master to perform synchronization between the starter node and the sync time master.

The time slave of the present invention includes calculating a plurality of clock offset values from the transmitted plurality of sync frames, and synchronizing using the calculated offset values.

The time slave of the present invention includes synchronizing on the basis of an average value of the calculated plurality of clock offset values.

The time slave of the present invention includes synchronizing on the basis of an average value of the offset values other than the maximum value or the minimum value among the calculated plurality of clock offset values.

The time slave of the present invention includes synchronizing on the basis of an average value of the remaining offset values except for the maximum value and the minimum value among the calculated plurality of clock offset values.

 The time slave of the present invention includes synchronizing on the basis of an average value of the remaining offset values excluding a plurality of maximum and minimum values among a plurality of calculated clock offset values.

The time master of the present invention extracts a time stamp from a master communication module connected to a time slave and a time master through a network, a clock received from a clock source, a master clock management module storing a current clock, and a clock value of a clock management module. And a master synchronization module for controlling a time stamp module, a clock management module, and a communication module mounted in the communication module to perform a synchronization algorithm.

Time slave according to the present invention is a slave communication module for receiving a sync frame associated with the synchronization over the network, the clock information according to the synchronization algorithm of the slave synchronization module and the slave synchronization module to perform a synchronization algorithm using the sync frame received in the slave communication module. It includes a slave clock management module for storing the.

The present invention provides a distributed synchronization method and apparatus in which the entire network synchronization is normally performed even when a part of the time master fails using a plurality of time masters.

The present invention provides a distributed synchronization method and apparatus in which effective synchronization is performed by placing a plurality of time masters that perform only synchronization functions, rather than having a plurality of masters controlling the entire system.

The present invention provides a distributed synchronization method and apparatus for synchronizing a slave based on an average value of offset values calculated from sync frames transmitted from a plurality of time masters, so that even if a defect occurs in some masters.

The present invention provides a distributed synchronization method and apparatus capable of stable synchronization even when a defect occurs in some masters by synchronizing slaves with average values excluding maximum and minimum values of offset values calculated from sync frames transmitted from a plurality of time masters. do.

The effects of the present invention are not limited to those mentioned above, and other effects not mentioned can be clearly understood by those skilled in the art from the following description.

1 is a diagram illustrating a conventional IEEE 1588 synchronization technique.
2 is a flowchart schematically illustrating a synchronization method according to the present invention.
3 illustrates an embodiment of a process of calculating an average value by removing a maximum value and a minimum value of an offset value a plurality of times.
4 is a block diagram illustrating mutual conversion between a first mode using one time master and a second mode using a plurality of time masters according to the present invention.
5 is a block diagram illustrating an embodiment of one of the devices according to the present invention.
6 is a block diagram showing the configuration of a time master according to the present invention.
7 is a flow chart for a synchronization technique that is mounted and operated in a synchronization module of a time master.
8 is a block diagram showing the configuration of a time slave according to the present invention.
9 is a flowchart illustrating a synchronization scheme that is mounted and operated in a synchronization module of a time slave.

Hereinafter, a fault-tolerant distributed synchronization method and apparatus according to the present invention will be described in detail with reference to the accompanying drawings.

1 is a diagram illustrating a conventional IEEE 1588 synchronization process. In the present invention, since the operation itself in which synchronization is performed between the master and the slave is the same, it will be briefly described.

IEEE 1588 consists of a time master and a plurality of time slaves, which synchronize the plurality of slaves to the master clock using the IEEE 1588 algorithm.

Specifically, the time master periodically transmits a sync message to the time slave. The sync message is used to send the time master's time stamp t1 to the time slave.

At this time, due to the influence of the operating system inside the time master, a delay may occur until the time stamp t1 is transmitted as an actual network message. In order to reduce the delay time, the actual time stamp t1 is transmitted to the time slave through a follow-up message. The sync message is sent to store the time stamp t2 at the time of receipt in the time slave.

The time slave receiving the follow-up message transmits a delay request message to the time master to calculate a network transmission delay time that occurs when the sync message is transmitted.

The time master receiving the delay request message transmits the time stamp t4 at the time of receiving the delay request message to the time slave through the delay response message.

The time slave may calculate a clock offset (offset_from_master) between the time master and the time slave using the following equations. The time slave can synchronize its clock to the time master by adding a clock offset (offset_form_master) to its clock.

oneway_delay = ((master_to_slave_delay + (slave_to_master_delay)) / 2

master_to_slave_delay = t2-t1

slave_to_master_delay = t4-t3

offset_from_master = t2-t1-oneway_delay (final offset calculation)

2 schematically illustrates a fault-tolerant distributed synchronization method according to the present invention.

In the fault-tolerant distributed synchronization method according to the present invention, one of a plurality of time masters is designated as a starter node, and a sink time master, which is a time master other than the starter node among the plurality of time masters, is synchronized with the starter node. Step S2, a step S3 in which the sync frames of the starter node and the sync time master are periodically transmitted to the time slave and a step S4 synchronized with the plurality of time masters using the sync frame in which the time slave is transmitted.

The master according to the present invention uses a time master that performs only the operations for synchronization, not the entire master required for system control. This allows efficient synchronization without overloading the system, even with multiple time masters.

One of the main features of the present invention is the use of a plurality of time masters. That is, a plurality of time masters operate for synchronization, and a time slave is synchronized to a plurality of time masters.

Since multiple time masters are used, synchronization is required between time masters. To this end, in step S1, one of the plurality of time masters is designated as a starter node. The starter node designation may be determined by system or user selection.

As another example, since a problem occurs when a starter that starts synchronization starts to fail, it is desirable to check whether a plurality of time masters operate properly before synchronization is performed. This allows one of the working time masters to be designated as a starter node.

It is named "sink time master" to specify the remaining nodes except the starter node among the plurality of time masters. That is, the plurality of time masters are composed of starter nodes and one or more sink time master nodes.

The starter node transmits the sync frame to the sink master node so that the sync frame master is synchronized with the starter node (step S2). Synchronization between all time masters is a prerequisite for synchronizing time slaves.

In step S4, a time slave receives a plurality of sync frames, a step S4-1 in which a plurality of clock offset values are calculated from the plurality of sync frames, and a plurality of time slaves in the plurality of clock offset values. S4-3 steps are synchronized to the time master.

The step S4, which is the core configuration of the present invention and enables the most various embodiments, is a process in which a time slave is synchronized to a plurality of time masters. Here, the plurality of time masters includes a starter node and a sink time master.

The plurality of time masters (starter node and one or more sync time masters) transmit a sync frame to the time slave. The time slave receives a plurality of sync frames and calculates a plurality of clock offset values. The offset value is calculated through the conventional IEEE 1588 method described with reference to FIG. 1.

The time slave is synchronized to the plurality of time masters using the calculated plurality of clock offset values. Specifically, it looks at what values can be synchronized.

In one embodiment, the average value of the calculated plurality of clock offset values may be calculated and used as a reference for synchronization. In this case, if some of the time masters fail, the average value can be calculated to be somewhat wrong. However, when using a large number of time masters, even if some of the time masters fail, the average value can be calculated within a range acceptable to the system.

In another embodiment, the plurality of clock offset values calculated above are arranged in a table such as a memory, and the time slaves are synchronized based on an intermediate value of the table rather than calculating an average value. In this case, since the process of calculating the average value is omitted, synchronization can be performed more quickly.

Median selection can also be chosen in other well-known ways through data structure theory or algorithm theory.

As another embodiment, the plurality of clock offset values calculated first may be arranged in a table such as a memory, and the time slave may be synchronized based on the calculated average value except for the maximum value or the minimum value. In this case, it would be desirable to have at least three time masters.

If a defect occurs in the time master, the offset value calculated from the time master in which the defect occurs is a meaningful embodiment because it belongs to the maximum value or the minimum value region.

As another embodiment, the plurality of clock offset values calculated first may be arranged in a table such as a memory, and the time slave may be synchronized based on the calculated average value except for the maximum value and the minimum value.

As another embodiment, the plurality of clock offset values calculated first may be arranged in a table such as a memory, and the average value may be calculated except for a plurality of offsets of the maximum value and the minimum value region. That is, the time slave is synchronized based on the average value of one or more remaining offset values after the process of removing the maximum value and the minimum value a plurality of times.

There is a problem of how many times the maximum value and the minimum value are to be removed, but it is preferable to remove the number of times defined by the user in advance.

For example, if it is defined as removing a predetermined number of times as in the above table, it is possible to calculate the average value of the offset value as soon as possible without any operation.

As another example, the number of times of removal may be set to two or three times when the deviation between the previous offset average value and the later calculated offset value average value suddenly increases on the system.

Since the average can be calculated from the remaining values except the maximum and minimum values, there is a limit to the number of times the maximum and minimum values are removed. That is, the removal must be performed with a limit condition that there is at least one remaining clock offset value except for the maximum value and the minimum value.

3 illustrates an embodiment of a method of excluding a maximum value and a minimum value a plurality of times. That is, the maximum and minimum values are removed twice, and the time slave is synchronized based on the remaining average values.

A distributed synchronization method using a plurality of time masters, comprising: a first mode (similar to a conventional synchronization method) in which only one of the plurality of time masters is used as a master to synchronize a time slave by the IEEE 1588 synchronization method, or the aforementioned method The fault-tolerant distributed synchronization method according to the invention enables a method having a second mode in which a plurality of time masters synchronize time slaves. That is, in a certain case, while going to the first mode, a specific condition may occur or may be changed to the second mode by the selection of a system. It may also be possible to change from the second mode to the first mode.

4 illustrates this idea. That is, INIT refers to an initial state in which synchronization is not performed, Non-Fault Tolerance (NFT) refers to a conventional method that is not fault-tolerant, and FT (Fault Tolerance) refers to a fault-tolerant synchronization method according to the present invention.

Hereinafter, a fault-tolerant distributed synchronization apparatus according to the present invention will be described in detail. However, descriptions common to the fault-tolerant distributed synchronization method will be omitted and the focus will be on the core configuration of the apparatus.

5 schematically shows an embodiment of one of the devices according to the invention. 5 illustrates a network system 240 having two time masters 200 and 210 and two time slaves 230. The apparatus of the present invention is applicable to various topologies such as ring, line, star, and tree topologies as well as the bus topology shown in FIG.

The fault-tolerant distributed synchronization apparatus according to the present invention is networked with a plurality of time masters, a designated starter node among a plurality of time masters, a sink time master which is a time master other than a starter node among a plurality of time masters, and a plurality of time masters. It includes a time slave is synchronized with a plurality of sync frames received from the master.

The starter node of the present invention is preferably designated among the time masters that operate normally after checking whether the plurality of time masters operate normally.

As in the method of the present invention, the starter node transmits a sync frame to the sync time master to perform synchronization between the starter node and the sync time master.

The time slave receives a plurality of sync frames as described in the method of the present invention, calculates a plurality of offset values, and synchronizes with the plurality of time masters.

As shown in FIG. 6, the time master 300 specifically includes a master communication module 330 connected to a time slave and a time master through a network 240, and receives a clock from a clock source to store a current clock. The clock management module 320, a time stamp module 340 which extracts a time stamp from the clock value of the clock management module and is mounted in the communication module, and a master synchronization module 310 that controls the clock management module and the communication module to perform a synchronization algorithm. ).

7 is a flow chart for a synchronization technique that is mounted and operated in a synchronization module of a time master.

In order to perform synchronization, synchronization between masters must first be completed. For this purpose, one of the time masters is designated as the starter node. The starter node sends a sync frame to synchronize other time masters based on its clock, after which the other time masters send their own sync messages. The order for this is as follows:

The time master first checks whether it is set as a starter node, and if it is a starter node, transmits a sync frame and sets a timer to periodically transmit a sync frame.

The sink time master, not the starter node, waits for a sync frame from the starter node. When the sync frame is received from the starter node, it synchronizes itself with the starter, sets a timer, and periodically sends the sync frame.

As shown in FIG. 8, the time slave 400 is a slave communication module 430 that receives a sync frame related to synchronization through a network 240, and a slave that performs a synchronization algorithm using the sync frame received by the slave communication module. And a slave clock management module 420 for storing clock information according to a synchronization algorithm of the synchronization module 410 and the slave synchronization module.

9 is a flowchart illustrating a synchronization scheme that is mounted and operated in a synchronization module of a time slave.

The time slave first initializes the timer and waits to receive a sync frame. When the sync frame is received, the clock offset with the node transmitting the sync frame is calculated and stored in the offset table using IEEE 1588. At this time, when receiving a sync frame several times, the offset for each sync frame is calculated and stored in one table. When one or more sync frames are received, when the timer expires, the time slave uses a fault-tolerant median algorithm to align one or more time offsets stored in the table to calculate the average except for the highest and lowest values. Synchronize your clock with the calculation result.

The embodiments and drawings attached to this specification are merely to clearly show some of the technical ideas included in the present invention, and those skilled in the art can easily infer within the scope of the technical ideas included in the specification and drawings of the present invention. Modifications that can be made and specific embodiments will be apparent that all fall within the scope of the present invention.

200: starter node 210, 300: time master
220,400: time slave 240: network
310: master synchronization module 320: master clock management module
330: master communication module 340: time stamp module
410: slave synchronization module 420: slave clock management module
430: slave communication module

Claims (21)

Checking whether the plurality of time masters are operating normally;
A step S1, in which any one of the time masters operating normally is designated, and one of the time masters operating normally among the plurality of time masters is designated as a starter node;
A step S2 of synchronizing the sync time master, which is a time master other than the starter node among the plurality of time masters, to the starter node using a sync frame transmitted from the starter node;
Step S3, in which the sync frames of the starter node and the sync time master are periodically transmitted to a time slave; And
In step S4, wherein the time slave is synchronized with the plurality of time masters using the transmitted sync frame,
The step S4
Step S4-1 in which the time slave receives a plurality of sync frames;
Step S4-2 of calculating a plurality of clock offset values from the plurality of sync frames; And
And a step S4-3 in which the time slave is synchronized to the plurality of time masters by using the calculated average value of the plurality of clock offset values. delete delete delete delete The method of claim 1,
In step S4-3, a fault-tolerant distributed synchronization method for synchronizing the time slave based on an average value calculated by excluding a maximum value or a minimum value among the plurality of clock offset values calculated in step S4-2. . The method of claim 1,
In the step S4-3, a fault-tolerant distributed synchronization method for synchronizing the time slave based on the calculated average value except for the maximum value and the minimum value among the plurality of clock offset values calculated in the step S4-2. . The method of claim 1,
In step S4-3, the step of excluding a maximum value and a minimum value among the plurality of clock offset values calculated in step S4-2 is performed a plurality of times, and the time slave is synchronized based on an average value of the remaining offset values. Fault-tolerant distributed synchronization method. 9. The method of claim 8,
Excluding the maximum value and the minimum value that is performed a plurality of times
A fault-tolerant distributed synchronization method, characterized in that a predetermined number of times of input is performed, with a limit condition that at least one remaining clock offset value is excluded except a maximum value and a minimum value. A first mode of synchronizing a time slave by an IEEE 1588 synchronization method using only one of the plurality of time masters as a master; or
10. A distributed synchronization method according to any one of claims 1 or 6 to 9, comprising a second mode in which a plurality of time masters synchronize time slaves.
And a mode change is possible between the first mode and the second mode. A starter node designated among a plurality of time masters that are normally operated after checking whether the plurality of time masters are normally operated;
A sync time master that is a time master other than the starter node among the plurality of time masters; And
A plurality of time frames connected to the time master and a network, receiving a plurality of sync frames from the plurality of time masters, calculating a plurality of clock offset values from the transmitted plurality of sync frames, and using the average value of the calculated offset values. Including time slaves that are synchronized
The starter node,
And a sync frame transmitted to the sync time master to perform synchronization between the starter node and the sync time master. delete delete delete delete The method of claim 11,
The time slave is a fault-tolerant distributed synchronization apparatus, characterized in that the synchronization based on the average value of the remaining offset value except the maximum value or the minimum value of the plurality of clock offset values. The method of claim 11,
The time slave is a fault-tolerant distributed synchronization apparatus, characterized in that the synchronization based on the average value of the remaining offset value except the maximum value and the minimum value of the plurality of clock offset values. The method of claim 11,
And the time slave performs a plurality of steps of excluding a maximum value and a minimum value among the calculated plurality of clock offset values and synchronizes based on an average value of the remaining offset values. 19. The method of claim 18,
Excluding the maximum value and the minimum value that is performed a plurality of times
A fault-tolerant distributed synchronization device, characterized in that a predetermined number of times of input is performed, with a limit condition that at least one remaining clock offset value is excluded except a maximum value and a minimum value. The method of claim 11,
The time master
A master communication module coupled to the time slave and the time master via a network;
A master clock management module which receives a clock from a clock source and stores a current clock;
A time stamp module extracting a time stamp from a clock value of the clock management module and mounting the time stamp on the communication module; And
And a master synchronization module for controlling the clock management module and the communication module to perform a synchronization algorithm. The method of claim 11,
The time slave is
A slave communication module to receive synchronization related sync frames over a network;
A slave synchronization module that performs a synchronization algorithm using the sync frame received by the slave communication module; And
A slave clock management module for storing clock information according to a synchronization algorithm of the slave synchronization module;
The fault-tolerant distributed synchronization device comprising a.

Images