JP7170957B2 - Distributed processing system, distributed processing method, distributed processing program and scheduling device - Google Patents

Description

The present invention relates to a distributed processing system that distributes and processes a plurality of tasks by a plurality of processors.

A distributed processing system in which a plurality of devices are connected by a network has been studied.
A distributed processing system is intended to execute many tasks in a state in which a plurality of devices are synchronized, or to synchronize and control a plurality of devices over a wide range.

In the distributed processing system, each device shares information on the processing results of tasks processed with each other, and executes processing of further tasks.
In a distributed processing system, periodic processing may be performed. In that case, if the period cannot be observed, the process to be executed next becomes unclear, and equipment control and task processing synchronization cannot be performed. Then, a serious error occurs, such as the distributed processing system stopping.
It is possible to realize a distributed processing system that performs periodic processing by reducing unnecessary waiting time for tasks or communication in order to observe the cycle, and by performing communication and processing with low delay and punctuality. Become.

A distributed processing system does not have to be a system in which multiple devices are connected via a network.
For example, the distributed processing system may be a system that uses bus or memory sharing without using a network, or a system that uses both bus and memory sharing. By using a bus or shared memory, information can be exchanged with very little delay.

Patent Literature 1 discloses a system that performs distributed processing using a plurality of CPU cores.
This system allocates task time (scheduling) in consideration of the increase in processing delay due to an increase in the load on each CPU core and the increase in processing waiting time due to an increase in communication time, and improves the efficiency of processing. come true.
CPU is an abbreviation for Central Processing Unit.

JP 2010-079622 A

The conventional method calculates the processing delay time when executing a single process and the processing delay time used for time allocation based on the CPU load. Further, in the conventional method, time allocation is performed using communication time when communication is actually performed.
However, in practice, various tasks and communications are performed. Each task and each communication is executed immediately if it can be executed when the request occurs. And regardless of whether each task is a regular task or a sudden task, and regardless of whether each communication is a regular communication or a sudden communication, a high-priority Tasks and communications are given priority, and each task and each communication may not be completed at the expected timing. As a result, a task and communication whose execution is delayed cause a delay in the next executed task and communication in a chain.

Only the content of the task and the priority of the task are input to each device. Task execution is performed by functions such as the OS of each device. At that time, the execution timing (execution time) of each task and each communication is determined by the OS in real time according to criteria such as priority in order from the executable task or communication. In other words, it is not possible to finely control the execution timing of each task and each communication.
In the current distributed processing system, each task and each communication are executed on a case-by-case basis, so detailed timings of each task and each communication cannot be calculated.
OS is an abbreviation for Operating System.

SUMMARY OF THE INVENTION An object of the present invention is to enable each task and each communication to be executed in a planned manner without executing each task and each communication ad hoc.

A distributed processing system according to the present invention is a system in which a plurality of tasks are distributedly processed by a plurality of processors.
The distributed processing system comprises:
the execution order of the plurality of tasks, the amount of processing required to execute each task, the amount of communication required for communication for giving the execution result of each task to the next task, the priority of each task, and the priority of each communication an information acquisition unit that acquires task information indicating a degree and an assignment destination that identifies a processor to which each task is assigned;
Designating the execution time assigned to each task and the execution time assigned to each communication based on the execution order of the plurality of tasks, the processing amount of each task, the communication amount of each communication, and the assignment destination of each task a tentative schedule determining unit that determines a schedule;
Based on the priority of each task and the priority of each communication, by adjusting the execution time of each task specified in the tentative schedule and the execution time of each communication specified in the tentative schedule, and a conflict resolution unit that generates a schedule in which conflicts between communications and conflicts between communications are resolved.

According to the present invention, it is possible to systematically execute each task and each communication without executing each task and each communication ad hoc.

1 is a configuration diagram of a distributed processing system 100 according to Embodiment 1. FIG. 1 is a configuration diagram of a scheduling device 200 according to Embodiment 1. FIG. 1 is a configuration diagram of a distributed processing device 300 according to Embodiment 1. FIG. 4 is a flowchart of a distributed processing method according to Embodiment 1; 4 shows task information 111 according to Embodiment 1. FIG. 4 shows a directed graph 112 according to the first embodiment; FIG. 4 shows a topology graph 113 according to the first embodiment; FIG. 4 shows the tentative schedule graph 114 according to the first embodiment; FIG. 4 is a flowchart of conflict resolution processing (S200) according to the first embodiment; 4 is a flowchart of conflict resolution processing (S200) according to the first embodiment; 4 is a flowchart of task adjustment processing (S300) according to the first embodiment; 4 is a flowchart of task adjustment processing (S300) according to the first embodiment; 4 is a flowchart of communication adjustment processing (S400) according to the first embodiment; 4 is a flowchart of communication adjustment processing (S400) according to the first embodiment; 4 is a flowchart of communication adjustment processing (S400) according to the first embodiment; FIG. 2 is a configuration diagram of a distributed processing system 100X according to Embodiment 2; FIG. 3 is a configuration diagram of a distributed processing device 300X according to Embodiment 2; 8 is a flowchart of a distributed processing method according to Embodiment 2; FIG. 4 is a configuration diagram of a distributed processing device 300 according to Embodiment 3; 1 is a hardware configuration diagram of a scheduling device 200 according to an embodiment; FIG. 3 is a hardware configuration diagram of the distributed processing device 300 according to the embodiment; FIG.

The same or corresponding elements are denoted by the same reference numerals in the embodiments and drawings. Descriptions of elements having the same reference numerals as those described will be omitted or simplified as appropriate. Arrows in the figure mainly indicate the flow of data or the flow of processing.

Embodiment 1.
The distributed processing system 100 will be described with reference to FIGS. 1 to 15. FIG.

*** Configuration description ***
The configuration of the distributed processing system 100 will be described based on FIG.
A distributed processing system 100 is a system in which a plurality of tasks are distributedly processed by a plurality of processors.

In the distributed processing system 100 , a device including each processor is called a distributed processing device 300 .

A distributed processing system 100 includes a plurality of distributed processing devices 300 .
Each distributed processing device 300 has a synchronization function.
A plurality of distributed processing apparatuses 300 are synchronized with each other by their respective synchronization functions.

In FIG. 1, the distributed processing system 100 comprises three distributed processing devices (300A-300C).
However, the distributed processing system 100 may include two distributed processing apparatuses 300 or four or more distributed processing apparatuses 300 .

Each of the plurality of devices controlled by the plurality of distributed processing apparatuses 300 is called a target device 101 .
A specific example of the target device 101 is a robot arm, servo or motor.

A distributed processing system 100 includes a plurality of target devices 101 .
In FIG. 1, the distributed processing system 100 comprises three target devices (101A-101C).
Each target device is controlled by one or more distributed processing units 300 . For example, there are multiple tasks Ax for controlling the target device 101A. A plurality of distributed processing apparatuses 300 cooperate to execute a plurality of tasks Ax.
However, the distributed processing system 100 may include two target devices 101 or four or more target devices 101 .

The distributed processing system 100 comprises a transfer device 102 .
The transfer device 102 transfers data between the distributed processing devices 300 . Examples of forwarding device 102 are routers, switches, bridges or hubs.

A plurality of distributed processing apparatuses 300 communicate with each other via the transfer apparatus 102 .
Specifically, communication is performed to give the execution result of each task to the next task.
However, the plurality of distributed processing apparatuses 300 may communicate with each other without going through the transfer apparatus 102 . In that case, the transfer device 102 is unnecessary, and the distributed processing devices 300 are connected to each other.

The distributed processing system 100 has a scheduling device 200 .
The scheduling device 200 performs scheduling for distributed processing of multiple tasks by multiple processors. Scheduling means allocating time for each task and each communication.
That is, the scheduling device 200 generates a schedule in which the execution time of each task and the execution time of each communication are specified.

Each distributed processing unit 300 executes each assigned task at the execution time of each task specified in the schedule, and communicates the execution result of each assigned task at the execution time of each communication specified in the schedule. do.

In FIG. 1, the scheduling device 200 is connected to the distributed processing device 300A.
However, the scheduling device 200 may be connected to the transfer device 102 .
Also, the scheduling device 200 may have the function of the transfer device 102 . In that case, the transfer device 102 is unnecessary, and the scheduling device 200 is connected to the multiple distributed processing devices 300 instead of the transfer device 102 .

The network configuration of distributed processing system 100 may be any network configuration.
The network of distributed processing system 100 may be either a wired network or a wireless network.

The configuration of the scheduling device 200 will be described based on FIG.
The scheduling device 200 is a computer comprising hardware such as a processor 201 , a memory 202 , an auxiliary storage device 203 and an external interface 204 . These pieces of hardware are connected to each other via signal lines.

A processor 201 is an IC that performs arithmetic processing and controls other hardware. For example, processor 201 is a CPU or system LSI.
IC is an abbreviation for Integrated Circuit.
CPU is an abbreviation for Central Processing Unit.
LSI is an abbreviation for Large Scale Integration.

Memory 202 is a volatile or non-volatile storage device. Memory 202 is also referred to as main storage or main memory. For example, memory 202 is RAM. The data stored in memory 202 is saved in auxiliary storage device 203 as needed.
RAM is an abbreviation for Random Access Memory.

Auxiliary storage device 203 is a non-volatile storage device. For example, the auxiliary storage device 203 is ROM, HDD, SSD or memory card. Data stored in the auxiliary storage device 203 is loaded into the memory 202 as required.
ROM is an abbreviation for Read Only Memory.
HDD is an abbreviation for Hard Disk Drive.
SSD is an abbreviation for Solid State Drive.

The external interface 204 is an interface for communicating with, or inputting/outputting from, an external device (for example, the distributed processing device 300A). That is, the external interface 204 serves as both a communication interface and an input/output interface. For example, the external interfaces 204 are Ethernet (registered trademark), USB, and UART ports.
USB is an abbreviation for Universal Serial Bus.
UART is an abbreviation for Universal Asynchronous Receiver Transmitter.

The scheduling device 200 includes elements such as a scheduling section 210 and a schedule output section 220 .
The scheduling unit 210 includes elements such as an information acquisition unit 211 , a tentative schedule determination unit 212 and a conflict resolution unit 213 .
These elements are implemented in software.

Auxiliary storage device 203 stores a scheduling program for causing the computer to function as scheduling section 210 and schedule output section 220 . The scheduling program is loaded into memory 202 and executed by processor 201 .
The auxiliary storage device 203 further stores an OS. At least part of the OS is loaded into memory 202 and executed by processor 201 .
The processor 201 executes the scheduling program while executing the OS.
OS is an abbreviation for Operating System.

Input/output data of the scheduling program is stored in the storage unit 290 .
Memory 202 functions as storage unit 290 . However, a storage device such as the auxiliary storage device 203 , a register within the processor 201 and a cache memory within the processor 201 may function as the storage unit 290 instead of or together with the memory 202 .

The scheduling device 200 may include multiple processors that substitute for the processor 201 . Multiple processors share the functions of processor 201 .

The scheduling program can be recorded (stored) in a non-volatile recording medium such as an optical disc or flash memory in a computer-readable manner.

The configuration of the distributed processing device 300 will be described based on FIG.
The distributed processing device 300 is a computer comprising hardware such as a processor 301 , a memory 302 , an auxiliary storage device 303 and an external interface 304 . These pieces of hardware are connected to each other via signal lines.

A processor 301 is an IC that performs arithmetic processing and controls other hardware. For example, processor 301 is a CPU or system LSI.
Memory 302 is a volatile or non-volatile storage device. Memory 302 is also referred to as main storage or main memory. For example, memory 302 is RAM. The data stored in the memory 302 is saved in the auxiliary storage device 303 as required.
Auxiliary storage device 303 is a non-volatile storage device. For example, the auxiliary storage device 303 is ROM, HDD, SSD or memory card. Data stored in the auxiliary storage device 303 is loaded into the memory 302 as required.
The external interface 304 is an interface for communicating with, or inputting/outputting from, an external device (for example, the target device 101 and the transfer device 102). That is, the external interface 304 serves as both a communication interface and an input/output interface. For example, the external interfaces 304 are Ethernet (registered trademark), USB, and UART ports.

The distributed processing device 300 includes elements such as an execution control unit 311 , a task execution unit 312 , a communication execution unit 313 , a device control unit 314 and a schedule reception unit 319 . These elements are implemented in software.

The auxiliary storage device 303 stores a distributed processing program for causing the computer to function as an execution control unit 311 , a task execution unit 312 , a communication execution unit 313 , a device control unit 314 and a schedule reception unit 319 . The distributed processing program is loaded into memory 302 and executed by processor 301 .
The auxiliary storage device 303 further stores an OS. At least part of the OS is loaded into memory 302 and executed by processor 301 .
The processor 301 executes the distributed processing program while executing the OS.

Input/output data of the distributed processing program is stored in the storage unit 390 . For example, the storage unit 390 stores a schedule 391 and the like.
Memory 302 functions as storage unit 390 . However, a storage device such as the auxiliary storage device 303 , a register within the processor 301 and a cache memory within the processor 301 may function as the storage unit 390 instead of or together with the memory 302 .

The distributed processing device 300 may include multiple processors in place of the processor 301 . Multiple processors share the functions of processor 301 .

The distributed processing program can be computer-readable and recorded (stored) in a non-volatile recording medium such as an optical disc or flash memory.

***Description of operation***
The operating procedure of the distributed processing system 100 corresponds to the distributed processing method. Further, the procedure of operation of the distributed processing system 100 corresponds to the procedure of processing by the distributed processing program. The scheduling program for scheduling device 200 is part of the distributed processing program for distributed processing system 100 .

The distributed processing method will be described based on FIG.
In step S110, the information acquisition unit 211 of the scheduling device 200 acquires system information via the external interface 204 or a recording medium.
System information is data that includes configuration information and task information.
The configuration information indicates the configuration of the distributed processing system 100. FIG.
The task information indicates multiple tasks to be executed by the distributed processing system 100 .

The task information 111 will be described based on FIG. Task information 111 is a specific example of task information.
The task information 111 indicates information for each of five tasks (A, B, C, D, A').
Specifically, the task information 111 indicates task name, previous task, next task, processing amount, communication amount, task priority, communication priority, and assignment destination.
A task name identifies a task.
A previous task identifies the task to be executed immediately before.
Next task identifies the next task to be executed.
The amount of processing is the amount of processing required to execute the task. The length of task execution time is determined according to the amount of processing. For example, the amount of processing can be represented by the length of processing time or the number of steps. The processing time length is the length of time required to execute the task under the standard execution conditions. For example, the execution condition is the condition of the processor 301 or the like.
The amount of communication is the amount of communication for giving the task execution result to the next task. For example, the amount of communication is represented by communication time length or data size. The communication time length is the length of time required for communication of execution results under standard communication conditions. For example, the communication conditions are conditions such as communication speed. Depending on the configuration of the distributed processing device 300 and how each task is assigned, communication of execution results may not be necessary. For example, when the task and the next task are executed by the same distributed processing device 300, the execution result of the task is shared among the distributed processing devices 300, so the amount of communication is extremely small. In other words, the amount of communication is so small that it can be ignored.
A task priority is the priority of a task.
Communication priority is the priority of communication for giving the execution result of a task to the next task.
Assignee identifies the processor 301 to which the task is assigned. That is, the assignee identifies the processor 301 that executes the task. For example, device A identifies processor 301 of distributed processing device 300A.

The task name, previous task and next task indicate dependencies between tasks.
Dependencies between tasks determine the execution order of multiple tasks.

Returning to FIG. 4, the description continues from step S120.
In step S120, the provisional schedule determination unit 212 of the scheduling device 200 determines a provisional schedule based on the system information.

The tentative schedule specifies the execution time assigned to each task and the execution time assigned to each communication.

A tentative schedule is determined without considering conflicts between tasks and communications.
Conflict between tasks means that at least part of the execution time overlaps between tasks assigned to the same distributed processing device 300 . Assume that the number of tasks that can be executed by one distributed processing device 300 at one time is one. In other words, each distributed processing device 300 cannot execute two or more tasks at the same timing.
Contention between communications means that at least part of the execution time overlaps between communications carried out using the same channel. Assume that each distributed processing device 300 performs full-duplex communication using a set of two communication paths. It is also assumed that the number of execution results that can be communicated at one time through each communication channel is one. In other words, each communication path cannot communicate two or more execution results at the same timing.

The provisional schedule for each distributed processing device 300 is determined as follows.
First, the provisional schedule determining unit 212 refers to the task information, and determines the execution order of the plurality of tasks based on the task name column, the previous task column, and the next task column.
The provisional schedule determination unit 212 also refers to the task information and determines the length of execution time of each task based on the processing amount column.
The provisional schedule determination unit 212 also refers to the task information and determines the length of execution time of each communication based on the column of traffic.
The provisional schedule determination unit 212 also refers to the task information and determines the distributed processing device 300 to which each task is assigned based on the assignment destination column.
Also, the provisional schedule determination unit 212 determines whether the assignment destination of each task is different from the assignment destination of the next task based on the assignment destination column of each task and the assignment destination column of the next task. If the allocation destination of each task is different from the allocation destination of the next task, the provisional schedule determination unit 212 allocates communication processing. If the assignee of each task is the same as the assignee of the next task, the provisional schedule determining unit 212 does not assign the communication process. This is because the assignment destination of the task is the same, so communication for sharing data for executing the task by communication is unnecessary.
Then, the provisional schedule determination unit 212 determines each distributed processing device based on the order of execution of the plurality of tasks, the length of execution time of each task, the length of execution time of each communication, and the assignment destination of each task. Execution time of each task in 300 and execution time of each communication in each distributed processing device 300 are determined. The provisional schedule determination unit 212 generates data indicating each determined information. For example, the provisional schedule determination unit 212 generates data corresponding to the provisional schedule graph 114 of FIG.

The directed graph 112 is shown in FIG. The directed graph 112 is a directed graph generated based on the task information 111 (see FIG. 5).
The directed graph 112 represents the following dependencies.
After task A is executed, task B, task C, and task D are executed based on the execution result of task A. FIG.
After execution of task B, task C and task D, task A' is executed based on the execution result of task B, the execution result of task C, and the execution result of task D. FIG.

The topology graph 113 is shown in FIG. The topology graph 113 is generated based on configuration information acquired together with the task information 111 .
The topology graph 113 represents the following configuration.
Port 1 of the distributed processing device 300 A is connected to port 2 of the transfer device 102 .
Port 4 of distributed processing device 300B is connected to port 3 of transfer device 102 .
Port 6 of the distributed processing device 300C is connected to port 5 of the transfer device 102 .

The tentative schedule graph 114 is shown in FIG. The provisional schedule graph 114 is generated based on the task information 111 (see FIG. 5) and configuration information (see FIG. 7) acquired together with the task information 111. FIG.
"CPUn" means the processor 301 of the distributed processing device 300n.
"Port xy" means a communication path from port x to port y.
A “switch” means a forwarding device 102 .
Circles marked with 'A', 'B', 'C', 'D' or 'A'' represent tasks.
"Tn" represents the communication n of the execution result.
"Sn" represents task n for transfer.
The provisional schedule graph 114 represents, for example, the following provisional schedule.
Task A is assigned an execution time of "0 to 10".
The communication T5 is assigned an execution time of "10-15".
An execution time of "10" is assigned to task S3. Assume that the time lag due to task Sn is zero. Transfer delays and buffering delays may be added depending on the actual processing.
The communication T6 is assigned an execution time of "10+α to 15+α". “α” means communication time for one frame.
Task D is assigned an execution time of "10 to 20".

Returning to FIG. 4, the description continues from step S200.
In step S200, the conflict resolution unit 213 of the scheduling device 200 executes conflict resolution processing based on the provisional schedule and system information.
Conflict resolution processing adjusts the execution time of each task specified in the provisional schedule and the execution time of each communication specified in the provisional schedule, thereby eliminating conflicts between tasks and communications. 391 is generated.
Details of the conflict resolution process (S200) will be described later.

In step S<b>130 , the schedule 391 is set for each distributed processing device 300 .
For example, the schedule output unit 220 of the scheduling device 200 transmits the schedule 391 to each distributed processing device 300 via the external interface 204 . The schedule reception unit 319 of each distributed processing device 300 receives the schedule 391 and stores the schedule 391 in the storage unit 390 . In this case, the schedule output unit 220 transmits the schedule 391 to the distributed processing device 300A. The distributed processing device 300A then transmits the schedule 391 to each of the distributed processing devices 300B and 300C via the transfer device 102 .
For example, the schedule output unit 220 of the scheduling device 200 writes the schedule 391 to a recording medium. A user inputs a schedule 391 to each distributed processing device 300 using a recording medium. The schedule receiving unit 319 of each distributed processing device 300 receives the schedule 391 and stores the schedule 391 in the storage unit 390 .

In step S<b>140 , the processor 301 of each distributed processing device 300 executes each assigned task at the execution time of each task specified in the schedule 391 .
Furthermore, the processor 301 of each distributed processing device 300 communicates the execution result of each assigned task at the execution time of each communication specified in the schedule 391 .

Specifically, the execution control unit 311 operates the task execution unit 312 and the communication execution unit 313 according to the schedule 391 .
The task execution unit 312 executes each assigned task.
The communication execution unit 313 communicates execution results of each assigned task.
The device control unit 314 controls the target device 101 according to the command from the task when the task for controlling the target device 101 is executed.

The execution control unit 311 operates the task execution unit 312 as follows.
The execution control unit 311 determines at each time whether the processor 301 is in an idle state.
When the processor 301 is in an idle state, the execution control unit 311 refers to the schedule 391 and determines whether or not there is a task that has not been executed. A task that has not been executed is called an unexecuted task.
If there is one or more unexecuted tasks, the execution control unit 311 refers to the schedule 391 and determines whether there is an unexecuted task assigned a start time within a specified period of time from the current time. An unexecuted task assigned a start time within a specified period of time from the current time is called an execution target task.
If there is an execution target task, the execution control unit 311 causes the task execution unit 312 to execute the execution target task.

The execution control unit 311 operates the communication execution unit 313 as follows.
The execution control unit 311 determines whether the communication channel is available at each time.
If the communication path is available, the execution control unit 311 refers to the schedule 391 and determines whether or not there is communication that has not been executed. Communication that has not been executed is referred to as unexecuted communication.
If there is one or more unexecuted communications, the execution control unit 311 refers to the schedule 391 and determines whether there is an unexecuted communication assigned a start time within a specified period of time from the current time. An unexecuted communication assigned a start time within a specified period of time from the current time is referred to as an execution target communication.
If there is communication to be executed, the execution control unit 311 causes the communication executing unit 313 to execute the communication to be executed.

The flow of the conflict resolution process (S200) will be described with reference to FIGS. 9 and 10. FIG.
In step S201, the conflict resolution unit 213 sets an initial value of priority to be processed. A priority to be processed is referred to as a target priority.
Specifically, the conflict resolution unit 213 sets the highest priority as the target priority. As a result, schedule assignment is determined in descending order of priority for tasks and communications.

In step S202, the conflict resolution unit 213 determines whether or not there is an undetermined task and an undetermined communication.
An undetermined task is a task whose execution time is not determined.
Undetermined communication is communication whose execution time is not determined.
Execution time is specified by a start time and an end time.

If there is at least one of the undetermined task and the undetermined communication, the process proceeds to step S203.
If there are no pending tasks or pending communications, the process ends.

In step S203, the conflict resolution unit 213 determines whether or not there are coordination candidate tasks and coordination candidate communications.

An adjustment candidate task is an undetermined task that has the same task priority as the target priority, is in a ready state, and has a pending flag indicating non-suspended (0).
The executable state is a state in which there is no previous task or the respective execution times of the previous task and previous communication are fixed.
The pending flag indicates pending (1) or not pending (0). The initial value of the pending flag is non-pending (0).

An adjustment candidate communication is an unconfirmed communication that has the same communication priority as the target priority, is in an executable state, and has a hold flag indicating non-hold (0).
The executable state is a state in which task execution time is fixed.

If there is at least one of the adjustment candidate task and the adjustment candidate communication, the process proceeds to step S205.
If there is no adjustment candidate task or adjustment candidate communication, the process proceeds to step S204.

In step S204, the conflict resolution unit 213 lowers the target priority by one.
After step S204, the process proceeds to step S202.

In step S205, the conflict resolution unit 213 selects the task or communication with the earliest execution order from among the coordination candidate tasks and coordination candidate communications. A selected task or selected communication is referred to as an adjustment target. In the policy that schedule allocation is determined in order of priority tasks and communications, tasks and communications having priorities equal to or higher than the priority to be adjusted are compared with those to be adjusted in order of earliest start time. That is, low priority tasks and communications are assigned avoiding high priority tasks and communications.
If the adjustment target is a task, the process proceeds to step S300.
If the adjustment target is communication, the process proceeds to step S400.

In step S300, the conflict resolution unit 213 executes task adjustment processing.
Task adjustment processing adjusts the execution time of a task with a lower priority when at least part of the execution time overlaps between tasks assigned to the same processor 301 .
Details of the task adjustment processing (S300) will be described later.

In step S400, the conflict resolution unit 213 executes communication adjustment processing.
The communication adjustment process adjusts the execution time of the communication with the lower priority when at least part of the execution time overlaps between the communications executed using the same communication channel.
Details of the communication adjustment process (S400) will be described later.

After step S300 or step S400, the process proceeds to step S211.

In step S211, the conflict resolution unit 213 determines the suspension flag to be adjusted.
If the suspension flag to be adjusted indicates non-suspension (0), the process proceeds to step S212.
If the suspension flag to be adjusted indicates suspension (1), the process proceeds to step S202.

In step S212, the conflict resolution unit 213 determines each update target, adjusts the execution time of each update target, and updates the state of each update target.

Describe the update target.
When the task adjustment process (S300) is executed, the update targets are each task executed after execution of the task whose execution time is adjusted and each communication executed after execution of the task whose execution time is adjusted. .
When the communication adjustment process (S400) is executed, the update targets are each task executed after execution of the communication whose execution time is adjusted and each communication executed after execution of the task whose execution time is adjusted. .

The execution time of each update target is adjusted as follows.
When the execution time to be adjusted is changed, the conflict resolution unit 213 shifts the execution time of each update target based on the execution time after update.

The state of each update target is updated as follows.
If the execution time to be adjusted has not been determined, the conflict resolution unit 213 sets each update target to an unexecutable state.

In step S213, the conflict resolution unit 213 sets the pending flag of each undetermined task and the pending flag of each undetermined communication to non-suspended (0).
After step S213, the process proceeds to step S201.

The flow of the task adjustment process (S300) will be described with reference to FIGS. 11 and 12. FIG.
A task to be adjusted is called a target task.

In step S301, the conflict resolution unit 213 determines whether or not there is a common task with the same priority or higher.
A common task having the same priority or higher is a common task having a task priority higher than the task priority of the target task.
A common task is a task that shares the same destination as the target task.
If there is a common task with the same priority or higher, the process proceeds to step S302.
If there is no common task with the same priority or higher, the process proceeds to step S313.

In step S302, the conflict resolution unit 213 determines whether or not there is a conflict task.
A competing task is a high priority common task that overlaps at least part of its execution time with the target task.
If there is a competing task, the process proceeds to step S303.
If there is no competing task, the process proceeds to step S311.

In step S303, the conflict resolution unit 213 adjusts the execution time of the target task.

The execution time of the target task is adjusted as follows.
When the execution order of the target task is before the execution order of the competing tasks, the conflict resolution unit 213 changes the execution time of the target task to the time before the execution time of the competing tasks.
If the execution order of the target task is after the execution order of the competing tasks, the conflict resolution unit 213 changes the execution time of the target task to a time after the execution time of the competing tasks.

The pending flag value of the target task remains non-suspended (0).
After step S303, the process ends.

In step S311, the conflict resolution unit 213 determines whether or not there is a potential task.
A potential task is a task that may conflict with the target task. Specifically, a potential task is a high-priority common task with an uncertain execution time.
If there is a potential task, the process proceeds to step S312.
If there is no possible task, the process proceeds to step S313.

In step S312, the conflict resolution unit 213 sets the pending flag of the target task to pending (1).
After step S312, the process ends.

In step S313, the conflict resolution unit 213 determines the execution time of the target task.
The pending flag value of the target task remains non-suspended (0).
After step S313, the process ends.

The communication adjustment process (S400) will be described with reference to FIGS. 13, 14 and 15. FIG.
A communication to be coordinated is referred to as a target communication.

In step S401, the conflict resolution unit 213 determines whether there is common communication with the same priority or higher.
A common communication having the same priority or higher is a common communication having a communication priority higher than the communication priority of the target communication.
Common communication is communication that shares a communication path with the target communication.
If there is common communication with the same priority or higher, the process proceeds to step S402.
If there is no common communication with the same priority or higher, the process proceeds to step S404.

In step S402, the contention resolution unit 213 determines whether or not there is contention communication.
A competing communication is a high priority common communication that overlaps at least a portion of its execution time with the target communication.
If there is contention communication, the process proceeds to step S403.
If there is no competing communication, the process proceeds to step S411.

In step S403, the conflict resolution unit 213 adjusts the execution time of the target communication.

The execution time of the target communication is adjusted as follows.
When the execution order of the target communication is before the execution order of the contention communication, the contention resolution unit 213 changes the execution time of the target communication to the time before the execution time of the contention communication.
If the execution order of the target communication is after the execution order of the contention communication, the contention resolution unit 213 changes the execution time of the target communication to a time after the execution time of the contention communication.

The value of the hold flag of the target communication remains non-hold (0).
After step S403, the process ends.

In step S404, the conflict resolution unit 213 determines the execution time of the target communication.
The value of the hold flag of the target communication remains non-hold (0).
After step S404, the process ends.

In step S411, the contention resolution unit 213 determines whether or not there is possible communication.
A potential communication is a communication that may conflict with the target communication. Specifically, the potential communication is a high-priority common communication whose execution time is not fixed.
If there is possible communication, the process proceeds to step S412.
If there is no possibility communication, the process proceeds to step S413.

In step S412, the contention resolution unit 213 sets the hold flag of the target communication to hold (1).
After step S412, the process ends.

In step S413, the conflict resolution unit 213 determines whether mixed communication and divided communication are applicable.
In other words, the contention resolution unit 213 confirms whether preparations have been made to implement mixed communication and split communication determinations present in the transition step. Specifically, contention resolution section 213 determines whether the start time of each contention communication that is the target of mixed communication and divided communication is fixed.
Mixed communication is communication mixed with target communication. Specifically, mixed communication is communication to which an execution time that partially overlaps with the execution time of the target communication is assigned.
A split communication is a potential communication that requires splitting of the target communication. Specifically, division communication is a possibility communication to which an execution time within the execution time of the target communication is assigned.
If mixed communication and split communication can be applied, the process proceeds to step S414.
If mixed communication and split communication cannot be applied, the process proceeds to step S421.

In step S414, the contention resolution unit 213 determines the start time of the target communication.
If the start time of the target communication has been determined, the process proceeds to step S415.
If the start time of the target communication has not been determined, the process proceeds to step S416.

In step S415, the contention resolution unit 213 sets the hold flag of the target communication to hold (1).
After step S415, the process ends.

In step S416, the contention resolution unit 213 determines the start time of the target communication.
The value of the hold flag of the target communication remains non-hold (0).
After step S416, the process ends.

In step S421, the contention resolution unit 213 determines the presence or absence of mixed communication and divided communication.
If there are both mixed communication and split communication, the process proceeds to step S423.
If at least one of the mixed communication and the divided communication does not exist, the process proceeds to step S422.

In step S422, the conflict resolution unit 213 determines the execution time of the target communication.
The value of the hold flag of the target communication remains non-hold (0).
After step S422, the process ends.

In step S423, the contention resolution unit 213 determines which of mixed communication and divided communication has an earlier start time. The communication with the earlier start time is referred to as the earlier communication.
If the previous communication is mixed communication, the process proceeds to step S424.
If the previous communication is divided communication, the process proceeds to step S425.

In step S424, the contention resolution unit 213 treats the target communication and the mixed communication as one communication. That is, the contention resolution unit 213 generates communication in which target communication and mixed communication are mixed.
The value of the hold flag of the target communication remains non-hold (0).
After step S424, the process ends.

In step S425, the contention resolution unit 213 divides the target communication into the first half communication and the second half communication.
The first half of the communication is the communication before the execution time of the divided communication.
The second half of the communication is the communication remaining after removing the first half of the communication from the target communication.

The conflict resolution unit 213 then determines the execution time of the first half of the target communication.
The value of the hold flag of the target communication remains non-hold (0).
After step S425, the process ends.

*** Effect of Embodiment 1 ***
According to Embodiment 1, each task and each communication can be executed systematically without executing each task and each communication ad hoc.

***Description of Example 1***
The scheduling device 200 may generate the schedule 391 in consideration of the communication time of priority frames.
A priority frame is a frame prioritized over the communication of execution results of each task. Specific priority frames are emergency frames and scheduled frames. Urgent frames are frames that should be communicated as soon as possible. A fixed frame is a frame whose communication timing is determined.
Specifically, the provisional schedule determination unit 212 reserves communication time for priority frames, and allocates time other than the reserved communication time to each task and each communication. For example, by giving the priority frame the highest priority in the task information 111, communication time for the priority frame is secured.

***Description of Example 2***
In step S110 (see FIG. 4), the information acquisition unit 211 may automatically acquire system information.
For example, the system information can be automatically acquired by executing an actual task or a task equivalent to the actual task, and by executing actual communication or communication equivalent to the actual communication. Small device delays and the like can be calculated and corrected.

***Description of Example 3***
The scheduling device 200 may generate the optimum schedule 391 for each distributed processing device 300 by changing the allocation destination of each task. In other words, the scheduling device 200 may calculate an optimal, shortest time allocation with less communication traffic by changing the allocation destination of each task. The optimal schedule 391 is the schedule with the shortest overall execution time of the multiple tasks, and the schedule with the least overall communication volume of the multiple tasks.

Embodiment 2.
An embodiment in which distributed processing is realized by one distributed processing apparatus 300 will be described mainly with reference to FIGS. 16 to 18 for differences from the first embodiment.

*** Configuration description ***
The configuration of the distributed processing system 100X will be described based on FIG.
The distributed processing system 100X includes one distributed processing device 300X instead of multiple distributed processing devices (300A to 300C) (see FIG. 1).
The distributed processing device 300X comprises a multi-core processor to replace multiple distributed processing devices (300A to 300C).

The configuration of the distributed processing device 300X will be described with reference to FIG.
The distributed processing device 300X comprises a plurality of processors (301A-301C).
The processor 301A replaces the distributed processing device 300A (see FIG. 1). That is, the processor 301A controls the target device 101A.
The processor 301B replaces the distributed processing device 300B (see FIG. 1). That is, the processor 301B controls the target device 101B.
The processor 301C replaces the distributed processing device 300C (see FIG. 1). That is, the processor 301C controls the target device 101C.

Each of the plurality of processors (301A to 301C) includes elements such as an execution control unit 311, a task execution unit 312, a communication execution unit 313, a device control unit 314, and a schedule reception unit 319.

***Description of operation***
The distributed processing method will be described with reference to FIG.
The distributed processing method in the second embodiment is the same as the distributed processing method in the first embodiment (see FIG. 4).
However, the plurality of tasks are distributedly processed by the plurality of processors 301 of the distributed processing device 300X instead of the plurality of distributed processing devices 300. FIG.

*** Effect of Embodiment 2 ***
According to the second embodiment, it is possible to realize distributed processing with one distributed processing device 300 .

Embodiment 3.
A mode of adjusting the timings of starting execution of tasks and communications in the distributed processing device 300 will be described mainly with reference to FIG. 19 for differences from the first embodiment.

*** Configuration description ***
The configuration of the distributed processing device 300 will be described with reference to FIG.
The distributed processing device 300 further includes a task processing waiting section 315 and a communication waiting section 316 .
The distributed processing program further causes the computer to function as a task processing waiting section 315 and a communication waiting section 316 .

***Description of operation***
The distributed processing method in Embodiment 3 is the same as the distributed processing method in Embodiment 1 (see FIG. 4).
However, in step S120, the provisional schedule determining unit 212 generates a provisional schedule by giving margins to the processing time of each task and the communication time. As a result, in step S200, a schedule 391 is generated with a margin between the processing time of each task and the communication time.

Also, in step S140, the task processing standby unit 315 and the communication standby unit 316 operate as follows.
The task processing standby unit 315 causes the task execution unit 312 to wait until the execution time of each task specified in the schedule 391 before the task execution unit 312 executes each task. For example, the task processing standby unit 315 does not input each task to the task execution unit 312 until the execution time of each task specified in the schedule 391 is reached.
The communication standby unit 316 causes the communication execution unit 313 to wait until the execution time of each communication specified in the schedule 391 after each task is executed. For example, the communication standby unit 316 does not input the execution result of each task to the communication execution unit 313 until the execution time of each communication specified in the schedule 391 .

*** Effect of Embodiment 3 ***
By the distributed processing method, communication timing is estimated from the processing time of each task, and a communication schedule is established. However, although the task start timing is synchronized with the communication schedule, the task execution timing in each processor is the processing time estimated from the processing content or the measured processing time. Therefore, the processing time may increase or decrease depending on the characteristics of the processor or unexpected processing such as interrupt processing. Alternatively, if the input information of each task changes, that is, if the information input changes according to the state of the controlled object even if the processing is performed periodically, the amount of calculation increases or decreases, and the processing time increases accordingly. may increase or decrease. For communications, the data to be sent may also taper off. Also, there is a possibility that the delay due to transmission processing in a switch or the like differs for each transmission timing. If such a phenomenon occurs, there is a possibility that frame transmission will start at an unintended timing only by granting transmission permission to a frame with a specific priority during a specific period. If the frame transmission matches a transmission schedule with the same priority, the frame will be transmitted at a different timing than originally intended. If the transmission of the frame A interferes with the transmission of the frame B that should have been transmitted at that timing, the conventionally planned schedule will be shifted and the schedule will be broken. The problem is that task processing is not executed at specific timings and frames are not transmitted at specific timings. In addition, task processing and communication processing may be perturbed.
As a countermeasure against these, the distributed processing device 300 is provided with a task processing waiting section 315 and a communication waiting section 316 . As a result, the distributed processing device 300 can execute each task and each communication according to the timing of the schedule 391 according to the schedule 391 . Moreover, as a countermeasure against fluctuation, the scheduling apparatus 200 can generate the schedule 391 by assuming fluctuation and giving margins to the processing time of each task and each communication time.

***Description of Examples***
Embodiment 3 may be implemented in combination with Embodiment 2. A single distributed processing device 300X substituting for a plurality of distributed processing devices (300A to 300C) may comprise the task processing waiting section 315 and the communication waiting section 316. FIG.

*** Supplement to the embodiment ***
The hardware configuration of the scheduling device 200 will be described based on FIG.
The scheduling device 200 comprises processing circuitry 209 .
The processing circuit 209 is hardware that implements the scheduling section 210 and the schedule output section 220 .
The processing circuitry 209 may be dedicated hardware, or may be the processor 201 that executes programs stored in the memory 202 .

If processing circuitry 209 is dedicated hardware, processing circuitry 209 may be, for example, a single circuit, multiple circuits, a programmed processor, a parallel programmed processor, an ASIC, an FPGA, or a combination thereof.
ASIC is an abbreviation for Application Specific Integrated Circuit.
FPGA is an abbreviation for Field Programmable Gate Array.

The scheduling device 200 may include multiple processing circuits that replace the processing circuit 209 . A plurality of processing circuits share the functions of the processing circuit 209 .

In the processing circuitry 209, some functions may be implemented in dedicated hardware and the remaining functions may be implemented in software or firmware.

Thus, each function of the scheduling device 200 can be realized by hardware, software, firmware, or a combination thereof.

The hardware configuration of the distributed processing device 300 will be described with reference to FIG.
The distributed processing device 300 includes processing circuitry 309 .
The processing circuit 309 is hardware that realizes the execution control unit 311 , the task execution unit 312 , the communication execution unit 313 , the device control unit 314 and the schedule reception unit 319 .
The processing circuit 309 may be dedicated hardware, or may be the processor 301 that executes a program stored in the memory 302 .

If processing circuitry 309 is dedicated hardware, processing circuitry 309 may be, for example, a single circuit, multiple circuits, a programmed processor, a parallel programmed processor, an ASIC, an FPGA, or a combination thereof.

The distributed processing device 300 may include a plurality of processing circuits that substitute for the processing circuit 309 . A plurality of processing circuits share the functions of the processing circuit 309 .

In the processing circuitry 309, some functions may be implemented in dedicated hardware and the remaining functions may be implemented in software or firmware.

Thus, each function of the distributed processing device 300 can be realized by hardware, software, firmware, or a combination thereof.

Each embodiment is an illustration of preferred modes and is not intended to limit the technical scope of the present invention. Each embodiment may be implemented partially or in combination with other embodiments. The procedures described using flowcharts and the like may be changed as appropriate.
The “unit”, which is an element of each device described in each embodiment, may be read as “processing” or “step”.

100 Distributed Processing System 101 Target Device 102 Transfer Device 111 Task Information 112 Directed Graph 113 Topology Graph 114 Temporary Schedule Graph 200 Scheduling Device 201 Processor 202 Memory 203 Auxiliary Storage Device 204 External Interface 209 Processing Circuit 210 Scheduling Unit 211 Information Acquisition Unit 212 Temporary Schedule Determination Unit 213 Conflict Resolution Unit 220 Schedule Output Unit 290 Storage Unit 300 Distributed Processing Unit 301 Processor 302 Memory 303 Auxiliary Storage Device 304 External Interface , 309 processing circuit, 311 execution control unit, 312 task execution unit, 313 communication execution unit, 314 device control unit, 315 task processing standby unit, 316 communication standby unit, 319 schedule reception unit, 390 storage unit, 391 schedule.

Claims (7)

A distributed processing system in which multiple tasks are distributedly processed by multiple processors,
the execution order of the plurality of tasks, the amount of processing required to execute each task, the amount of communication required for communication for giving the execution result of each task to the next task, the priority of each task, and the priority of each communication an information acquisition unit that acquires task information indicating a degree and an assignment destination that identifies a processor to which each task is assigned;
Designating the execution time assigned to each task and the execution time assigned to each communication based on the execution order of the plurality of tasks, the processing amount of each task, the communication amount of each communication, and the assignment destination of each task a tentative schedule determining unit that determines a schedule;
Based on the priority of each task and the priority of each communication, by adjusting the execution time of each task specified in the tentative schedule and the execution time of each communication specified in the tentative schedule, a conflict resolution unit that generates a schedule in which conflicts between and conflicts between communications are resolved;
Distributed processing system with The conflict resolution unit adjusts the execution time of a task with a lower priority when at least part of the execution time overlaps between tasks assigned to the same processor, and after execution of the task whose execution time is adjusted 2. The distributed processing system according to claim 1, wherein the execution time of each task to be executed and the execution time of each communication executed after execution of the task whose execution time has been adjusted are adjusted. The conflict resolution unit adjusts the execution time of communication with lower priority when at least part of the execution time overlaps between communications executed using the same communication path, and the communication whose execution time is adjusted. 3. The distributed processing system according to claim 1, wherein the execution time of each task to be executed after execution of and the execution time of each communication to be executed after execution of the communication whose execution time has been adjusted are adjusted. The distributed processing system comprises the plurality of processors,
Each processor executes each assigned task at the execution time of each task specified in the schedule, and communicates the execution result of each assigned task at the execution time of each communication specified in the schedule. 4. The distributed processing system according to claim 1. A distributed processing method for distributed processing of a plurality of tasks by a plurality of processors,
The information acquisition unit obtains the execution order of the plurality of tasks, the amount of processing required to execute each task, the amount of communication required for communication for giving the execution result of each task to the next task, and the priority of each task. , obtaining task information indicating the priority of each communication and the assignee identifying the processor to which each task is assigned;
A provisional schedule determining unit determines the execution time assigned to each task and the execution assigned to each communication based on the execution order of the plurality of tasks, the processing amount of each task, the communication amount of each communication, and the assignment destination of each task. Decide on a tentative schedule, specifying times and
A conflict resolution unit adjusts the execution time of each task specified in the provisional schedule and the execution time of each communication specified in the provisional schedule based on the priority of each task and the priority of each communication. A distributed processing method for generating a schedule in which competition between tasks and competition between communications are resolved. A distributed processing program for distributed processing of a plurality of tasks by a plurality of processors,
the execution order of the plurality of tasks, the amount of processing required to execute each task, the amount of communication required for communication for giving the execution result of each task to the next task, the priority of each task, and the priority of each communication an information acquisition process for acquiring task information indicating a degree and an assignment destination that identifies a processor to which each task is assigned;
Designating the execution time assigned to each task and the execution time assigned to each communication based on the execution order of the plurality of tasks, the processing amount of each task, the communication amount of each communication, and the assignment destination of each task a tentative schedule determination process for determining a schedule;
Based on the priority of each task and the priority of each communication, by adjusting the execution time of each task specified in the tentative schedule and the execution time of each communication specified in the tentative schedule, a conflict resolution process for generating a schedule in which conflicts between and conflicts between communications are resolved;
Distributed processing program for executing on a computer. The execution order of multiple tasks, the amount of processing required to execute each task, the amount of communication required for communication to give the execution result of each task to the next task, the priority of each task, and the priority of each communication and an assignment destination that identifies a processor to which each task is assigned, and an information acquisition unit that acquires task information indicating
Designating the execution time assigned to each task and the execution time assigned to each communication based on the execution order of the plurality of tasks, the processing amount of each task, the communication amount of each communication, and the assignment destination of each task a tentative schedule determining unit that determines a schedule;
Based on the priority of each task and the priority of each communication, by adjusting the execution time of each task specified in the tentative schedule and the execution time of each communication specified in the tentative schedule, a conflict resolution unit that generates a schedule in which conflicts between and conflicts between communications are resolved;
A scheduling device comprising:

Images