JP4523921B2 - Computer resource dynamic controller - Google Patents

Description

  The present invention relates to an apparatus and a system for preventing frequent occurrence of deadline mistakes due to inadequate processing in a distributed system in which a plurality of computers form a network and cooperate to process a plurality of processes in parallel.

  According to the conventional computer resource dynamic control method, for example, according to Patent Document 1 as Conventional Example 1, a process based on the processing time and the data content of the processing data, the data size, and the previous processing time for each repetition period. A means for predicting the next processing time and a task execution order plan determined for each range of the predicted processing time are stored, and are determined in advance based on the data of the next processing time obtained by the prediction. By determining the processing order of the next cycle based on the proposed execution order, it is possible to prevent the occurrence of a process deadline miss.

In addition, according to other patent document 2 as Conventional Example 2, the process amount to be processed is unitized, for example, 8 units for a certain process, 5 units for another process, etc. Thus, a method of assigning a processing unit to each processor is shown. Therefore, it has a process management means for managing the unit amount of the process, a processor management means for managing the processor to be executed, and an allocated resource amount determination means.
JP-A-4-171538 JP-A-6-28323

In the conventional example 1 described above, process migration (transfer to another computer) cannot be performed, and even if there is an empty resource in another computer, there is no choice but to change or degenerate the processing contents. There is a problem that other computer resources cannot be used efficiently, or that deadline mistakes caused by a computer (processor) entrusted with the process of overloading continue.
Further, in the above conventional example 1, the next processing time is predicted mainly based on the data size and value, but the processing time is not determined by the data size. Furthermore, each process receives various data, but the content of the data is often unknown until just before the process receives, and cannot be predicted in advance. For this reason, it is impossible to predict the next periodic processing time by the method of Conventional Example 1.
In the conventional example 2, the process amount is also divided and managed in units, and the computer is not allocated according to the contents or quality of the process or the processing amount is not changed depending on the computer environment. There is a problem that it is not possible to prevent the occurrence of deadline mistakes.

The present invention has been made to solve the above-described problems. When a deadline miss of each process constituting the system occurs, process migration is performed to a computer having a sufficient free memory amount and CPU time. The purpose is to prevent the continuous occurrence of process deadline mistakes.
In addition, by predicting the future processing time of the process from the processing time of the past process, if it is predicted that a future deadline miss will occur, the process migration is performed to prevent the occurrence of a process deadline miss. For the purpose.

The computer resource dynamic control device according to the present invention connects a plurality of computers that process a target process, and connects to a system that processes the plurality of processes in a coordinated manner.
A computer load collection unit that monitors the resource usage status of each of the above computers, collects the usage status information of this resource of the computer, and detects a shortage of this resource;
A processing time collecting unit that monitors the processing time of the process and collects the processing time of the process, and detects the occurrence of a deadline miss that is a process that exceeds the deadline time that defines the processing limit time of the process;
The processing time of the above-mentioned process processed in the system, the use amount of the computer resource used in the processing of this process, and the deadline time are stored together with the process identifier, and the owned resource amount of the above-mentioned computer connected to the system is calculated. A system configuration information table memory for storing together with the identifier of
A transition process determining unit that recommends reassignment of processing of the process or related process to a computer different from the current computer by referring to the system configuration information table / memory according to the notification that the occurrence of the deadline miss is detected;
In response to the notification recommending the reallocation, the processing time of the process is within the deadline time by referring to the processing time of the process in the system configuration information table memory and the amount of resources owned by the computer. A resource allocation unit that extracts a process and a candidate computer for which reassignment has been recommended, causes the extracted process to be allocated to the extracted candidate computer, and simultaneously updates the system configuration information table and memory;

According to the present invention, a deadline miss of a process is detected, and the amount of computer resources required by the process and the amount of resources possessed by each computer are managed. By allocating to other computers or other computers with sufficient computer resources, there is an effect of preventing the continuation of deadline mistakes.
In addition, if the cause of the deadline mistake is a computer failure, all processes that were running on the failed computer can be reassigned to other computers, and processing can be continued, resulting in a highly available system. There is an effect.

Embodiment 1 FIG.
FIG. 1 is a diagram showing a configuration of a computer resource dynamic control system according to the present embodiment. That is, the resource dynamic control system according to the present embodiment shares the components necessary for the system in a distributed computer system environment in which a plurality of computers 100 are connected by the network 102, and the processing as a whole is performed. Prevent frequent deadline mistakes due to not being in time. In the figure, each computer 100 shares the processing of the process 101 or performs it independently, and processes a new process according to resource allocation.
First, the hardware configuration of a computer will be described. FIG. 2 is a diagram illustrating a configuration when the computer 100 is a computer resource dynamic control device 100a. In these drawings, a computer 100 or a computer resource dynamic control device 100a includes a processor (CPU) 11, a memory 12 for storing programs and processes 101, a processing time monitoring unit 1, a processing time collection unit 2, and a migration process determination unit 3. , Computer load monitoring unit 4, computer load amount collection unit 5, resource allocation unit 6, system configuration information management table / memory 7, processing time history table / memory 8, computer load history table / memory 9, computer state table / memory 10 , Input device 13, output device 14, communication interface 15, and process 101 operating on each computer.
The detailed connection of each computer is not shown in the configuration of FIG. 1 to show an overall configuration example, but each computer 100 has the same configuration as the detailed configuration of FIG. For example, in the rightmost computer, only the processing time monitoring unit 1, the computer load monitoring unit 4, the resource allocation unit 6, and the process 101 written in the memory 12 (not shown) are displayed in a frame, but the processor shown in FIG. 11 and the memory 12 via an internal bus 16. Further, the constituent elements such as the resource allocation unit 6 obtain the function by developing a program having functions as described in the following functional description in the memory, and reading and executing the program by the processor 11.
This process 101 is read by a CD reader which is a kind of the input device 13 or another recording medium reader, or is taken in by the communication interface 15 from another computer via a communication line or the like. As shown, it is actually expanded and stored in the memory 12. The process 101 processed in the present embodiment may be an application in which the processing load greatly varies depending on the content of data rather than the size of the data, such as a radar application.

Hereinafter, the function of each component in the computer resource dynamic control device 100a or the computer resource dynamic control system in the present embodiment will be described.
FIG. 3 is a diagram illustrating the function of the processing time monitoring unit 1 that monitors the data processing time of the process according to the present embodiment. That is, when each processor 11 that performs distributed processing executes the processing of the assigned process 101, the time at which the processing is started after receiving data (hereinafter referred to as “processing start time”), and the time at which the processing is completed ( (Hereinafter referred to as “processing end time”) is provided with a step of reporting to the processing time monitoring unit 1 and the time (in actual use) of the CPU (processor) for the calculation between the processing start time and the processing end time ( (Hereinafter referred to as “CPU time”). The steps related to the operation will be described later in the description of the operation. Furthermore, the processing time monitoring unit 1 has a role of notifying the processing time collecting unit 2 of the above processing start time, processing end time, and CPU time as necessary.
One processing time monitoring unit 1 may be provided for each computer 100, one for each process 101, or one for the entire system. FIG. 1 shows a configuration in which one computer 100 is provided as an example of the configuration.

4, 5, and 6 are diagrams for explaining main functions of the processing time collection unit 2 in the present embodiment. As shown in FIG. 4A, the processing time collection unit 2 receives the processing start time, the processing end time, and the CPU time from the processing time monitoring unit 1, and divides these data for each process, thereby processing the processing time history table. Hold in memory 8 a certain number of times in the past. FIG. 4B is a diagram showing an example of data held in the processing time history table memory 8.
Further, as shown in FIG. 5A, the number of deadline misses that have occurred in the history that is held is investigated, and if a deadline miss has occurred more than an allowable number of times, It has a function of notifying the migration process decision unit 3 that a deadline miss has occurred a specified number of times.
Further, as shown in FIG. 6, when there is a request from the migration process determination unit 3 to transmit a history of processing start time, end time, and CPU time of a certain process, a processing time history having data of the corresponding process It has a function of transmitting data held in the table memory 8. The number of past data necessary for the processing time collecting unit 2 to execute the function, the number of past data held in the processing time history table memory 8, the deadline time defining the processing limit time of the process, and this deadline time The allowable number of deadline misses that exceed the processing is described in the system configuration information management table memory 7.
One processing time collection unit 2 may be provided for each process, one for each computer, or one for the entire system.

FIG. 7 is a diagram illustrating the function of the migration process determination unit 3 in the present embodiment. When the migration process determination unit 3 receives a notification from the processing time collection unit 2 that the deadline miss has occurred more than the specified number of times, it makes an inquiry to the processing time collection unit 2, the computer load collection unit 5 and the like to process other processes. Hardware resources should be reassigned to eliminate deadline errors by investigating time fluctuations, computer status (normal or abnormal), CPU and memory utilization, etc. Has the ability to detect processes.
One migration process determining unit 3 may be provided in the system.

  FIG. 8 is a diagram for explaining the function of the computer load monitoring unit 4 in the present embodiment. The computer load monitoring unit 4 is provided on each computer 100 and has a function of periodically acquiring the memory usage status and CPU usage rate of each computer 100 and notifying the computer load amount collecting unit 5 of these. This information is periodically notified to the computer load amount collecting unit 5 that the computer 100 is operating normally.

9, FIG. 10, FIG. 11, and FIG. 12 are diagrams for explaining the main functions of the computer load amount collection unit 5 in the present embodiment. As shown in FIG. 9, the computer load amount collection unit 5 receives a notification from the computer load monitoring unit 4, classifies the memory availability, the usage status of the CPU 11, and the reception time of the above information for each computer, and calculates the computer load history. Stored in the table memory 9.
Further, when there is a request from the migration process determination unit 3 or the resource allocation unit 6 to transmit information on the usage rate of the CPU 11, the amount of free memory, and normality / abnormality of the computer, as shown in FIG. It also has a function of extracting the corresponding data from the data held in the history table memory 9 and transmitting it.
Further, when the data is not received from the computer load monitoring unit 4, the entry of the computer in which the corresponding data is not received in the computer status table memory 10 indicates “abnormal” as shown in FIG. 11B. Thus, it has a function of notifying the migration process determining unit 3 that a failure has occurred at the same time.
Further, it has a function of referring to the data in the computer load history table memory 9 and notifying the migration process determining unit 3 when there is a computer whose free memory amount or free CPU resource is reduced. As shown in FIG. 12, when there is a request for obtaining a list of normal computers 100 from the migration process determination unit 3 or the resource allocation unit 6, the computer state table / memory 10 is referred to and the normal operation is performed. It also has a function of returning a list of identifiers of the computer 100.

  FIG. 13 is a diagram illustrating the function of the resource allocation unit 6 in the present embodiment. The resource allocation unit 6 has a function of newly allocating a process selected by the migration process determination unit 3 to be migrated to another computer 100. When the resource allocation unit 6 allocates a different computer to the process, the computer load amount collection unit 5 is inquired about the load of the CPU 11 of each computer 100 and the usage status of the memory 12, or the system configuration information management table memory 7 The CPU 101 starts the process 101 on the computer 100 that is less likely to cause a memory shortage or a deadline miss by acquiring information on the CPU time and memory amount required by the process 101. Further, after the process 101 is started, the process operating on the original computer is stopped.

  FIG. 14 is a diagram showing an example of the data configuration of the system configuration information management table / memory 7 in the present embodiment. The system configuration information management table / memory 7 includes, as information related to software, a “process name” that is the name of the process 101, a process start time, a process end time, and a CPU that are stored to count the number of occurrences of the most recent deadline miss. "Number of histories stored in the processing time history table" indicating the number of time histories, "Average CPU time", "Deadline time", "Deadline miss allowable number in history", "Operation" The usage amount of computer resources such as “in-computer”, “process identifier”, “consumed memory”, and the like is held for each process 101. As hardware information, “host name” for identifying the computer 100 existing in the system, owned resource amount such as “installed memory amount” of the computer, and “number of histories held in the computer load history table” are described for each computer. .

  FIG. 15 is a diagram showing an example of the data configuration of the processing time history table memory 8 in the present embodiment. This table exists for each process, and “processing start time”, “processing end time”, and “CPU time” are stored for each entry in the table. In this table, when the processing time monitoring unit 1 transmits “processing start time”, “processing end time”, and “CPU time” to the processing time collecting unit 2, the data is written by the processing time collecting unit 2. The number of entries is determined by the system configuration information management table. When data is written in all entries, the oldest data is deleted in order. If a deadline miss occurs more than the specified number of times, all the corresponding data in the table for managing the data of the process that performed the resource allocation is deleted after the resource allocation is completed.

FIG. 16 is a diagram showing an example of the data configuration of the computer load history table / memory 9 in the present embodiment. This table exists for each computer, and data of “data acquisition time”, “CPU usage rate”, and “free memory amount” is stored for each entry in the table. This table is written by the computer load amount collection unit 5 when the computer load monitoring unit 4 sends data of “data acquisition time”, “CPU usage rate”, and “free memory amount” to the computer load amount collection unit 5. The number of entries in the table is determined by the system configuration information management table. When data is written in all entries, old data is deleted in order.
FIG. 17 is a diagram showing an example of the data configuration of the computer state table / memory 10 in the present embodiment. In this table, there is an entry for storing “normal / abnormal” for each computer, and the computer load amount collection unit 5 receives “data acquisition time”, “CPU usage rate”, “ If the “Available Memory” data is not received, it is determined that an abnormality has occurred in the computer, and the entry of the corresponding computer is indicated as “abnormal”.
As shown in FIG. 2, the computer resource dynamic control apparatus 100a may have all the necessary components. However, as shown in FIG. 1, some computers need to collect processing time. The constituent elements such as the prediction unit 2 may be distributed and these computer groups may constitute a computer resource dynamic control system as a whole.

Next, the operation of each component of the computer resource dynamic control device or computer resource dynamic control system in the present embodiment will be described.
FIG. 18 is a diagram illustrating an example of an operation flow of the processing time monitoring unit 1. The operation will be described with reference to the function explanatory diagram of FIG. In step S1 (hereinafter, description of steps is omitted), when the processor 11 starts processing of each process 101, the processing time monitoring unit 1 acquires the processing start time of the process. In S2, when the processor 11 finishes the process, the process end time and the CPU time in the process time from S1 to S2 are acquired. In S3, the processing time collection unit 2 is notified of the processing start time, the processing end time, and the CPU time. This is a specific operation by the function of the processing time monitoring unit 1 shown in FIG.
The processing time monitoring unit 1 repeats the processing of S1 to S3 as long as the processor 11 executes the processing operation of the process 101. The acquisition of the CPU time in the process of each process is obtained by, for example, a times () system call, a get (age) system call, a ps command, etc. if the operating system is UNIX (registered trademark).

  FIG. 19 is a diagram illustrating an example of an operation flow of the processing time collection unit 2. The operation will be described with reference to the function explanatory diagrams of FIGS. In S11, when processing start time, end time, and CPU time data for processing a process is received from the processing time monitoring unit 1, in S12, the process processing data is written in a certain entry in the processing time history table memory 8. Write these data. In S13, it is checked whether or not a deadline miss has occurred in the processes of all processes from the process start time and process end time written in S12. Whether or not a deadline miss has occurred is obtained by subtracting the process start time from the process end time to obtain the time required for the process, and this time and the process deadline indicated in the system configuration information management table memory 7 It can be obtained by comparing the line times. If a deadline miss has occurred, in S14, all data held in the processing time history table memory 8 are referred to and it is checked whether or not a deadline miss has occurred more than the specified number of times. The specified number of allowable deadline misses is held in the system configuration information management table memory 7. If a deadline miss has occurred more than the specified number of times, in S15, the migration process determination unit 3 is notified and all data written in the processing time history table memory 8 is deleted. If no deadline miss has occurred in S13, the process proceeds to S16, and the entry for writing data next in the processing time history table memory 8 is advanced by one. When there are no more empty entries in the processing time history table memory 8, thereafter, the oldest data is overwritten in order in S12.

  FIG. 20 is a diagram showing an example of an operation flow of the computer load monitoring unit 4 in the present embodiment. The operation will be described with reference to the function explanatory diagram of FIG. In S <b> 1001, the computer load monitoring unit 4 acquires the processing start time by the processor 11. In step S1002, the computer load and the memory usage are acquired. In step S1003, the computer load amount collecting unit 5 is notified of the acquired current time, CPU usage rate, and memory usage amount. For example, if the operating system is UNIX (registered trademark), the CPU usage rate acquisition method can be obtained by the mpstat command, the top command, and the memory usage acquisition method by the vmstat command.

FIG. 21 is a diagram illustrating an example of an operation flow of the computer load amount collection unit 5 in the present embodiment. The operation will be described with reference to the function explanatory diagrams of FIGS. In step S1011, the process waits for a certain period of time. In S1013, the column of the computer that transmitted the data in the computer state table memory 10 is set to “normal”. In step S1014, data is written to the corresponding entry in the computer load history table memory 9. In the computer load history table, as shown in FIG. 9 (b), entries are divided for each computer, so the entry is written in the entry of the computer that sent the notification. In S1015, it is checked whether the written data indicates a resource shortage. The resource shortage may be, for example, a case where the CPU usage rate reaches 100% or the amount of used memory exceeds the amount of installed memory. Further, the upper limit of the CPU usage rate and the used memory amount may be set by the user, and if the upper limit is exceeded, the resource may be insufficient. If a resource shortage has occurred, in S1016, the migration process determination unit 3 is notified of the shortage of resources and the computer name. In S1017, in order to write data next, the entry in the computer load history table / memory 9 is advanced by one. When there are no more empty entries in the computer load history table memory 9, the oldest data is overwritten in order thereafter.
If no data arrives from the computer load monitoring unit 4 after waiting for a predetermined time in S1011, the computer load monitoring unit 4 that has not transmitted the data in the computer state table memory 10 in S1018. The column of the computer to be monitored is changed to “abnormal”, and the migration process determination unit 3 is notified in S1019. The procedure of S1011 to S1019 shows the processing of collecting the computer load for one computer 100, and this is performed for all the computers 100.

  FIG. 22 is a diagram illustrating an example of an operation flow of the migration process determination unit 3 when the processing time collection unit 2 is notified that a deadline miss has occurred. In FIG. 5, as shown in FIG. 5, in S101, the processing time collection unit 2 receives a notification notifying that a deadline miss has occurred a predetermined number of times or more. Along with the notification, for example, the name of the process in which a deadline miss has occurred is also received. When the name of the process in which the deadline miss has occurred is received in S101, in S102, the computer that processes the process in which the deadline miss has occurred is identified from the system configuration information management table memory 7. In step S103, another process operating on the computer is identified and its name is acquired. In S104, the computer load amount collecting unit 5 is inquired as to whether or not the computer is operating normally. Then, in S105, if the computer identified in S103 is not operating normally, it can be seen that the cause of the deadline error is that the computer has stopped. In that case, in order to transfer all the processes 101 detected in S103 to another computer, in S106, these process names are notified to the resource allocation unit 6 as targets for resource reallocation.

  If the computer is operating normally in S105, in S107, a process in which the CPU time is increased among the processes found in S103 is investigated. Whether the CPU time has increased or not is determined because the system configuration information management table holds the average CPU time of each process, so that the average CPU time is obtained by looking at the system configuration information management table and compared with the current CPU time. I understand. Then, the process in which the deadline miss has occurred and the process found in S107 are set as candidates for resource reallocation, and the resource allocation unit 6 is notified to perform resource allocation for any of them.

  FIG. 23 is a diagram illustrating an example of an operation flow performed by the migration process determination unit 3 when a notification is received from the computer load collection unit. The operation will be described with reference to the function explanatory diagram of FIG. If the load amount is larger than that of the computer load amount collection unit 5, it is considered that the notification may arrive if there is a computer failure or the memory capacity is insufficient. When such a notification is received from the computer load amount collecting unit 5 in S111, the process 101 being processed on the computer 100 in which an abnormality has occurred is identified by referring to the system configuration information management table memory 7 in S112. To do. In S113, it is checked whether the content of the notification from the computer load amount collecting unit 5 is a computer failure or a memory shortage. If the content of the notification is a computer failure, in S114, the resource allocation unit 6 is notified of all processes examined in S112 as resources to be reassigned. If the content of the notification is insufficient, in step S112, all the processes 101 identified and found are candidates for resource reallocation so that any one of these processes is transferred to another computer 100. Notify the allocation unit 6.

FIG. 24 is a diagram illustrating an example of an operation flow of the resource allocation unit 6. The operation will be described with reference to the function explanatory diagram of FIG. In S121, the migration process determination unit 3 receives a list of resource reassignment candidates or processes to be reassigned. When the list of resource reassignment candidates arrives, it is a case where a deadline miss has occurred and the memory is insufficient. In this case, if at least one of the processes shown in the received list is addressed Good. When the target list for resource reallocation arrives, it is a case where a computer failure has occurred. In this case, it is necessary to perform resource reallocation for all the received processes.
In S122, the process names described in the list received in S121 are stored in a table in the order of, for example, the CPU time. In S124, one process name is extracted from the process table created in S122. In S126, a list of normally operating computers 100 is acquired from the computer load amount collection unit 5, and is stored in a table in order of increasing load, for example. Then, S127 and the subsequent steps are repeated for all the computers shown in the table of S126 or until an assignment destination computer is found.
In S127, one computer with the highest load is taken out from the table showing the list of computers 100. In S128, it is confirmed whether or not the process 101 extracted in S124 can operate on the computer 100 selected in S127. As a method of confirming whether or not it can be operated, for example, a method of simply judging based on the CPU usage rate of a computer, or actually determining the processing timing based on the processing cycle and CPU time of each process, a deadline mistake is detected. A method of analyzing whether or not scheduling can be performed without occurring can be considered.

If the resource allocation unit 6 determines in S129 that the process 101 extracted from the table in S124 is operable on the computer 100 selected in S127, then in S130, the free memory of the computer extracted in S126. The amount is checked, and it is confirmed whether or not the process taken out in S124 is more than the amount of memory used. If sufficient memory is available, in S132, the computer selected in S127 is determined as the migration destination of the process extracted in S124. In S133, the process determined in S124 is activated and processed by the computer determined in S132, and the process of the process performed by the computer that has been operating is stopped.
In S134, the process extracted from the table in S124 of the system configuration information management table memory 7, the computer to which the process is assigned (that is, the computer extracted in S127) and the process extracted from the table in S124. The information related to the current CPU time is rewritten, and the process returns to S123 to perform the process indicated by the next entry in the table created in S122.
When a list of processes is given as a “resource reassignment candidate process group” from the migration process determination unit 3, it is not necessary to perform resource reassignment for all the notified processes 101. If the above process reassignment is determined, the process may be terminated without performing the process of assigning other processes shown in the table created in S122.

If it is determined in S129 that the computer 100 selected in S127 causes a deadline miss when the process 101 is processed, the process returns to S126, and the computer having the next largest load is selected and the processes from S127 are performed.
In S131, if the computer selected in S127 does not have the amount of free memory necessary to process the process selected in S124, the process returns to S126, and S127 and subsequent steps are executed for the computer in the next entry in the table. To do. In S126, if S127 is executed for all the computers and the allocation destination is not found, the resource allocation of the process is abandoned and the process proceeds to S135, the process selected in S124 is stopped, and the process proceeds to S122 in S136. Delete the stopped process name from the created table.
Note that the reallocation method shown here is merely an example, and there are many examples in which the operations of S124 to S136 are applied, but other examples are omitted here.

In FIG. 25, the computer resource dynamic control device 100a or the computer resource dynamic control system in which the components are distributed is linked based on the operations shown in FIGS. FIG. 5 is a diagram showing an overall operation flow that operates and prevents the continuation of a deadline miss.
In S141, the processing time collection unit 2 or the computer load amount collection unit 5 detects the occurrence of a deadline miss or the abnormality of the computer, the CPU, and the memory / resource shortage, and notifies the migration process determination unit 3 of them. The operation of the processing time collection unit 2 in this case is as shown in S11 to S16, and the operation of the computer load amount collection unit 5 is as shown in S1011 to S1018. Thereafter, the migration process determination unit 3 investigates the contents of the notification in S142. If the content of the notification is a computer failure, the resource allocation unit 6 allocates resources according to the procedures of S121 to S136 for all processes operating on the failed computer in S144. If it is not a computer failure, in S145, the migration process determination unit 3 finds a process that causes a deadline error, a process that causes a resource shortage, and a process that performs resource reallocation. In S146, the resource allocation unit 6 performs resource allocation for these processes in the procedure of S121 to S136. After the resource allocation is completed, the resource allocation unit 6 in S147, the identifier of the process that newly allocated the resource in the system configuration information management table memory 7, the allocation destination computer that processes the process, and the CPU time required for the process Rewrite information about the new information.

In other words, in this embodiment, a plurality of computers are connected to a network, and a plurality of processes operate in cooperation or independently on each computer, and each process performs processing at a fixed time period and performs processing within a time limit. In a distributed real-time system that is required to complete the process, it has a processing time monitoring means for monitoring the processing time of the process and a computer load monitoring means for monitoring the resource usage status of the computer. In response to the change of the resources given to the process, the system prevents the continuation of process deadline mistakes. By adding the following configuration, process migration to other computers with sufficient free resources is possible. , Prevents the recurrence of deadline mistakes in all processes after process migration It shows a computer resource dynamic control method, characterized by. And
11) A list of processes constituting the system, an average processing time of each process, a consumed memory amount, a CPU time, a time limit for processing per process, a list of computers constituting the system, and a memory amount installed in the computer are retained. System configuration information management table.
12) Processing time collection means for holding the history of fluctuations in the processing time and CPU time of all processes operating on the system and detecting the occurrence of a process deadline miss.
13) Computer load collection means for holding a history of resource usage of all the computers operating on the system, notifying other means of usage of past computer resources, and detecting occurrence of resource shortage.
14) A migration process determination unit that determines a process for reallocating resources in order to continue real-time processing when a failure such as a deadline miss, CPU resource shortage, or deadline miss occurs.
15) Resource allocation means for acquiring a necessary resource amount from the system configuration information management table for the process determined by the migration process determining means and performing process migration on a computer having the necessary free resources is provided.

In other words, the computer resource dynamic control device according to the present embodiment connects a plurality of computers that process the target process, and connects to a system that processes the plurality of processes in cooperation.
A computer load collection unit that monitors the resource usage status of each of the above computers, collects the usage status information of this resource of the computer, and detects a shortage of this resource;
A processing time collecting unit that monitors the processing time of the process and collects the processing time of the process, and detects the occurrence of a deadline miss that is a process that exceeds the deadline time that defines the processing limit time of the process;
The processing time of the above-mentioned process processed in the system, the use amount of the computer resource used in the processing of this process, and the deadline time are stored together with the process identifier, and the owned resource amount of the above-mentioned computer connected to the system is calculated. A system configuration information table memory for storing together with the identifier of
A transition process determination unit that recommends reassignment of processing of the process to a computer different from the current computer by referring to the system configuration information table / memory by notification of occurrence of the deadline miss;
In response to the reassignment recommendation notification, the processing time of the process in the system configuration information table / memory is referred to the amount of resources owned by the computer so that the processing time of the process is within the deadline time. A candidate computer, and a resource allocation unit that updates the system configuration information table and memory at the same time that the extracted process is allocated and executed by the extracted candidate computer.
In this way, a deadline miss of a process is detected, and the process of the process 101 in which the deadline miss has occurred more than the specified number of times is assigned to another lightly loaded computer 100, or the process causing the deadline miss is calculated by the computer. By allocating to other computers with sufficient resources, it is possible to prevent continuation of deadline mistakes.
In addition, when the cause is a computer failure, the process can be continued by reassigning all the processes operating on the failed computer to other computers.

Embodiment 2. FIG.
In the first embodiment, a case has been described in which a process in which a deadline error has occurred is transferred to a light-loading computer based on the results of processing the process by the computer. An apparatus and system for reducing the probability of occurrence of a deadline miss by moving the process to another computer when a line miss is likely to occur will be described.
FIG. 26 is a diagram showing an example of the configuration of a computer resource dynamic control system in the present embodiment. The computer resource dynamic control system according to the present embodiment includes, in addition to the configuration according to the first embodiment, a deadline miss occurrence time prediction unit 21, a computer load prediction unit 22, a CPU time prediction unit 23, and a memory usage amount prediction unit 24. I have. The functions and operations of the processing time monitoring unit 1, the computer load monitoring unit 4, the system configuration information management table / memory 7 or the computer state table / memory 10 of the other components are the same as those in the first embodiment, and will be described. Omitted.
Similarly, FIG. 27 is a diagram showing a configuration of the computer resource dynamic control device 100b in the present embodiment. That is, instead of performing dynamic control as a system with the distributed arrangement shown in FIG. 26, a specific computer may be provided with all the components and become a control device.
Further, the processing time collection unit includes the deadline miss occurrence time prediction unit 21 and is configured as the processing time collection / prediction unit 2b, and the computer load collection unit includes the computer load prediction unit 22 to collect the computer load collection. -It is good also as a structure used as the estimation part 5b.

Hereinafter, the migration process determination unit 3b, the resource allocation unit 6b, the deadline miss occurrence time prediction unit 21 or the processing time collection / prediction unit 2b, the computer load prediction unit 22 or the computer load amount collection / prediction unit 5b in the present embodiment, The functions of the CPU time prediction unit 23 and the memory usage amount prediction unit 24 will be described.
The migration process determination unit 3b in the present embodiment has a function of receiving a notification from the processing time collection unit 2 as well as receiving a notification of the occurrence of a deadline miss from the deadline miss occurrence time prediction unit 21. This is different from Form 1. In addition, when a notice of the occurrence of a deadline miss is received from the deadline miss occurrence time prediction unit 21, the deadline miss is also inquired by inquiring to the computer load prediction unit 22, the memory usage prediction unit 24, and the CPU time prediction unit 23. Obtain the predicted value of the computer load, memory usage, and CPU time (the time that the processor of the computer processes the process) at the time of occurrence, and based on these predicted values, determine the cause of the occurrence of a deadline miss that may occur in the future The point to be estimated is different from the first embodiment.
When the migration process determining unit 3b receives a deadline miss notification from the deadline miss occurrence time predicting unit 21, the resource allocating unit 6b in the present embodiment has a computer load predicting unit 22, a CPU time predicting unit 23, a memory The difference from the first embodiment is that an appropriate computer 100 is selected as a process reallocation destination by making an inquiry to the usage amount prediction unit 24.

  FIG. 28 is a diagram illustrating the function of the deadline miss occurrence time prediction unit 21 in the present embodiment. The deadline miss occurrence time prediction unit 21 has a function of calculating the relationship between the time and the processing time from the processing start time and the processing end time indicated in the processing time history table memory 8 by using statistics or the like, and expressing it by an approximate expression. Have. Further, it has a role of predicting a deadline miss occurrence time using the obtained approximate expression and notifying the processing time collection / prediction unit 2b or the transition process determination unit 3b directly to the predicted deadline occurrence time. For example, the example of FIG. 28 shows a case where processing is performed at a cycle of 50 milliseconds and the processing time is increased by 10 milliseconds each time processing is performed. The approximate expression in this case can be obtained as y = 0.2 × (x−processing start time in the oldest data) +10 milliseconds (y is the processing time, x is the processing start time). Is 100 milliseconds, it can be seen that a deadline miss occurs at 10:10:10 seconds 450 milliseconds. In this case, the deadline miss occurrence time prediction unit 21 notifies the transition process determination unit 3b of a notice that “a deadline miss will occur at 10: 10: 10: 450 ms”.

FIG. 29 is a diagram showing the function of the computer load prediction unit 22 in the present embodiment. The computer load predicting unit 22 has a function of calculating the relationship between time and CPU load variation from a CPU usage rate shown in the computer load history table memory 9 by a technique such as statistics and expressing the variation with an approximate expression. . Further, it has a function of obtaining the CPU usage rate of each computer at the estimated deadline miss occurrence time by using the obtained approximate expression and notifying these values in response to a request from the migration process determining unit 3b or the like. Further, it has a function of obtaining a time when the CPU usage rate exceeds a certain upper limit value and notifying the migration process determination unit 3b. The certain upper limit value may be 100%, or may be determined in advance by the user.
For example, FIG. 29 shows a case where the load on a certain computer increases by 0.5% every 50 milliseconds. In this case, the approximate expression of the CPU usage rate can be obtained as y = 0.01 × (x−the reception time of the oldest data) +10 (y is the CPU usage rate and x is the time). From 3b, if there is an inquiry about the CPU load at 10:10:10 seconds 450 milliseconds, it will be answered 14.5%. If there is no data at the same time as the inquiry, the closest time data shown in the table is acquired and answered. Also, assuming that the allowable upper limit value of the CPU usage rate is 100%, from this equation, the CPU usage rate becomes 100% at 10:10:28.
The computer load prediction unit 22 may be provided for each process or may be one in the system. However, in the case of one system, approximate equations are obtained for all processes.

FIG. 30 is a diagram illustrating a function of the CPU time prediction unit 23 in the present embodiment. The CPU time prediction unit 23 has a function of obtaining the relationship between the time and the fluctuation of the CPU time from the CPU time of each process shown in the processing time history table memory 8 by using statistics or the like and expressing it by an approximate expression. Further, it has a function of obtaining and notifying the CPU time of each process at the time of predicting the occurrence of a deadline miss using the obtained approximate expression in response to a request from the migration process determining unit 3b and the like.
For example, FIG. 30 shows a case where the CPU time of a certain computer increases by 10 milliseconds every 50 milliseconds. In this case, an approximate expression of CPU time can be obtained as y = 0.2 × (x−processing start time in the oldest received data) +9 (y is CPU time, x is processing start time). If there is an inquiry about the CPU time of 10: 10: 10: 450 milliseconds from the process determination unit 3b, the CPU time prediction unit 23 will reply that the time is 99 milliseconds.

FIG. 31 is a diagram showing the function of the memory usage amount prediction unit 24 in the present embodiment. The memory usage amount prediction unit 24 has a function of obtaining the time and memory usage variation from the memory usage shown in the computer load history table / memory 9 by using statistics or the like, and expressing the variation by an approximate expression. Further, it has a function of obtaining the time when the memory shortage occurs based on the approximate expression of the memory usage, and notifying the migration process determining unit 3b. Further, it has a function of obtaining the memory usage of each computer at the estimated deadline miss occurrence time using the obtained approximate expression and notifying it in response to a request from the migration process determining unit 3b or the like. The amount of free memory when the amount of free memory becomes insufficient may be given to the system by the user, or when the amount of free memory becomes 0, the memory may be insufficient. Here, to simplify the description, it is assumed that the memory is insufficient when the free memory becomes zero.
For example, the example of FIG. 31 shows a case where the memory usage of a certain computer increases by 2 MB every 50 milliseconds. In this case, the approximate expression of the memory usage can be obtained as y = 0.04 × (x−the reception time of the oldest data) +100 (y is the memory usage, x is the time). If the amount of installed memory in this computer is 180 MB, a memory shortage will occur after 2 seconds, and this is notified to the processing time collection / prediction unit 2b or directly to the migration process determination unit 3b. . In addition, when there is an inquiry about the memory usage of 10: 10: 10: 450 milliseconds from the migration process determination unit 3b or the like, 118 MB is replied.
Although there may be one memory usage prediction unit 24 in the computer or one in the system, approximate expressions are obtained for all the computers.

FIG. 32 is a diagram illustrating an example of an operation flow of the deadline miss occurrence time prediction unit 21 in the present embodiment. The operation will be described with reference to the function explanatory diagram of FIG. In S21, the deadline miss occurrence time prediction unit 21 refers to all entries in the processing time history table memory 8 and obtains the start time and processing time for each entry. The processing time is obtained by subtracting the processing start time from the processing end time. In S22, the relation between the processing start time and the processing time is expressed by an approximate expression using a statistical technique. As a method of obtaining the approximate expression, for example, a least square method or the like can be considered. Further, the order of the approximate expression may be determined in advance, or a method may be considered in which the order is gradually increased from the first order after determining the upper limit of the order and an accurate approximate expression is obtained. Then, using the obtained formula, the deadline miss occurrence time is obtained in S23. For example, when the formula obtained in S22 is y = a _n x ⁿ + a _n-1 x ^n-1 +... + Ax + b (y: processing time, x: time), the value of y is dead. It can be obtained from the value of x when the line time value is substituted.
In S24, it is determined whether a deadline miss will occur in the future. Whether a deadline miss will occur in the future can be determined, for example, based on whether the time obtained in S23 is before or after the current time. If it is determined that a deadline miss will occur in the future, in S25, the name of the process in which the occurrence of a deadline miss is expected is notified to the migration process determination unit 3b via the processing time collection / prediction unit or directly. This operation may be performed without being synchronized with the processing of the process, or may be performed every time the processing of the process is completed.

FIG. 33 is a diagram showing an example of an operation flow of the computer load prediction unit 22 in the present embodiment. The operation will be described with reference to the function explanatory diagram of FIG. In S201, all entries in the computer load history table memory 9 are referred to determine the load acquisition time of each entry and the computer load at that time. In S202, the relation between the computer load and the load acquisition time is expressed by an approximate expression using a statistical technique. As a method of obtaining the approximate expression, the least square method, for example, can be considered as in the deadline miss occurrence time predicting means. In addition, the order of the approximation formula may be determined in advance as in the deadline miss occurrence time prediction means, or after gradually determining the upper limit of the order and approximating the formula closest to the data fluctuation. A method of formulating may be used.
In S203, the time when the CPU usage rate becomes 100% or the determined upper limit value is obtained from the equation obtained in S202. If it is predicted that the CPU usage rate will reach the upper limit, in S205, the time when the CPU usage rate reaches the upper limit and the computer name are notified. Note that the processing of S201 to S205 may be performed in synchronization with the notification of the computer load information from the computer load amount collection unit 5, or may be performed asynchronously.

FIG. 34 is a diagram showing an example of an operation flow of the CPU time prediction unit 23 in the present embodiment. The operation will be described with reference to the function explanatory diagram of FIG. In S211, the processing start time and CPU time in each entry are acquired by referring to all entries in the processing time history table / memory. In S212, the relation between the CPU time acquisition time and the CPU time is expressed by an approximate expression using a statistical technique. As a method of obtaining the approximate expression, the least square method described above can be considered. The order of the approximate expression is processed in the same manner as described above. This operation may be performed either synchronously or asynchronously with the process processing.
FIG. 35 is a diagram illustrating an example of an operation flow of the memory usage amount prediction unit 24 in the present embodiment. In S221, all entries in the computer load history table / memory 9 are referred to, and the load acquisition time in each entry and the memory usage at that time are obtained. In S222, the relationship between the load acquisition time and the memory usage is expressed by the same method as that expressed by the above approximate expression. This processing may also be performed in synchronization with the notification to the computer load amount collection unit 5 or asynchronously. In S223, the time when the memory shortage occurs is obtained. For example, if the formula obtained in S222 is y = a _n x ⁿ + a _n-1 x ^n-1 +... + A ₁ x + b (y: memory usage, x: time) For example, what is necessary is just to obtain | require the value of x at the time of substituting the memory amount mounted on a computer, or the upper limit memory amount determined by the user. If the calculated time is later than the current time, it is predicted that a memory shortage will occur, and in S225, the time when the memory shortage occurs and the computer name thereof are notified to the migration process determination unit 3b.

FIG. 36 is a diagram illustrating an example order of an operation flow in the transition process determination unit 3b according to the present embodiment when a prediction notification of the occurrence of a deadline miss has arrived from the deadline miss occurrence time prediction unit 21.
In S231, the deadline miss occurrence time predicting unit 21 receives a notification notifying that a deadline miss occurs at a time. For example, it is conceivable that the name of the process in which a deadline miss occurs is received along with the notification. When the name of the process in which the deadline miss occurs in S231, the name of the computer that is processing the process in which the deadline miss occurs is acquired by referring to the system configuration information management table memory 7 in S232. In step S233, the name of another process 101 operating on the computer 100 is acquired. In S234, among the processes found in S233, a process in which the processing time of the CPU increases at the predicted time of occurrence of a deadline miss is investigated. A process whose CPU time has increased indicates a process whose processing load has actually increased. Whether or not the CPU time has increased is determined by acquiring the average CPU time from the system configuration information management table because the system configuration information management table holds the average CPU time of each process. Further, it can be determined by inquiring and obtaining the CPU time of the process at the deadline miss occurrence time and comparing these two values. In S235, the resource allocation unit is configured to shift the process predicted to cause a deadline miss and the process found in S234 as at least one of these as candidate processes for resource reassignment. 6b is notified.

FIG. 37 shows an example of the operation flow when each prediction notification arrives from the other operation flow of the migration process determination unit 3b in this embodiment, that is, the memory usage amount prediction unit 24 or the computer load prediction unit 22. It is a figure.
When a notice of occurrence of a deadline miss is received from the memory usage prediction unit 24 or the computer load prediction unit 22 in S241, the computer in which a memory shortage occurs in S242 is referred to the system configuration information management table memory 7 The name of the process 101 operating on 100 is acquired, the process found in S242 is set as a resource reallocation candidate, and these processes are notified to the resource allocation unit 6b.
The operation of the migration process determination unit 3b when receiving a notification from the processing time collection / prediction unit 2b and the computer load amount collection unit 5 is the same as that of the first embodiment.

FIG. 38 is a diagram illustrating an example of an operation flow of the resource allocation unit 6b in the present embodiment.
When the migration process determination unit 3b receives a resource reassignment candidate and a list of resource reassignment process processes in S251, the process name described in the list is used in S252 when a deadline miss occurs, for example. The table is held in order of increasing CPU time. The CPU time when a deadline miss occurs is obtained by inquiring the CPU time prediction unit 23.
In S254, one process name is extracted from the process name table created in S252. In S255, a list of normally operating computers 100 is acquired from the computer load amount collection unit 5, and stored in the table in descending order of load, for example, when a deadline miss occurs. The computer load at the deadline miss occurrence time is obtained by inquiring the computer load prediction unit 22.
In S257, for example, one computer having the highest CPU usage rate is extracted from the table shown in the computer list. In S258, it is confirmed whether or not the process 101 extracted in S254 can operate on the computer 100 selected in S257. As a method of confirming whether or not it can be operated, for example, a method of simply judging based on the CPU usage rate of a computer or an actual investigation of which process occupies the CPU based on the cycle of each process. A method of analyzing whether or not scheduling can be performed without causing a deadline error is conceivable.

If it is determined in S259 that the process extracted from the table in S254 is operable on the computer 100 selected in S257, then in S260, the computer 100 extracted in S257 is free at the deadline miss occurrence time. The amount of memory is checked, and it is confirmed whether or not the amount of memory used by the process 101 extracted in S254 is larger than the amount used. The free memory amount at the deadline miss occurrence time is obtained by inquiring the memory use amount at the deadline miss occurrence time to the memory use amount prediction unit 24, and is further extracted from the system configuration information management table memory 7 in S256. Can be obtained by referring to the amount of installed memory and subtracting the former value from the latter value.
If there is a sufficient amount of free memory, in S262, the computer 100 selected in S257 is determined as the reassignment destination of the process 101 extracted in S254. In step S263, the process 101 extracted in step S254 is started on the computer 100 determined in step S261, and the process operating in the original computer is stopped. In S264, the process extracted from the table in S254 of the system configuration information management table memory 7 and the computer 100 (the computer extracted in S256) that is the allocation destination of the process and the process 101 extracted from the table in S254. The CPU time at the time of occurrence of a deadline miss is rewritten and the process returns to S253, and the processes of S253 to S264 are performed for the process indicated by the next entry in the table created in S252.

In addition, when the dynamic resource control device or system is operating in response to a notification from the migration process determination unit 3b from the deadline miss occurrence time prediction unit 21 or the memory usage amount prediction unit 24, that is, a candidate process group for resource reallocation If a list of processes 101 is given and it is sufficient that resource reallocation can be performed for at least one of the processes shown in the list, resource reallocation is performed for all the notified processes. You may end without.
In S259, when it is determined that the process 101 is processed by the computer 100 selected in S257, a deadline error occurs, the process returns to S256, and S257 and subsequent steps are executed for all the other computers 100 shown in the table created in S255. To do.
In S261, if there is not enough free memory in the computer 100 selected in S257, the process returns to S256, and S257 and subsequent steps are executed for all the computers shown in the table created in S255. In S256, it is checked whether or not all computers can be operated. If unassigned, it is determined that there is no assignable computer, and resource allocation is abandoned, and the own process is stopped in S266. In step S265, the process is deleted from the table, and the processes in and after step S253 are performed for the next process in the table.
Note that the reallocation method shown here is merely an example, and there are many examples in which the operations of S254 to S264 are applied, but other examples are omitted here.

FIG. 39 is a diagram showing an outline of the operation flow of the computer resource dynamic control device or system in the present embodiment when a cooperative operation is performed based on the operation of each component.
In the figure, in S271, the processing time collection unit 2, the computer load amount collection unit 5, the deadline miss occurrence time prediction unit 21, and the memory usage amount prediction unit 24 are configured to generate a deadline miss, a deadline miss occurrence prediction, The migration process determining unit 3b is notified of a computer abnormality, CPU / memory resource shortage, and CPU / memory / resource shortage prediction. Detection of the occurrence of a deadline miss is performed according to the procedures of S11 to S16. Detection of occurrence prediction of a deadline miss is performed in the steps S21 to S25. Computer abnormality detection and CPU / memory resource shortage detection are performed in steps S1011 to S1016. Prediction of the occurrence of a deadline miss is performed according to the procedures of S21 to S25. Prediction of CPU resource shortage is performed in the procedure of steps S201 to S205. Prediction of memory resource shortage is performed by the procedure of steps S221 to S225.
Then, in S272, the migration process determination unit 3b investigates the content of the notification. If the content of the notification is a computer failure, in S274, the resource allocation unit 6b allocates resources to all the processes 101 operating on the failed computer 100 in the procedure of S251 to S265.

  If it is not a computer failure, in S275, the migration process determining unit 3b finds a process that is considered to be the cause of the occurrence of a deadline error or a process that is considered to be the cause of the resource shortage, and in S276, the resource allocation unit 6b Resource allocation is performed according to the procedures of S251 to S265. In S277 after the resource allocation is completed, the resource allocation unit 6b rewrites the identifier of the process that has newly allocated the resource, information on the allocation destination computer 100 on which the process operates, and the like in the system configuration information management table memory 7. .

In other words, in this embodiment, a plurality of computers are connected to a network, and a plurality of processes operate in cooperation or independently on each computer, and each process performs processing at a fixed time period and performs processing within a time limit. In a distributed real-time system that is required to complete the process, it has a processing time monitoring means for monitoring the processing time of the process and a computer load monitoring means for monitoring the resource usage status of the computer. A system that prevents the occurrence of process deadline misses by changing the resources given to the process accordingly, and by adding the following configuration to other computers that have free resources before the deadline miss occurs A computer resource that prevents deadline errors from occurring It shows the control system. That is,
21) A system configuration information management table that holds an average processing time of each process, an amount of consumed memory, a CPU time, a time limit for processing per process, and a list of computers constituting the system.
22) Processing time collecting means for holding a history of fluctuations in processing time and CPU time of each process operating on the system and detecting occurrence of a process deadline miss.
23) A processing time predicting unit that learns the variation in the processing time of the process from the history of the variation in the processing time of each process, and predicts the occurrence and occurrence time of a future deadline miss.
24) Computer load collecting means for holding a history of resource usage of all the computers operating on the system, notifying other means of the usage status of past computer resources, and detecting occurrence of resource shortage.
25) Learning future fluctuations in memory usage based on fluctuations in the memory usage of computers, detecting future memory shortages, and predicting the occurrence of process deadline misses and CPU resource shortages Memory usage amount predicting means for notifying the memory usage status at time.
26) Learning future fluctuations in the CPU usage status from fluctuations in the CPU usage status of the computer, detecting the occurrence of a future CPU resource shortage, and the CPU usage at the time when the memory shortage occurs and the estimated deadline miss occurrence time CPU usage rate predicting means for notifying the situation.
27) A migration process determining unit that determines a process to which resources should be reassigned when a failure such as a deadline miss, CPU resource shortage, or memory resource shortage occurs.
28) Resource allocation means for acquiring a necessary resource amount for the process determined by the migration process means from the system configuration information management table and performing process migration on a computer having the necessary free resources is provided.

In other words, the computer resource dynamic control apparatus according to the present embodiment predicts the usage time of the processor based on the change in the processor usage status of the computer, and predicts the occurrence time of the deadline miss by the current processor. A predictor;
A memory usage amount predicting unit that predicts a necessary amount of the memory based on a change in the memory usage status of the computer,
The computer load amount collection unit has a computer load amount collection / prediction unit that predicts the required amount of resources and calculation function in addition to the collection of computer resource usage information.
The processing time collection unit has a processing time collection / prediction unit that predicts the occurrence time of a deadline miss by the current computer in addition to collecting the processing time of the process,
When the migration process determination unit receives the notification of the necessary amount prediction of the resource and the calculation function or the prediction of the occurrence of the deadline miss, it recommends the reallocation of the process based on the prediction,
The resource allocation unit is configured to cause the candidate computer that has extracted the recommended extracted process to execute allocation based on the prediction of the occurrence time of the deadline miss and the prediction of the required amount of memory.
As described above, in addition to the first embodiment, the occurrence of a deadline miss is predicted by learning from fluctuations in the processing time of the process, and the deadline miss occurrence is predicted to enable resource reallocation. In addition to soft real-time processing of the system, hard real-time processing can also be realized. Also, by allocating resources by predicting the CPU load and memory usage when a deadline miss occurs, it is possible to allocate resources to the computer 100 that is less prone to deadline misses. Deadline miss recurrence can be prevented with high probability.

Embodiment 3 FIG.
A description will be given of a configuration and an operation in which, when a deadline miss is predicted to occur, the process is transferred to another computer earlier to reliably prevent the occurrence of a deadline miss.
The configuration of the computer resource dynamic control system or apparatus in the present embodiment is the same as that in the second embodiment. The processing time monitoring unit 1, processing time collection / prediction unit 2b, computer load monitoring unit 4, computer load amount collection unit 5, processing time history table / memory 8, computer load history table / memory 9, computer according to the third embodiment The functions and operations of the state table memory 10 and the CPU time prediction unit 23 are the same as those in the second embodiment.
An example of the data configuration of the system configuration information management table / memory 7c in the present embodiment is shown in FIG. In the system configuration information management table / memory 7c according to the present embodiment, in addition to the information shown in the system configuration information management table / memory 7 according to the second embodiment, information relating to the resource dynamic control system includes “reassignment time” “Process startup time” is stored as information about the software. The time required for reallocation refers to the notification of deadline miss occurrence, deadline miss prediction, memory shortage prediction, computer failure, etc., until the resource allocation unit 6c completes the resource reallocation after the notification to the migration process determination unit 3c. The time required for this is maintained. This time may be the worst value in the past, or may be an average value in the past if there is no significant change in the allocation completion time. The process start time is the time from when the dynamic control system starts the process 101 to when the process starts processing. This time includes, for example, a time until the process is loaded into the memory and an initialization process specific to the process.

FIG. 41 shows an operation flow of the deadline miss occurrence time prediction unit 21c in the present embodiment. In S31, S21 to S24 are executed. At this time, if it is determined in S24 that a deadline miss will occur in the future, the current time is acquired in S32, and the remaining time until the deadline miss occurrence time is obtained. In S33, the sum of the time required for resource reallocation, the time required for process activation, and the processing cycle time per process is obtained. In S34, the time obtained in S32 is compared with the time obtained in S33. If the time in S32 is short, that is, the remaining time until the occurrence of a deadline miss is the time required for handling, the time required for starting the process, and the processing cycle. If it is shorter than the sum of the times, the name and the current time of the process in which the occurrence of a deadline miss is expected are notified to the migration process determination unit 3c in S35. In S34, when the time of S32 is shorter, that is, when resource reassignment is started immediately after S34, it can be estimated that resource reassignment is completed at a time before one cycle of the time when a deadline miss occurs. Continues monitoring the process time without reallocating resources.
Note that the processing of S31 to S35 may be performed in synchronization with the end of the processing of the process, or may be performed asynchronously with the processing of the process. In this example, whether or not to start resource allocation is determined based on whether or not resource allocation is completed one cycle before the deadline miss occurs. You may decide.

FIG. 42 shows an operation flow of the memory usage prediction unit 24c in the present embodiment.
In S301, S221 to S224 are executed. If it is determined in S224 that a memory shortage will occur in the future, the current time is acquired in S302, and the remaining time until the memory shortage occurs is obtained. In S303, the sum of the time required for resource reallocation, the time required for process activation, and the processing cycle time of the process is obtained. In S304, the time obtained in S302 is compared with the time obtained in S303, and the time obtained in S302 is shorter, that is, the remaining time until the memory shortage occurs is the time required for resource reallocation and the time required for starting the process. If it is shorter than the sum of the processing cycle times, in S305, the time when the memory shortage occurs, its computer, and the current time are notified to the migration process determination unit 3c. In S304, if the time obtained in S302 is shorter, that is, if resource reallocation is completed before one cycle of the time when memory shortage occurs when resource reallocation is performed immediately, Continue monitoring. Note that it is determined whether or not resource allocation is started based on whether or not resource allocation is completed one cycle before the deadline miss occurs. However, it is not necessarily one cycle and can be arbitrarily determined by the user. Good.

FIG. 43 shows an operation flow of the computer load predicting unit 22c in the present embodiment.
In S311, S201 to S204 are executed. If it is determined in S204 that the CPU usage rate is greater than or equal to a certain upper limit value, the current time is acquired in S312 and the remaining time until the CPU usage rate reaches the upper limit value is obtained. In S313, the sum of the time required for resource reallocation, the time required for process activation, and the process cycle time of the process is obtained. In S314, the time obtained in S312 is compared with the time obtained in S313, and the time obtained in S312, that is, the remaining time until the CPU usage rate reaches the upper limit value, the time required for resource reallocation and the process activation required If it is shorter than the time and the interval time of the periodic processing, in S315, the time when the CPU reaches the upper limit value, its computer, and the current time are notified to the migration process determination unit 3c. In S314, the time of S312 is shorter, that is, even if the resource reallocation is performed at the current time, nothing is done if the resource reallocation ends one cycle or more before the time when the CPU resource shortage occurs. Continue monitoring the CPU usage rate.

FIG. 44 shows an operation flow of the migration process determination unit 3c when a prediction notification of deadline miss occurrence is received from the deadline miss occurrence time prediction unit 21c in the present embodiment.
In S321, when the deadline miss occurrence time predicting unit 21c receives the process 101 in which the occurrence of a deadline miss is predicted and the current time acquired in S32, in S322, steps S232 to S234 in the second embodiment are performed. Execute. In S323, the process found in S234 and the process in which the occurrence of a deadline miss is predicted are determined as resource reassignment candidates, and the time sent from the candidate process group and the deadline miss occurrence time prediction unit 21c is determined as a resource. It transmits to the allocation part 6c.
FIG. 45 shows an operation flow of the migration process determination unit 3c when a notification is received from the memory usage prediction unit 24c or the computer load prediction unit 22c in the present embodiment.
In S331, a notification is received from the memory usage prediction unit 24c or the computer load prediction unit 22c. The contents of the notification are the computer in which the abnormality has occurred, the current time acquired by the memory usage prediction unit 24c in S312 and the current time acquired by the computer load prediction unit 22c in S322. In S332, the system configuration information management table / memory 7c is referred to, and a process operating on the computer notified in S331 is acquired. In S333, the process found in S322 is set as a resource reassignment target or a resource allocation candidate process, and the resource allocation unit 6c is notified of these processes and the time received in S321.

FIG. 46 shows an operation flow of the resource allocation unit 6c in the present embodiment.
In S341, from the migration process determination unit 3c, the resource allocation target, the process list as allocation candidates, and the times acquired by the deadline miss occurrence time prediction unit 21c, the computer load prediction unit 22c, and the memory usage prediction unit 24c are received. To do. In S342, S252 to S263 are executed. If it is determined in 253 that the table in S252 is empty or it is not necessary to deal with it, the process proceeds to S343 to acquire the current time. In S344, the time required for resource allocation is obtained by subtracting the time received in S341 from the time obtained in S343. In S345, the required resource allocation time registered in the system configuration information management table / memory 7c is acquired. Then, the time required for the current resource allocation obtained in S344 and the resource allocation time registered in the system configuration information table / memory 7c are obtained. Then, the value of the resource allocation required time in the system configuration information management table memory 7c is rewritten to the value obtained in S344.

FIG. 47 is a diagram showing an outline of the operation flow of the computer resource dynamic control device or system in the present embodiment when a cooperative operation is performed based on the operation of each component described above.
In S351, the processing time collection / prediction unit 2b, the computer load amount collection unit 5, the deadline miss occurrence time prediction unit 21c, and the memory usage amount prediction unit 24c The migration process determination unit 3c is notified of the abnormalities of CPU, memory resource shortage, and CPU / memory resource shortage. Detection of the occurrence of a deadline miss is performed according to the procedures of S11 to S16. Detection of occurrence prediction of a deadline miss is performed in the procedure of S31 to S35. Computer abnormality detection and CPU / memory resource shortage detection are performed in steps S1011 to S1016. Prediction of CPU resource shortage is performed according to the procedures of S31 to S35. Prediction of memory resource shortage is performed according to the procedure of S301 to S305.
Then, in S352, the migration process determination unit 3c investigates the content of the notification. If the content of the notification is a computer failure, in S355, the resource allocation unit 6c allocates resources to all processes operating on the failed computer in the sequence of S341 to S347.
If it is not a computer failure, in S355, a candidate for the process causing the deadline error or the process causing the resource shortage is found, and in S356, the resource allocator 6c allocates resources in the procedure of S341 to S347. I do. After completing the resource allocation, in S357, the resource allocation unit 6c rewrites the identifier of the process to which the resource allocation is newly performed, information on the allocation destination computer on which the process operates, and the like in the system configuration information management table memory 7c.

  As described above, the system configuration information management table / memory 7c holds the worst value of the time required for resource allocation, and the time required for resource allocation and the time for starting the process from the estimated occurrence time of deadline miss, memory shortage, and CPU resource shortage. By starting the resource allocation by the sum of the period processing time and the resource processing, the real-time processing of the process can be continued at least one period before the occurrence of the deadline miss.

That is, this embodiment is characterized in that the process migration is completed immediately before the occurrence of a deadline miss by adding an element having the following characteristics in addition to the computer resource dynamic control system in the previous embodiment. This shows a computer resource dynamic control method. That is,
31) In addition to the system configuration information management table in the previous embodiment, process migration after detecting deadline miss occurrence, deadline miss occurrence prediction, memory shortage, memory shortage prediction, computer failure in each process System configuration information management table that holds the time to complete the process and the time from the completion of process migration to the start of the process.
32) The processing time predicting means in the previous embodiment learns the process time variation from the history of the process time variation of each process, predicts the occurrence and occurrence time of a future deadline miss, Processing time prediction means having a function of notifying the migration process determination means of the occurrence prediction of a deadline miss only before the time required for the process to start processing after completing the migration.
33) In the memory usage prediction means in the previous embodiment, a deadline miss occurrence time or CPU resource shortage occurs by predicting future fluctuations in the memory usage based on fluctuations in the memory usage of each computer. Notifying the memory usage at the time to run, predicting the time when memory shortage will occur, causing the migration process decision means to run out of memory only as long as necessary until the process starts after the process migration is completed Memory usage prediction means with a function to notify the prediction.
34) In the CPU usage rate prediction means in the previous embodiment, a deadline miss occurrence time or memory shortage occurs by predicting future fluctuations in CPU usage based on fluctuations in CPU usage of each computer. CPU usage at the time to be executed is predicted, the time when CPU resource shortage occurs is predicted, and the migration process decision means is short of CPU resources only before the process migration is completed and the process starts processing. CPU usage rate predicting means having a function of notifying the occurrence prediction.

In other words, in the computer resource dynamic control device according to the present embodiment, in addition to the device according to the previous embodiment, the system configuration information table / memory includes a reassignment time required for reassigning a process to a candidate computer and the candidate. Memorize the process start time required to start the process on the computer,
The resource allocation unit is characterized in that the extracted candidate computer allocates and executes the process extracted before the sum of the reassignment required time and the process start required time.
Compared to the second embodiment, it is possible to examine the behavior of changes in application processing time, memory usage, and CPU usage until just before the occurrence of a deadline miss, and to allocate resources based on more accurate system behavior. It can be carried out. As a result, if the behavior changes immediately before and it is found that a deadline miss, memory shortage, and CPU resource shortage do not occur, the processing overhead added with resource allocation can be reduced compared to the second embodiment. it can. In addition, the worst value of resource allocation time is maintained in the system configuration information table, and continuously updated, and resource allocation processing can be completed before a deadline miss occurs by starting resource allocation in advance of the worst time. Increases nature.

Embodiment 4 FIG.
In the present embodiment, a configuration and operation for preferentially processing a process having high importance when resources are insufficient will be described.
The components of the computer resource dynamic control device or the computer resource dynamic control system in the present embodiment are the same as those in the first embodiment. That is, the processing time monitoring unit 1, the processing time collection / prediction unit 2b, the migration process determination unit 3, the computer load monitoring unit 4, the computer load amount collection unit 5, the processing time history table / memory 8, the computer load history table / memory 9, The configuration of the computer state table memory 10 is the same as that of the first embodiment.
The configuration of the system configuration information management table / memory 7d in the present embodiment is shown in FIG. In addition to the information shown in the system configuration information management table / memory 7 of the first embodiment, the system configuration information management table / memory 7d of the present embodiment holds the “importance” of each process. Here, the importance indicates the order in which processes are stopped when the processes are stopped in order to create free resources when there are no more computer resources for operating all the processes constituting the system. In other words, by stopping the processes according to the order shown in the importance level, the more important processes will run even if there are not enough resources, and the less important processes will run when there is not enough computer resources in the entire system. Stop.

An example of the operation flow of the resource allocation unit 6d in the present embodiment is shown in FIG.
In S401, when the migration process determination unit 3 receives a resource allocation candidate and a list of resource allocation target processes 101, the list received in S401 is rearranged in descending order of importance in S402, and the table is displayed in the table. Hold. In S404, the process name having high importance at the top of the table in S402 is extracted. In step S405, the names of the computers 100 that operate normally are stored in the table in descending order of load. In S406, the computer name at the head of the table created in S405 is extracted. In step S407, the CPU usage status of the computer 100 taken out in step S406 is checked, and it is checked whether or not the process 101 selected in step S404 can operate on the computer 100 selected in step S406 without causing a deadline error. If it is determined that the operation can be performed without a deadline mistake, in S409, the free memory amount of the computer 100 extracted in S406 is checked to check whether the process 101 selected in S404 is operable. If it is determined that the operation is possible, in S411, the computer 100 selected in S406 is determined as the migration destination of the process 101 selected in S404, and in S412, the computer 100 whose allocation is determined in S411 is determined. The process 101 selected in S404 is started, and the process selected in S404 that was operating on the original computer is stopped. In S413, the assigned computer and process and the processing time of the current process in the system configuration information management table are rewritten.

If it is determined in S408 that a deadline miss will occur when operated, the computer 100 indicated in the next entry of the table created in S405 is selected in S415. If there is a computer indicated in the next entry, all the computers have not been investigated yet, so the process proceeds to S406, and the processes after S406 are performed for the computer selected in S415.
In S415, if there is no computer shown in the next entry, all the computers 100 are already checked to see if they can be operated, and it is found that none of the computers can be operated. This indicates that there is no free computer resource. In this case, the process proceeds to S417, and it is checked whether there is a process having a lower importance than the process 101 in S404. If there is a process, the process proceeds to S419, and the process having the lowest importance among the operating processes is stopped. In S420, if there is a stopped process in the table created in S402, the entry is deleted. A computer resource is vacant because a process of low importance stops. Then, the process proceeds to S405, and the resources that can be allocated to the process 101 are examined again by rearranging in the order of the computer load.
In S417, when there is no process having a lower importance than the process currently being investigated, the process is terminated by stopping the own process in S418. If the process reallocation is completed in S413 or the process is stopped in S418, the first entry from the table, that is, the process name selected in S404 is deleted in S414. For the process indicated in the next entry, the processing from step S404 is repeated.

FIG. 50 is a diagram showing an outline of the operation flow of the computer resource dynamic control device or system when all the components operate in cooperation in this embodiment.
In S421 in the figure, the processing time collection / prediction unit 2b or the computer load amount collection unit 5 notifies the migration process determination unit 3 of the occurrence of a deadline miss or a computer abnormality. The operation of the processing time collection / prediction unit 2b in this case is as shown in S11 to S16, and the operation of the computer load amount collection unit 5 is as shown in S1011 to S1018. Then, in S422, the migration process determination unit 3 investigates the cause of the deadline miss. In S423, it is investigated whether or not the cause is a failure of the computer. If the failure is a computer, the resource allocation unit 6 applies to all the processes 101 operating on the failed computer 100 in S424. Resources are allocated according to the procedures of S401 to S420. At the time of resource allocation, if there is no available resource that can be assigned to any computer, the resources 101 are stopped in order from the least important process 101 to create an available resource, and resource allocation is attempted again.
If it is not a computer failure, in S425, the migration process determination unit 3 finds a process that causes a deadline miss, a process that causes a resource shortage, and a process that performs resource reallocation. In S426, the resource allocation unit 6 allocates resources to the process that causes the deadline miss or the process that is considered to be the cause of the resource shortage in the procedure of S401 to S420. At the time of resource allocation, if there is no available resource that can be assigned to any of the computers 100, the available resources are created by stopping in order from the process with the lower importance, and resource allocation is attempted again. In S427, after the resource allocation is completed, the resource allocation unit 6 identifies the identifier of the process 101 that has newly allocated the resource in the system configuration information management table memory 7d, the allocation destination computer 100 in which the process operates, Rewrite information related to CPU time required for.

That is, this embodiment gives priority to important processing when the system does not have enough resources to operate all processes by adding the following features to the computer resource dynamic control system of the first embodiment. This shows a computer resource dynamic control system that enables real-time processing to be continued. That is,
41) In addition to the system configuration table of the first embodiment, a system configuration information management table that holds the order of processes to be preferentially operated when there are not enough system resources to operate all processes.
42) In the resource allocation unit according to the first embodiment, when there is no free resource for the process determined by the migration process determination unit to operate, the rank of the rank is determined based on the process operation order of the system configuration information management table. Resource allocation means for creating a free resource in a computer by stopping a low process and performing process migration on the computer is provided.

In other words, the computer resource dynamic control device in the present embodiment, in addition to the device in the first embodiment, the system configuration information table memory stores the processing priority of the process,
The resource allocating unit refers to the system configuration information table / memory when the occurrence of a deadline miss is predicted in the current computer, and assigns a process having a higher process priority to the candidate computer to execute the process. And
As described above, when a plurality of independent processes are operating in the system, even when the system resources for operating all the processes 101 are lost due to a computer failure or the like, the process is given importance. By stopping the less important processes, it is possible to create a free space in the system resources, and to continue the operation with priority given to the processes with higher importance.

Embodiment 5 FIG.
In the present embodiment, a description will be given of an apparatus and a system that can perform allocation in consideration of the performance difference when the performance of each computer 100 varies.
The components of the dynamic resource control apparatus or system in the present embodiment are the same as those in the first embodiment. In this embodiment, that is, processing time monitoring unit 1, processing time collection / prediction unit 2b, migration process determination unit 3, computer load monitoring unit 4, computer load amount collection unit 5, processing time history table / memory 8 or computer state The configuration of the table memory 10 is the same as that of the first embodiment.
FIG. 51 shows an example of information held in the system configuration information management table / memory 7e in the present embodiment. In the system configuration information management table / memory 7e, in addition to the information shown in the system configuration information management table / memory 7 in the first embodiment, “operable computer name” for each application as the process 101 and “performance” for each computer 100 are displayed. Value "is shown. As the performance value, it is conceivable that benchmark program data indicating the performance value of the computer, such as SPECInt, is shown.

An example of the operation flow of the resource allocation unit 6e in the present embodiment is shown in FIG.
In FIG. 5, when a resource allocation candidate and a list of processes to be allocated are received from the migration process determination unit 3 in S501, the list of processes 101 is arranged in order of increasing CPU time, for example, and stored in a table in S502. To do. If the table created in S502 is empty, the process ends.
If it is not empty, one name of the process 101 is extracted from the table created in S502 in S504. Then, the computers 100 in which the process 101 of S504 is operable and operating normally are rearranged in the descending order of the computer load, for example, and held in a table. A list of computers that can operate each process is shown in the system configuration information management table memory 7e, and the identifier of the computer is obtained by referring to the list. Then, S507 and the subsequent steps are performed for all the computers in the table created in S505 or until the computer 100 as the resource allocation destination is determined. In S507, one computer is selected from the table created in S505. In S508, the performance value of the computer 100 in S507 and the performance value of the currently operating computer are acquired with reference to the system configuration information management table memory 7e. In step S509, the current CPU time of the process 101 selected in step S504 is corrected based on the value obtained in step S508. For example, when the performance value of the currently operating computer is 100 and the performance value of the computer selected in S507 is 85, the corrected CPU time is 100/85 times. In S510, based on the corrected value, it is confirmed whether or not the process selected in S504 can operate without causing a deadline miss on the computer selected in S507.

If it is operable, the process proceeds to S512, and it is confirmed whether or not the computer selected in S507 has enough free memory to operate the computer selected in S504. If there is a sufficient amount of free memory, in S514, the computer selected in S507 is determined as the resource allocation destination computer. In step S515, the process 101 selected in step S504 is started on the computer 100 determined in step S514, and the processes that have been operating so far are stopped. In S516, the “computer in operation” column and the “average CPU time” column in the system configuration management table memory 7e are rewritten to the current values. In S517, the process name to which the resource reallocation has been performed is deleted from the table created in S504, and the process returns to S503, and the processes after S503 are performed on the remaining processes in the table created in S502.
In S506, if S507 and subsequent steps are performed for all the computers shown in the table of S505, it means that no assignable computer was found. Therefore, the resource reallocation of the process selected in S504 is abandoned, and the process goes to S518. The process 101 is stopped and the stopped process name is deleted from the table created in S502.
In S511 and S513, if the process selected in S504 is not operable, the process returns to S506, and the processing of S507 and subsequent steps is performed on the computer of the next entry.

FIG. 53 shows an example of the operation flow of the dynamic resource control apparatus or the entire system performed in cooperation with each component in this embodiment.
In S521, the processing time collection / prediction unit 2b or the computer load amount collection unit 5 detects the occurrence of a deadline miss or the abnormality of the computer, the CPU, or the memory resource shortage, and notifies the migration process determination unit 3 of the detection. . The operation of the processing time collection / prediction unit 2b in this case is as shown in S11 to S16, and the operation of the computer load amount collection unit 5 is as shown in S1011 to S1018. Then, in S522, the migration process determination unit 3 investigates the content of the notification. If the content of the notification is a computer failure, in S524, the resource allocation unit 6e allocates resources to all the processes 101 operating on the failed computer by the procedures of S501 to S518. If it is not a computer failure, in S525, the migration process determination unit 3 finds a process that is considered to be a cause of occurrence of a deadline miss and a process that is considered to be a cause of occurrence of resource shortage. In S146, the resource allocation unit 6e allocates resources to the process found in S525 according to the procedures in S501 to S518. After the resource allocation is completed, in S527, the resource allocation unit 6e performs processing in the system configuration information management table / memory 7e for the identifier of the process 101 that has newly allocated the resource, the allocation-destination computer 100 in which the process operates, and processing. Rewrite information about CPU time required.

In other words, the present embodiment adds the following features to the computer resource dynamic control system of the first embodiment, so that a computer that does not cause a deadline error even when the performance of the computers constituting the system is uneven. The computer resource dynamic control system is characterized by enabling the process migration. That is,
51) A system configuration information management table that holds a list of computers that can operate each process and a value indicating the performance of each computer in addition to the system configuration information management table of the first embodiment.
52) In the resource allocating unit of the first embodiment, a computer capable of operating the process determined by the migration process determining unit is set as a computer candidate as a process migration destination, and the performance value of the computer included in the system configuration information management table is used. Resource allocation means is provided for performing process migration on a computer having sufficient free resources by calculating computer resources required by the process when the process is transferred to the computer candidate.

In other words, in the computer resource dynamic control device according to the present embodiment, the system configuration information table memory stores an identifier of an operable computer capable of processing the target process and a performance value of the computer,
When the resource allocation unit receives the reassignment recommendation notification, the resource allocation unit refers to the system configuration information table / memory, checks the identifier of the operable computer and the performance value of the computer, and allocates the process to the candidate computer for execution. It is characterized by.
As described above, the system configuration information management table / memory 7e indicates the computer name in which each process 101 can operate and the performance value of the computer, so that the OS, CPU architecture, and performance value of the computers in the system are different. Even in such a case, it is possible to correctly determine an operable computer, correct CPU time, etc., and assign the computer to a computer that does not cause a deadline error.

Embodiment 6 FIG.
A configuration and operation for preventing the occurrence of a deadline miss by migrating a computer that processes a corresponding process within a short time even when the deadline miss occurrence prediction time is approaching will be described. Specifically, assuming that the computer to which such a process should be transferred is activated, the process is put on standby.
The configuration of the resource dynamic control system in this embodiment is shown in FIG. Further, FIG. 55 shows the configuration of the computer resource dynamic control device 100c in the case of being combined as one device. The components in the dynamic resource control system or apparatus in the present embodiment have a replica management unit 25 in addition to the configuration in the third embodiment.
Other processing time monitoring unit 1, processing time collection unit 2 (or processing time collection / prediction unit 2b), transition process determination unit 3, computer load monitoring unit 4, computer load amount collection unit 5 (or computer) in this embodiment Load collection / prediction unit 5b), resource allocation unit 6, system configuration information management table / memory 7, processing time history table / memory 8 or computer state table / memory 10, deadline miss occurrence time prediction unit 21, computer load prediction The configurations of the unit 22, the CPU time prediction unit 23, and the memory usage amount prediction unit 24 are the same as those in the third embodiment.

Here, the process replica is a process copied on the premise of migration that is activated and waiting on a computer that newly processes a process that is expected to be migrated. Therefore, it is a process that has already started, but is waiting for processing because, for example, the data to be processed has not yet arrived due to the transition. Further, it is assumed that the replica uses memory resources but does not use CPU resources, that is, the replica CPU time is zero.
The replica management unit 25 in the present embodiment refers to the data held by the computer load amount collection unit 5 and activates the replica on the computer 100 having sufficient free memory. When it is predicted that a deadline miss or the like will occur, when the resource allocation unit 6 reallocates resources to the process, the resource allocation unit 6 is notified of which computer 100 the process replica is operating on. Have a role. Also, when a memory shortage is predicted on the computer on which the replica operates, the replica of the process is stopped.

Examples of the operation flow of the replica management unit 25 in this embodiment are shown in FIGS. FIG. 56 shows a procedure for starting a replica at the time of starting the system. In S61, one process for operating the replica is selected. As a criterion for selecting a process, for example, the order in which the memory usage is large may be considered. In general, the larger the memory usage, the longer it takes for the process to start up, so it can be considered that it takes more time to complete the process reallocation. As another method, a method in which a processing cycle is short and a time required for resource allocation needs to be shortened is selected. In S 62, the memory usage of all the computers 100 is acquired from the computer load amount collecting unit 5. In step S63, a process replica is activated on a computer having a free memory capacity sufficient for the process to operate. In S63, with regard to the number of replicas to be activated, the replicas may be activated on all computers, for example, several replicas may be activated in the descending order of the amount of free memory, or one computer may be activated. It may be activated only on the top.
FIG. 57 is a diagram showing an operation by the replica management unit 25 when a memory of a certain computer 100 is insufficient during system operation. The operation of the memory usage prediction unit 24 when a memory shortage occurs will be described separately. When a replica stop request for a certain computer is received from the memory usage prediction unit 24 in S601, the replica operating on the specified computer 100 is stopped in S602. Here, all the replicas to be stopped may be stopped, may be stopped in descending order of memory usage, or may be stopped in order of increasing memory usage. In S603, the amount of memory used by a computer other than the computer whose replica has been stopped is acquired from the computer load amount collection unit 5, and in S604, the replica 100 is stored in the computer 100 having a memory free capacity sufficient for operation. Restart. S604 is executed for all replicas stopped in S602.

  FIG. 58 shows an operation flow of the memory usage prediction unit 24 in the present embodiment. In S611, S221 to S224 are executed. If a memory shortage is predicted, a check is made to see if a process replica operates on the computer 100 where memory shortage is likely to occur. If it is operating, the process proceeds to S613, and the replica management unit 25 is stopped. Request and stop replica. In S614, the time when the memory shortage occurs is obtained again. In this case, the formula for obtaining the time when the memory shortage occurs is obtained by subtracting only the memory usage amount used by the stopped replica from the formula obtained in S222. For example, if the equation obtained in S222 is y = ax + b (y: processing time, x: time) and the sum of the memory usage used by the stopped replica is c, the equation is y = Ax + b-c. By giving the amount of memory mounted on the computer to y in this equation, the time when memory shortage occurs is obtained. In S615, the remaining time until the memory shortage occurs is determined from the time when the memory shortage occurs in S223 or S614 and the current time. In step S616, it is checked whether a replica of a process operating on the computer is operating on another computer. Whether or not the replica is operating can be determined by inquiring the replica management unit 25. If the replica is operating on another computer, in S617, the sum of the resource reallocation time and one cycle processing time is obtained. If the replica is not operating, the sum of the time required for replica reassignment, the time required for process activation, and the time for one periodic process is obtained. Then, the time obtained in S617 or S618 is compared with the time obtained in S615, and if the time obtained in S615 is shorter, the process in which a deadline miss occurs in the deadline miss occurrence time prediction unit 21 in S620. And notify the current time.

FIG. 59 shows an example of the operation flow of the resource allocation unit 6f in the present embodiment. In S621, when the migration process determination unit 3 receives a resource reassignment candidate and a list of processes 101 to be reassigned, the process is rearranged in the descending order of the CPU time and stored in the table. Then, the operations after S624 are executed until the table becomes empty.
In S624, one process name is extracted from the table in S622. In S625, the identifiers of the computers 100 on which the replicas of the process in S624 operate are obtained from the replica management unit 25, arranged in the descending order of CPU usage, and stored in the table. Hold. Then, S627 and subsequent steps are executed for all the computers shown in the table created in S625. First, in S627, the computer name is extracted from the table created in S625. Then, it is confirmed whether or not the process extracted in S624 can operate on the computer extracted in S627. If it is operable, in S631, the computer selected in step S627 is determined as the assignment destination, and in S632, the process is started on the computer determined in S631, and the process that was operating on the original computer is stopped. . In step S633, information on the process, computer, and current processing time in the system configuration information management table / memory 7 is rewritten. In step S634, a request is made to activate a replica of the process extracted in step S624. In S636, the name of the process extracted in S624 is deleted from the table, another process is selected from the table, and S623 and subsequent steps are performed. In S629, if the computer selected in S627 does not have enough free CPUs, the process returns to S627, and S627 to S629 are executed for the other computers. If S627 to S629 are executed for all the computers in the table created in S625 and the allocation destination is not determined, the resource reallocation is abandoned, the own process is stopped in S63, and S624 is performed in S636. The process selected in step S3 is deleted from the table, and S623 and subsequent steps are performed for other processes. When the replica is not operating, the operation is performed in the same procedure as in the third embodiment.

FIG. 60 shows an overall operation flow of the dynamic resource control system or apparatus in the present embodiment.
In S651, the processing time collection unit 2, the computer load amount collection unit 5, the deadline miss occurrence time prediction unit 21, and the memory usage amount prediction unit 24 determine that a deadline miss has occurred, a deadline miss has occurred, and the computer has an abnormality. CPU / memory resource shortage and CPU / memory resource shortage prediction are notified to the migration process determining unit 3. Detection of the occurrence of a deadline miss is performed in steps S11 to S16, and detection of the occurrence prediction of a deadline miss is performed in steps S31 to S35. Computer abnormality detection and CPU / memory resource shortage detection are performed in steps S1011 to S1016, and CPU resource shortage prediction is performed in steps S31 to S35. Prediction of memory resource shortage is performed according to the procedure of S301 to S305. Then, in S652, the migration process determination unit 3 investigates the content of the notification. If the content of the notification is a computer failure, in S655, the resource allocation unit 6f allocates resources to all processes operating on the failed computer in the procedure of S341 to S347.
If it is not a computer failure, in S655, the resource allocation unit 6f allocates resources in the steps S621 to S634 with respect to the process that is considered to be the cause of the deadline error occurrence and the process that is considered to be the cause of the resource shortage. Do. In S656, after the resource allocation is completed, the resource allocation unit 6f rewrites the identifier of the process to which the resource has been newly allocated and information on the allocation destination computer in which the process is operating, in the system configuration information management table memory 7. .

In other words, the computer resource dynamic control system according to the present embodiment adds the following features to the computer resource dynamic control system expression according to the third embodiment, so that even when the remaining time until the deadline miss occurs is short, A computer resource dynamic control system characterized by preventing occurrence of a line miss will be described. That is,
61) Referring to data stored in the computer load collection means, a process copy is started in advance on a computer having a sufficient amount of free resources, and the computer on which the process copy performing the process migration operates is notified to the resource allocation means. Replica management means.
62) In the resource allocation unit of the third embodiment, a computer on which a copy of the process determined by the migration process determination unit is operating is set as a computer candidate to be a process migration destination, and has sufficient free CPU resources among the computer candidates. Resource allocation means for performing process migration on the computer is provided.

In other words, in addition to the apparatus of the third embodiment, the computer resource dynamic control apparatus according to the present embodiment starts the process of a predetermined process for the selected computer and terminates the process of the process. With a replica management unit that manages the state,
When the resource allocation unit receives the reassignment advisory notice, the resource allocation unit causes the operable computer to execute the process allocation if the selected computer is activated by the replica management unit and is operable. Features.
In this way, the replica management unit 25 activates a replica of each process in advance and makes it stand by, so that the time required from the start of resource reallocation to the actual processing of the process. Even when the remaining time until the occurrence of a deadline miss or memory shortage is short, the migration process can be completed in a short time, and the hard real-time process of the system can be continued.

Embodiment 7 FIG.
When a process constitutes part of pipeline processing, processing is affected before and after the stage of pipeline processing, and thus a restriction is added. The occurrence of a deadline miss in such a case has a wide range of effects, and appropriate cause elimination is necessary. In this embodiment, the transition in such pipeline processing will be described.
FIG. 61 shows an example of a processing form of a process to be applied by the computer resource dynamic control system in the present embodiment. As shown in the figure, it is assumed that a plurality of processes are connected to one input (hereinafter referred to as pipeline processing), and each process performs processing one after another to complete one cycle processing. In addition, the system is configured by the existence of a plurality of independent pipeline processes. Each process 101 constituting the pipeline processing may operate on the same computer 100 or may operate on a different computer. When processing is completed in a single process, such as the process to be applied in Embodiments 1 to 6, pipeline processing with a pipeline length of 1 is performed in this embodiment. ing.
The configuration of the resource dynamic control system in the present embodiment is the same as that in the fifth embodiment.
Processing time monitoring unit 1, computer load monitoring unit 4, computer load amount collection unit 5, resource allocation unit 6, processing time history table / memory 8, computer load history table / memory 9, computer state table / memory in the present embodiment The role and configuration of 10 are the same as those of the first embodiment, and detailed description thereof is omitted.

An example of the data configuration of the system configuration information management table / memory 7g in the present embodiment is shown in FIG. In addition to the information shown in the first embodiment, the system configuration information management table memory 7g according to the present embodiment includes “pipeline process name” and “process constituting the pipeline process” as “information related to pipeline process”. , “Pipeline processing deadline” and “Pipeline processing cycle” are shown. For example, in FIG. 62, there is a process group constituting pipeline processing called pipeline 1, and the pipeline processing is realized by the process A, process B, and process C transmitting data in order and performing processing. It is shown that the time from the start of the process A to the next process after the process from the process A to the process C is completed within one second is 200 milliseconds.
In addition to the role shown in the first embodiment, the processing time collection unit 2 in the present embodiment processes the process start time of the process that performs the first process of the pipeline process and the process of the process that performs the last process of the pipeline process. It plays a role of obtaining the time required for one pipeline process from the end time. Further, the processing time history of all processes held in the processing time history table memory 8 is referred to and the number of times that the pipeline processing that constitutes the system causes a deadline miss is investigated. When the deadline miss has occurred more than the allowable number of times, it has a role of notifying the migration process determining unit 3 that the deadline miss has occurred more than the prescribed number in the pipeline processing. The number of deadline misses in pipeline processing is obtained by obtaining the processing start time of the process that performs the first processing of the pipeline processing and the processing end time of the process that performs the last processing of the pipeline processing, subtracting both values, It is obtained by comparing with the value of “pipeline processing deadline” shown in the system configuration information management table. This is performed for all data shown in the processing time history table.

The operation flow of the processing time monitoring unit 1, the computer load monitoring unit 4, and the computer load amount collecting unit 5 in the present embodiment is the same as that of the first embodiment.
The operation of the processing time collection unit 2g in the present embodiment is shown in FIG. In S71, data is received from the processing time monitoring unit 1. In S72, data is written in the processing time history table memory 8. If the data written in S72 is the process completion time of the process that performs the process at the end of the pipeline process, the process proceeds to S74, and the process start time in the same process cycle of the process that performed the first process of the pipeline process is obtained. Similarly, in S75, the processing completion time in the same processing cycle of the process that performed the last processing of the pipeline processing is obtained. In S76, the processing time for one pipeline process is obtained from S74 and S75, and it is checked whether or not a deadline miss has occurred. If a deadline miss has occurred, check how many times the deadline miss has occurred in the data remaining in the processing time history table memory 8, and if a deadline miss has occurred more than the specified number of times , Notify the migration process determination unit 3. In step S79, one entry in the processing time history table is recommended first in order to store the processing data of the next cycle in each process. In S73, if the written data is not the data of the process that performs the process at the end of the pipeline process, or if no deadline miss has occurred in S76, the process proceeds directly to S79.

FIG. 64 shows the operation of the migration process determination unit 3g in the present embodiment.
When a deadline miss notification for pipeline processing is received from the processing time collection unit 2 in S701, a list of process names constituting the pipeline processing in which the deadline miss has occurred is acquired in S702. In step S703, a process having a significantly increased processing time is found for all the processes acquired in step S702. The found process is regarded as a process that causes a deadline error. There may be multiple processes that cause a deadline error. For all processes found in S703, S704 and subsequent steps are executed. In step S705, one of the processes found in step S703 is selected. In step S706, a computer on which the process operates is found. In step S707, another process operating on the computer found in step S704 is found. In step S708, it is checked whether the computer found in step S706 operates normally. Are candidates for resource reallocation. In this case, it is only necessary that resource reallocation can be performed for either the process having the largest increase in CPU time in S707 or the process in S705. In step S710, the resource allocation unit 6 is notified. If the computer is not operating normally in S708, the process proceeds to S711, and all processes found in S707 are subject to resource reallocation and are notified to the resource allocation unit 6.

FIG. 65 shows an example of the operation flow of the dynamic resource management system or apparatus that operates in cooperation with all the components in the present embodiment.
In S711, the processing time collection unit 2g or the computer load amount collection unit 5 notifies the migration process determination unit 3 of the occurrence of a deadline miss in pipeline processing, a computer failure, a memory shortage, and the like. The processing time collection unit 2g detects the occurrence of a deadline miss in the steps S71 to S79, and the computer load amount collection unit 5 detects a computer failure and a memory / CPU resource shortage in the steps S1011 to S1019.
Then, in S712, the migration process determining unit 3g analyzes the content of the notification, and if the cause is a computer failure, the process proceeds to S716, and the resource allocation unit 6 operates on the failed computer in the procedure of S121 to S136. Restart the process that was running on another computer. If the cause is not a failure, the migration process determination unit 3g finds a process that is considered to be the cause of the occurrence of a deadline miss or a resource shortage in steps S701 to S710 in S714. The allocator 6 reallocates resources according to the procedure from S121 to S136. In step S717, the information on the computer to which the resource reallocation has been performed is rewritten in the system configuration table memory 7g.

In other words, the computer resource dynamic control method of the present embodiment adds the following features to the computer resource dynamic control method of the first embodiment, so that a plurality of processes are connected in a pipeline, and each process is sequentially performed. When a deadline time is given to a process in which one cycle process is completed by performing the process at a time, a process migration that causes a deadline occurs when a deadline miss occurs This is a computer resource dynamic control system characterized by preventing the continuation of a deadline miss in the processing of this shape by performing the above. That is,
71) In addition to the system configuration information management table of the first embodiment, a system configuration information management table that holds a process group that performs one periodic process, a periodic processing time, a processing time limit, and a processing name.
72) A processing time collecting means in the processing time collecting means of the first embodiment, which obtains the time required for the entire series of processes constituted by a plurality of processes and notifies the transition process determining means when a deadline miss occurs.
73) When the notification is received from the processing time collecting unit in the migration process determining unit of the first embodiment, the cause of the occurrence of the deadline miss is investigated by examining the variation in the processing time of all the processes constituting the series of processes. A transition process determining means for obtaining the process and reporting the process to the resource allocating means.

In other words, the computer resource dynamic control device in the present embodiment, in addition to the device in the first embodiment, the system configuration information table memory stores the process name and deadline time constituting the pipeline processing,
When the migration process decision unit receives the occurrence or prediction of a deadline miss, it refers to the system configuration information table / memory to identify the pipeline process that causes the occurrence and makes a recommendation. Features.
As described above, according to the present embodiment, when a deadline is given to a pipeline process constituted by a plurality of processes, even when a deadline miss occurs in the pipeline process, the cause is removed and the deadline is removed. Mistakes can be resolved.
It should be noted that the above embodiments may be combined, and in that case, an effect obtained by combining the features described in each embodiment can be obtained.

It is a figure which shows the structure of the computer resource dynamic control system in Embodiment 1 of this invention. 2 is a diagram illustrating a configuration of a computer resource dynamic control device according to Embodiment 1. FIG. 6 is a diagram for explaining a function of a processing time monitoring unit according to Embodiment 1. FIG. 6 is a diagram for explaining a function of a processing time collection unit according to Embodiment 1. FIG. 6 is a diagram for explaining a function of a processing time collection unit according to Embodiment 1. FIG. 6 is a diagram for explaining a function of a processing time collection unit according to Embodiment 1. FIG. 6 is a diagram for explaining a function of a migration process determination unit in the first embodiment. FIG. 3 is a diagram for explaining functions of a computer load monitoring unit in Embodiment 1. FIG. FIG. 3 is a diagram for explaining a function of a computer load amount collection unit in the first embodiment. FIG. 3 is a diagram for explaining a function of a computer load amount collection unit in the first embodiment. FIG. 3 is a diagram for explaining a function of a computer load amount collection unit in the first embodiment. FIG. 3 is a diagram for explaining a function of a computer load amount collection unit in the first embodiment. FIG. 3 is a diagram for explaining a function of a resource allocation unit in the first embodiment. 6 is a diagram illustrating a data configuration example of a system configuration information management table in the first embodiment. FIG. 6 is a diagram illustrating a data configuration example of a processing time history table in the first embodiment. FIG. 6 is a diagram illustrating a data configuration example of a computer load history table according to the first embodiment. FIG. 3 is a diagram illustrating a data configuration example of a computer state table according to Embodiment 1. FIG. 6 is a diagram showing an operation flow of a processing time monitoring unit in the first embodiment. FIG. 6 is a diagram illustrating an operation flow of a processing time collection unit according to Embodiment 1. FIG. 6 is a diagram illustrating an operation flow of a computer load monitoring unit according to the first embodiment. FIG. FIG. 6 is a diagram illustrating an operation flow of a computer load amount collection unit according to the first embodiment. 6 is a diagram illustrating an operation flow of a migration process determination unit according to Embodiment 1. FIG. 6 is a diagram illustrating an operation flow of a migration process determination unit according to Embodiment 1. FIG. 6 is a diagram illustrating an operation flow of a resource allocation unit in Embodiment 1. FIG. FIG. 3 is a diagram showing an overall operation flow performed by the computer resource dynamic control device and system in the first embodiment. It is a figure which shows the structure of the computer resource dynamic control system in Embodiment 2 of this invention. It is a figure which shows the structure of the computer resource dynamic control apparatus in Embodiment 2. FIG. It is a figure explaining the function of the deadline miss occurrence time prediction part in Embodiment 2. FIG. It is a figure explaining the function of the computer load estimation part in Embodiment 2. FIG. FIG. 10 is a diagram for explaining a function of a CPU time prediction unit in the second embodiment. FIG. 10 is a diagram for explaining a function of a memory usage amount prediction unit in the second embodiment. FIG. 10 is a diagram showing an operation flow of a deadline miss occurrence time prediction unit in the second embodiment. It is a figure which shows the operation | movement flow of the computer load estimation part in Embodiment 2. FIG. FIG. 10 is a diagram showing an operation flow of a CPU time prediction unit in the second embodiment. FIG. 10 is a diagram illustrating an operation flow of a memory usage amount prediction unit in the second embodiment. FIG. 10 is a diagram illustrating an operation flow of a migration process determination unit in the second embodiment. FIG. 10 is a diagram illustrating an operation flow of a migration process determination unit in the second embodiment. FIG. 10 is a diagram showing an operation flow of a resource allocation unit in the second embodiment. It is a figure which shows the comprehensive operation | movement flow which the computer resource dynamic control apparatus and system in Embodiment 2 perform. It is a figure which shows the example of a data structure of the system configuration information management table in Embodiment 3 of this invention. FIG. 10 is a diagram showing an operation flow of a deadline miss occurrence time prediction unit in the third embodiment. FIG. 10 is a diagram illustrating an operation flow of a memory usage amount prediction unit in the third embodiment. FIG. 10 is a diagram showing an operation flow of a computer load prediction unit in the third embodiment. FIG. 10 is a diagram showing an operation flow of a migration process determination unit in the third embodiment. FIG. 10 is a diagram showing an operation flow of a migration process determination unit in the third embodiment. FIG. 10 is a diagram showing an operation flow of a resource allocation unit in the third embodiment. FIG. 10 is a diagram showing an overall operation flow performed by the computer resource dynamic control device and system in the third embodiment. It is a figure which shows the example of a data structure of the system configuration information management table in Embodiment 4 of this invention. FIG. 20 is a diagram illustrating an operation flow of a resource allocation unit in the fourth embodiment. FIG. 20 is a diagram showing an overall operation flow performed by the computer resource dynamic control device and system in the fourth embodiment. It is a figure which shows the example of a data structure of the system configuration information management table in Embodiment 5 of this invention. FIG. 20 is a diagram showing an operation flow of a resource allocation unit in the fifth embodiment. FIG. 20 is a diagram showing an overall operation flow performed by the computer resource dynamic control device and system in the fifth embodiment. It is a figure which shows the structure of the computer resource dynamic control system in Embodiment 6 of this invention. FIG. 20 is a diagram illustrating a configuration of a computer resource dynamic control device according to a sixth embodiment. FIG. 20 is a diagram showing an operation flow of a replica management unit in the sixth embodiment. FIG. 20 is a diagram showing an operation flow of a replica management unit in the sixth embodiment. FIG. 20 is a diagram illustrating an operation flow of a memory usage amount prediction unit in the sixth embodiment. FIG. 20 is a diagram showing an operation flow of a resource allocation unit in the sixth embodiment. FIG. 20 is a diagram showing an overall operation flow performed by the computer resource dynamic control device and system in the sixth embodiment. It is a figure which shows the example of the process which the computer resource dynamic control system in Embodiment 7 of this invention makes object. FIG. 20 is a diagram showing a data configuration example of a system configuration information management table in the seventh embodiment. FIG. 23 is a diagram showing an operation flow of a processing time collection unit in the seventh embodiment. FIG. 23 is a diagram showing an operation flow of a migration process determination unit in the seventh embodiment. FIG. 20 is a diagram showing an overall operation flow performed by the computer resource dynamic control device and system in the seventh embodiment.

Explanation of symbols

  1 processing time monitoring unit, 2 processing time collection unit, 2b processing time collection / prediction unit, 3 migration process decision unit, 4 computer load monitoring unit, 5 computer load collection unit, 5b computer load collection / prediction unit, 6 resources Allocation unit, 7 system configuration information management table / memory, 8 processing time history table / memory, 9 computer load history table / memory, 10 computer state table / memory, 11 processor (CPU), 12 memory, 13 input device, 14 output Apparatus, 15 communication interface, 16 internal bus, 21 deadline miss occurrence time prediction unit, 22 computer load prediction unit, 23 CPU time prediction unit, 24 memory usage prediction unit, 25 replica management unit, 100 computer, 100a, 100b, 100c Computer resource dynamic controller, 101 process.

Claims (2)

The computer has a processor and a memory connected multiple, a process performed repeatedly at a predetermined cycle by a processor in which a predetermined computer has among the plurality of computers, other than the predetermined computer of the plurality of computers In a computer resource dynamic control device that controls execution of other computers,
The process is executed with a deadline time that represents the limit of the processing time from the start to the end of execution of the process, and a processing time that exceeds the deadline time when the process is repeatedly executed at a fixed period. A system configuration information table memory that pre- stores an allowable number of times allowed to be performed and a consumed memory amount representing an amount of memory consumed when the process is executed ;
A processing time history table memory that stores at least one processing start time at which execution of the process is started and processing end time at which execution of the process is ended as processing time history information;
The processing start time and the processing end time output when the predetermined computer finishes executing the process are input from the predetermined computer, and the input processing start time and processing end time are input to the processing time history. Store as processing time history information in table memory,
In response to each processing time history information of one or more processing time history information stored in the processing time history table memory, a processing time obtained by subtracting the processing start time from the processing end time of the processing time history information is obtained.
By comparing the processing time corresponding to each processing time history information obtained and the deadline time stored in the system configuration information table memory, the number of processing times exceeding the deadline time is obtained,
The obtained number exceeds the allowable number of times stored in the system configuration information table memory, and the processing time corresponding to the latest processing time history information among the one or more processing time history information is set to the deadline time. If exceeded, a processing time collection unit that notifies that the number of processing times exceeding the deadline time exceeds the allowable number of times, and
When the processing time collection unit is notified that the number of the processing times exceeding the deadline time exceeds the allowable number of times, a transition process determination unit that determines to cause the other computer to execute the process;
The processor usage rate indicating the usage rate of the processor of the computer and the free memory amount indicating the free memory of the computer are input from each computer of the plurality of computers, and the processor usage rate and free space of each input computer are input. A computer load collection unit for outputting a memory amount;
When it is determined by the migration process determination unit that the other process is to be executed by another computer, the processor usage rate and the free memory amount of each computer output by the computer load amount collection unit Enter processor usage and free memory,
Compare the amount of free memory of the computer with the amount of consumed memory stored in the system configuration information table / memory in order of the processor usage rate of the other computer that has been input. computer determines, computer resource dynamic control device according to claim the resource allocation unit for controlling to cause to execute the process on the determined other computer further comprising a <br/>. The system configuration information table / memory stores in advance a memory amount representing the capacity of the memory of the computer corresponding to each of the computers,
The computer load amount collection unit inputs a memory usage amount representing a memory usage amount of the computer together with the processor usage rate from each computer of the plurality of computers, and the processor usage rate and the memory usage amount of each inputted computer And
The computer resource dynamic control device is
The computer load history table memory for storing one or more computer load history information corresponding to each computer, using the processor usage rate and memory usage output by the computer load amount collection unit as computer load history information,
The processing time is obtained corresponding to each processing time history information of the one or more processing time history information stored in the processing time history table memory, and the dead time is determined based on one or more fluctuations of the obtained processing time. It is determined whether or not the processing time exceeding the line time will occur in the future, and when it is determined that the processing time exceeding the deadline time will occur in the future, the processing time exceeding the deadline time occurs A deadline miss occurrence time predicting unit that predicts and outputs a time based on the variation of the one or more processing times;
Corresponding to the occurrence time predicted by the deadline miss occurrence time predicting unit based on the processor usage rate fluctuation of one or more computer load history information corresponding to each computer stored in the computer load history table memory A computer load prediction unit that calculates and outputs the processor usage rate of each computer
Corresponding to the occurrence time predicted by the deadline miss occurrence time prediction unit based on the variation in memory usage of one or more computer load history information corresponding to each computer stored in the computer load history table memory A memory usage prediction unit that calculates and outputs the memory usage of each computer ;
With
When the occurrence time is output by the deadline miss occurrence time prediction unit , the migration process determination unit determines to cause the other process to execute the process,
Said resource allocation unit,
When the occurrence time is output by the deadline miss occurrence time prediction unit, when the migration process determination unit determines to cause the other process to be executed by the migration process determination unit, the output from the computer load prediction unit Enter the processor usage rate of other computers among the processor usage rates of multiple computers corresponding to the time of occurrence.
In order of the processor usage rate of the other computers you entered,
(1) Obtain the memory amount of the other computer from the memory amount corresponding to each computer stored in the system configuration information table memory.
(2) Input the memory usage of the other computer among the memory usages of the plurality of computers corresponding to the occurrence time output by the memory usage prediction unit,
(3) The free memory amount obtained by subtracting the input memory usage from the acquired memory amount is obtained, and the obtained free memory amount is compared with the consumed memory amount stored in the system configuration information table memory. Determine whether the amount is greater than or equal to the amount of memory consumed,
When it is determined that the amount of free memory is equal to or greater than the amount of consumed memory, control is performed to cause another computer having the amount of free memory equal to or greater than the amount of consumed memory to execute the process. The computer resource dynamic control apparatus according to claim 1.

Images