JP2015185036A - Load distribution method, system, device and program - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a method and system capable of avoiding deterioration of processing capacity of a whole system, and being suitable for load distribution on a confluence.SOLUTION: A system comprises: at least one first device 21 for processing inputted data and outputting processing result data; and a second device 22 for executing processing with respect to the processing result data outputted from the first device 21. The at least one first device 21 grasps a load state of the second device 22, based on a data amount of pieces of data which are in a processing standby state on the second device 22, and the first device 21 executes the processing which is executed on the second device 22 by an amount according to the load state of the second device 22.

Description

  The present invention relates to a load balancing method, system, apparatus, and program.

  In a system including a plurality of devices capable of operating in parallel, a part (branch unit) in which the output of one device is distributed as an input to the plurality of devices, and the output from the plurality of devices is input to a smaller number of other devices As such, a system including a portion that merges (a merging portion) is known. For example, Patent Document 1 discloses an information processing apparatus including a branching unit that selects a transmission path to a processor and outputs a packet to the transmission path, and a merging unit that merges packets from each processor. Yes.

  In this type of system, in the branching unit, for example, the output from one output source (transmission source) device is distributed as an input to each of a plurality of branch destination devices. In the merge unit, for example, outputs from a plurality of transmission source devices are merged in, for example, a queue and processed. The plurality of devices may be a plurality of switches or the like, or may be a plurality of processors, a CPU (Central Processing Unit) core, or the like.

JP 2000-207382 A

  The analysis of related technology is given below.

  In a system including a plurality of devices capable of operating in parallel as described above, it is known that a portion with a large load can become a bottleneck of the processing capacity of the entire system if there is a bias in load among the devices. For this reason, in order to maintain the processing performance of the system, it is desirable that the load on the apparatus is averaged between the apparatuses.

  For example, a well-known round robin method (assigned in order at fixed time intervals and possessed as a load balancing control for averaging the load among a plurality of devices to which the branching is performed. When the time elapses, it can be applied to the end of the rescheduling target).

  On the other hand, with respect to load control at a confluence section that merges a plurality of outputs, for example, a back pressure method of a flow control method used in a half-duplex connection hub or switch is known. In the back pressure system, for example, when the reception buffer is likely to overflow, control is performed such as transmitting a collision signal to the transmission source to temporarily interrupt data transmission.

  As described above, the load control in the merging unit is performed such that, when the merging destination device detects an increase in the load, the output source (transmission source) device is not transmitted. For this reason, a time during which processing cannot proceed is generated in the output source device. As a result, the processing capacity of the entire system may be reduced.

  Therefore, the present invention has been devised in view of the above problems, and its purpose is to avoid a decrease in the processing capacity of the entire system, for example, a system, apparatus, and the like that are suitable for load control at a junction or the like, It is to provide a load balancing method and program.

  According to one aspect of the present invention, at least one first device that processes input data and outputs processing result data, and the processing result data output from the first device has a predetermined value. A second device that executes processing, and at least one of the first devices grasps the load status of the second device from the amount of data waiting for processing in the second device, and A system is provided that includes means for controlling the predetermined processing performed by the second device to be executed by the first device by an amount corresponding to the load status of the second device.

  According to another aspect of the present invention, means for processing input data and outputting processing result data to another device, and a load status of the other device based on a data amount waiting for processing in the other device Thus, an apparatus is provided that includes means for controlling to execute predetermined processing performed by the other apparatus according to the load status of the other apparatus.

According to still another aspect of the present invention, at least one first device processes input data and outputs processing result data.
The second device executes a predetermined process on the processing result data output from the first device,
At least one of the first devices grasps the load status of the second device from the amount of data waiting for processing in the second device,
There is provided a load distribution method for controlling the predetermined processing performed by the second device to be executed by the first device by an amount corresponding to the load status of the second device.

According to still another aspect of the present invention, a computer constituting an apparatus that processes input data and outputs processing result data to another apparatus,
From the amount of data waiting for processing in the other device, grasp the load status of the other device,
There is provided a program for executing a process for controlling to execute a predetermined process performed in the other apparatus by an amount corresponding to a load state of the other apparatus.

  According to still another aspect of the present invention, a computer-readable recording medium (for example, a semiconductor storage or a magnetic / optical storage) that records the program is provided.

  According to the present invention, it is possible to realize load distribution that avoids a decrease in the processing capacity of the entire system, and it is suitable for load control to, for example, a junction.

(A) is a figure explaining one Embodiment of this invention. (B) is a diagram illustrating the contents of a queue. It is a flowchart explaining operation | movement of one Embodiment of this invention. It is a figure explaining one Embodiment of this invention. It is a figure which illustrates typically an example of the composition of the device of one embodiment of the present invention. It is a figure explaining the basic concept of this invention.

  Embodiments of the present invention will be described with reference to the drawings. First, the basic concept of the present invention will be described. According to the present invention, referring to FIG. 5, the system 20 performs at least one first device 21 that performs processing 211 on the input data 23 and outputs processing result data 24, and the first device 21. And a second device 22 that executes processing 221 on the processing result data 24 output from the device 21. The first device 21 grasps the load status of the second device 22 from the information 213 regarding the amount of data in the processing waiting state 222 in the second device 22, and is performed by the second device 22. The processing unit 221 includes a control unit 212 that controls the first device 21 to execute the process 221 in accordance with the load status of the second device 22. The first device 21 and the second device 22 may be processing devices such as a processor and a CPU core. The processing 211 and the control means 212 in the first device 21 may of course implement their functions by a program executed by a computer, a processor, a CPU, or the like that constitutes the first device 21. .

  By controlling the processing 221 performed by the second device 22 to be executed by the first device 21 by an amount corresponding to the load status of the second device 22, the load at the junction or the like In performing the control, it is possible to avoid a decrease in the processing capacity of the entire system.

  According to one aspect of the present invention, the second device (22 in FIG. 5 or, for example, device 14 in FIG. 1A) has been processed by one or more of the first devices. A queue (222 in FIG. 5 or 15 in FIG. 1A) for inputting data and holding it until the execution of processing is provided, and at least one of the first devices (21 in FIG. 5 or FIG. ), For example, the device 11), for the input data, the first device (FIG. 5) instead of the second device (22 in FIG. 5 or the device 14 in FIG. 1A). 21 or the number of processes executed by the device 11 in FIG. 1A, for example, is set to the waiting time of the second device (22 in FIG. 5 or the device 14 in FIG. 1A). You may make it determine based on the length information of a matrix.

According to one embodiment of the present invention, at least one of the first devices (for example, the device 11 in FIG. 1A) performs a first process (FIG. 1A) on the input data. Process-2) or the first process (process-2 in FIG. 1A) and the second process (process-3 in FIG. 1A) are executed.
The first device (eg, device 11 in FIG. 1A) is the first device (eg, device 11 in FIG. 1A) and the second process (process 3 in FIG. 1A). ) And an offload counter (117 in FIG. 4) that counts the number of data executed, and the queue (15 in FIG. 1 (A)) of the second device (eg, device 14 in FIG. 1 (A)). Based on the length information, the upper limit of the number of data to be executed by the first device (eg, the device 11 in FIG. 1A) is the second process (Process 3 in FIG. 1A). It is good also as a structure provided with the means (threshold value calculation part 118 of FIG. 4) which acquires a threshold value.
The first device (for example, the device 11 in FIG. 1A) executes the first processing (processing-2 in FIG. 1A) on the input data, and then performs an offload counter. If the count value exceeds the threshold, the second process (process-3 in FIG. 1A) is not executed, and the process result data of the first process is stored in the second process. It is good also as a structure which transmits to an apparatus (for example, apparatus 14 of FIG. 1 (A)). The second device (eg, the device 14 in FIG. 1A) is the first device for the data for which the first process (Process-2 in FIG. 1A) has been executed. Process 2 (process-3 in FIG. 1A) may be executed.
On the other hand, when the count value of the offload counter is equal to or smaller than the threshold, the first device (for example, the device 11 in FIG. 1A) performs the second processing (processing 3 in FIG. 1A). ) Is further executed to increment the offload counter (117 in FIG. 4) by one, and the first process (process-2 in FIG. 1A) and the second process (FIG. 1A) The processing result data subjected to the process 3) may be transmitted to the second apparatus (for example, the apparatus 14 in FIG. 1A).

  According to one aspect of the present invention, the first device (for example, the device 11 in FIG. 1A) receives the data subjected to the first processing (processing-2 in FIG. 1A), Or, it indicates to what processing the processing result data subjected to the first processing (processing-2 in FIG. 1A) and the second processing (processing-3 in FIG. 1A) has been completed. It is good also as a structure which transmits to said 2nd apparatus (for example, apparatus 14 of FIG. 1 (A)) with completion step information. The second device (eg, device 14 in FIG. 1A) transmits to the queue (15 in FIG. 1A) from the first device (eg, device 11 in FIG. 1A). The processed result data and the completed step information may be stored in association with each other (see FIG. 1B).

  According to one aspect of the present invention, the first device (eg, device 11 in FIG. 1A) is the queue of the second device (eg, device 14 in FIG. 1A). With respect to the length of 15) in FIG. 1A, the range of the length is divided into a plurality of sections, the threshold is set in advance corresponding to each of the sections, and the acquired second device A threshold value corresponding to the queue length may be acquired from the queue length (see FIG. 4).

  Although the basic concept and the form thereof have been described above, exemplary embodiments will be described below.

  According to the embodiment of the present invention, in the system (1) composed of devices operating in parallel (10, 11, 12, 13, 14 in FIG. 1 (A)), a plurality of devices (11, 12, 13). ) And the device (14) that receives the output of those devices so that the work amount between these devices (11, 12, 13, 14) becomes average. Although not particularly limited, for example, in a multi-core CPU system in which each device is a processing device, for example, a CPU core, with respect to a plurality of cores and a core that receives the outputs of the plurality of cores, the average workload of these cores To do. In FIG. 1A, 2 represents an input channel and 3 represents an output channel.

  According to the embodiment of the present invention, when a device (14) that is a merging destination performs a relatively high load on a part of processing performed at a relatively low load, a plurality of output sources (transmissions) Control is performed so as to be executed by the original device (11, 12, 13). As a result, for example, the load between the output source devices (11, 12, 13) and the merging destination device (14) can be averaged. At that time, the output source devices (11, 12, 13) do not perform control for reducing the processing performance of the system, such as stopping the processing being executed.

  According to the embodiment of the present invention, the output source device (11, 12, 13) determines the load status (load) of the device (14) from the data length held in the queue (15) and placed in the process waiting state. May be determined.

Further, according to the embodiment of the present invention, the movement of the process is performed in stages by performing a certain calculation on the number of data accumulated in the queue (15). For example, as a ratio of processing executed in the output source device (11, 12, 13),
For example, 10% (1 out of 10 data is processed up to processing-3, otherwise only processing-2 is performed),
20% (2 out of 10 data is processed up to process-3, otherwise only process-2 is performed),
30% (same as 20%). This prevents the load on the output source devices (11, 12, 13) from increasing due to the movement of the process.

<Embodiment 1>
An embodiment of the present invention will be described with reference to FIG. In FIG. 1A, an information processing system 1 having an input channel 2 and an output channel 3 is illustrated. Receives data from the input channel 2, performs four processes of process-1, process-2, process-3, and process-4 in order on the received data, and transmits process result data from the output channel 3 ing.

  The devices 10, 11, 12, 13, and 14 can be operated in parallel. The devices 10, 11, 12, 13, and 14 may be configured with a CPU core. The devices 11, 12, 13, and 14 include a queue that stores and holds data exchanged between the devices. In FIG. 1A, only queue 15 of device 14 is shown for simplicity.

  The device 10 comprises a queue (not shown) consisting of buffers for storing data received on the input channel 2. Each device 10, 11, 12, 13, 14 operates in parallel, but the processing time is not constant. Therefore, each device has a queue and stores data transmitted from the output source device until the processing is executed by the device itself.

  The queues (not shown) of the devices 11, 12, and 13 store data output from the device 10. The queue 15 of the device 14 stores data output from the devices 11, 12, and 13. Although not particularly limited, the device 14 queues the output from the devices 11, 12, and 13 sequentially in the queue 15, for example, in the order of arrival, and dequeues the output from the queue 15 when processing is performed. Although not particularly limited, the queue of each device 11-14 may be constituted by, for example, a first-in first-out (FIFO) buffer.

  The outline of the operation of the present embodiment will be described with reference to FIG.

  Receive (input) data from input channel 2. Processing-1 for the input data is executed by the apparatus 10 in a fixed manner. The device 10 executes the process 1 and further distributes the output to the devices 11, 12, and 13.

  Process-2 is executed by one of the apparatuses 11, 12, and 13 according to the distribution result of the apparatus 10.

  There are two types of processing-3: a case where the devices 11, 12 and 13 execute and a case where the device 14 executes. However, each data is executed only once by any device. That is, for example, processing-3 is performed on the data distributed by the device 10 in the output source device 11 and the processing result data is transmitted to the device 14, and the device 14 applies the processing-3 to the data. Process-3 is not executed.

  Process-4 is executed by the device 14.

  Although not particularly limited, in the example of FIG. 1A, it is assumed that the process-2 is a process (a large amount of processing) that requires a longer processing time than other processes. For this reason, the process-2 is assigned to the three apparatuses 11, 12, and 13, and these apparatuses 11, 12, and 13 are operated in parallel to avoid a decrease in system performance (throughput).

  The data stored in the queues (not shown) of the devices 11, 12, and 13 and the data stored in the queue 15 of the devices 14 have different properties. That is, the data stored in the queues (not shown) of the devices 11, 12, and 13 are the results of executing the process-1. On the other hand, there are two types of data stored in the queue 15 of the device 14, that is, a result of executing the process-2 and a result of executing the process-3.

  Therefore, the queue 15 of the apparatus 14 stores data that is a set of processing results and information indicating which processing has been completed. FIG. 1B is a diagram schematically illustrating an example of the contents of the queue 15. In the example shown in FIG. 1B, the first, second, and fourth data in the queue 15 have completed steps 2, that is, processing 2 is completed, and processing results are XXX, YYY, and WWW. It represents something. The third data in the queue 15 indicates that the completion step is 3, that is, the process 3 is completed, and the process result is ZZZ.

  In the present embodiment, in the devices 11, 12, and 13, whether only the process 2 is executed on the data, or whether the process 3 is executed in addition to the process-2 is determined based on the data stored in the queue 15. The structure is determined by the length. The apparatuses 11, 12, and 13 check the length of the data in the queue 15 of the apparatus 14 after executing process-2 for each received data. The apparatuses 11, 12, and 13 determine how much of the data (percent) is to be executed up to the process 3 according to the data length of the queue 15 of the apparatus 14. The apparatuses 11, 12, and 13 execute up to process-3 for the data of the ratio (threshold value) among a plurality of data to be processed. For data exceeding the ratio, processing up to process-2 is executed.

  The processing results of the devices 11, 12, and 13 are transmitted to the device 14 together with the processing steps (completion steps) executed by the devices 11, 12, and 13, and stored in the queue 15 of the device 14. Device 14 retrieves the data from queue 15 and checks the completion step.

  If it is the completion step 2, the apparatus 14 performs the process-3 and the process-4 on the processing result, and outputs the processing result from the output channel 3. If the completion step is 3, the device 14 executes only the process-4 on the process result and outputs the process result from the output channel 3.

  FIG. 4 is a diagram illustrating the configuration of the device 11 (output source device) in FIG. Note that the devices 11 to 13 in FIG. 1A may have the same configuration. The basic configurations of the devices 10 and 14 may be the same as those of the device 11.

  Referring to FIG. 4, the device 11 includes, as a data path, a receiving unit 111 that receives data, and a reception buffer that temporarily stores and holds data received by the receiving unit 111 (processing waiting data received from the device 10). A queue 112, a processing unit 113 that executes processing on data extracted from the queue 112, and a transmission unit 114 that transmits a processing result and a completion step in the processing unit 113 are provided. In addition, in the apparatuses 10 and 14 in which the processing to be executed is fixed, the transmission unit 114 may output only the processing result, and the completion step may not be output.

  The apparatus 11 performs scheduling such as writing of received data to the queue 112 at the data path receiving unit 111, reading of data from the queue 112, and switching of processing executed by the processing unit 113 to another processing, A control unit 115 that controls output from the transmission unit 114 is provided.

  Further, as a control system, a reception counter 116 that counts data received from the queue 112, and an off-load counter that counts the number of cases in which the processing unit 113 executes the processing 3 among the received data in the device 11. 117, a queue length acquisition unit 119 for acquiring length information of the queue (15 in FIG. 1A) of the subsequent device (14 in FIG. 1A), and the queue length of the device 14 A threshold value calculation unit 118 for deriving a threshold value from the depth information is provided. The control unit 115 controls each of these units 116-119.

  The off-load counter 117 is a counter for controlling load reduction (off-load) in the device 14, and processing 3 in the processing unit 113 of the device 11 is executed according to the degree of load on the device 14. This is a counter for managing the number of data items.

  If it is determined that the load level of the device 14 is relatively high, the device 11 performs the process 3 on more received data. The process-3 performed by the device 14 when the load is relatively low is performed by the device 11 instead of the device 14 when the device 14 is relatively high load. 11, process 3 is executed for more received data.

  In order to realize this load distribution control, the control unit 115 executes the process 2 and then the value of the off-load counter 117 is larger than the threshold value calculated by the threshold value calculation unit 118 based on the queue length of the device 14. Output to the device 14 together with the completion step = 2. When the value of the off-load counter 117 is equal to or smaller than the threshold value, the process 3 is executed, the off-load counter 117 is incremented by one, and output to the apparatus 14 together with the completion step = 3.

  In FIG. 4, for the sake of simplicity, the control unit 115 is illustrated as a configuration in which the data path units 111 to 114 and the control system units 116 to 119 are controlled by control signal lines. It is good also as a structure provided with the control part of the control system separately, or it is good also as a structure provided with each part 116-119 of the control system in the control part 115.

  FIG. 2 is a flowchart illustrating the operation of one embodiment of the present invention. FIG. 3 is a diagram for explaining the operation of the embodiment of the present invention. With reference to FIGS. 2 to 4, the detailed operation of the devices 11, 12, and 13 will be described. The devices 11, 12, and 13 operate according to the same procedure, and will be described as the device 11.

  In the apparatus 11, the count value and the threshold value of the reception counter 116 are set to 0 to 100. However, it is a matter of course that the value is not limited to this value (for example, 100 may be changed to another value). When the threshold value is 100, the maximum number of processing-3 processing items (counted by the off-load counter 117) for received data in the apparatus 11 is 100.

  For example, when the device 11 is initialized, such as during power-up or reset, the control unit 115 of the device 11 clears the reception counter 116 and the off-load counter 117 to 0 (step S1).

  The device 11 enters a loop (infinite loop) for processing data from the device 10 (for example, data processing is repeated except that data processing is stopped by interruption or reset based on an external stop instruction or the like).

  The processing unit 113 of the device 11 receives data from the queue 112 (step S2). In response to this, the control unit 115 increases the count value of the reception counter 116 by one (step S3). In the apparatus 11, when the processing unit 113 receives data from the queue 112, the length of the queue 112 (corresponding to the amount of data held in the process waiting state) decreases by one unit.

  As a result, when the count value of the reception counter 116 exceeds 100 (YES branch of step S4), the control unit 115 clears the reception counter 116 and the off-load counter 117 to 0 (step S5), and then step S6. Move on. If the count value of the reception counter 116 is 100 or less, for example (NO branch of step S4), the process proceeds to step S6.

  In step S6, process-2 is performed on the data received by the processing unit 113.

  The queue length acquisition unit 119 acquires length information of the queue 15 (see FIG. 1A) of the joining apparatus 14 (step S7). The device 14 may be configured to notify the device 11 of the length information of the queue 15 in the own device from the management unit that manages the queue 15 via a control communication channel (not shown).

  The threshold value calculation unit 118 calculates (derivates) the threshold value by performing a predetermined calculation (or using a table lookup method) on the queue 15 (FIG. 1A) of the acquired joining device 14. Then, the calculated threshold value is supplied to the control unit 115 (step S8).

  The threshold is set according to a specific system, for example. Although not particularly limited, as a simple example, as illustrated in FIG. 3, when the length of the queue 15 of the acquired device 14 is “18”, the threshold calculated by the threshold calculation unit 118 is 10. If the length of the queue 15 of the acquired device 14 is “0” to “9”, it is determined that the load on the device 14 is low, so the threshold value is “0”. In this case, the apparatus 11 does not execute the process 3 on the received data, and the apparatus 14 executes the process 3.

  The control unit 116 compares the threshold value calculated by the threshold value calculation unit 118 with the count value of the off-load counter 117. When the count value of the off-load counter 117 exceeds the threshold value (YES branch of step S9), the completion step = 2 (information indicating “processing up to processing-2”) and the result of processing-2 are It transmits to the apparatus 14 (step S10), and returns to step S2.

  In the apparatus 11, when the threshold value calculated by the threshold value calculation unit 118 is equal to or larger than the count value of the off-load counter 117 (NO branch in step S9), the processing unit 113 receives the received data (for example, processing in step S6-2). Process-3 is executed on the data that has been executed (step S11).

  Thereafter, the control unit 115 increases the count value of the off-load counter 117 by +1 (step S12).

  The control unit 115 transmits the result of processing-3 to the device 14 together with the completion step = 3 (information indicating that “processing up to processing-3”) from the transmission unit 114 (step S13). Return.

  It should be noted that in the device 11, the data that is processed in processing-2 and processing-3, and that increases the count value of the off-load counter 117 by +1 (data that is extracted from the queue 112) is the same as the output source device (FIG. 1). The data unit may be a fixed length (fixed length) such as packet data, frame data, etc. transmitted from the device 10) to the device 11. Alternatively, variable length data may be used. In this case, when processing-3 is performed on data having a data amount twice the unit data length (size), the count value of the off-load counter 117 is increased by twice the unit data length, and thus +2. It may be.

  Although depending on the implementation, in the queue length acquisition process of the joining destination device in step S7 of FIG. 2, for each received data, the processing time for obtaining the queue length of the joining destination device is taken into consideration. Instead of acquiring the queue length of the destination device (14 in FIG. 1A), the queue length may be acquired every predetermined period.

  According to the above-described embodiment, the following operational effects are obtained.

A) When the load of the joining device (14 in FIG. 1 (A)) increases, the load is shared with the data output device (11, 12, 13 in FIG. 1 (A)). It is configured. As a result, the load is averaged over the entire system. In other words, it enables effective utilization of resources and capabilities in the entire system and contributes to an improvement in throughput.

B) It is not necessary to add a dedicated device or process to measure the load of the joining device (14 in FIG. 1A). A device queue length management function is typically provided in devices with queues for resource management. This is because the output source device uses this function to acquire the queue length information of the joining destination device and determine the load status of the joining destination device.

C) There is an effect that the load distribution of the processing is promoted between the devices 11, 12, and 13. In the apparatuses 11, 12, and 13, the process-2 (or the process 2 and the process-3) is executed, or the amount of the process-3 that the apparatuses 11, 12, and 13 additionally execute is the data processing capability. The higher the device, the higher. This is because an apparatus capable of processing a large amount of data increases the number of data items for executing process-3.

<Modification 1>
As a modification of the embodiment, there is application to pipeline processing. In the embodiment shown in FIG. 1, there are three devices 11 to 13 in the front stage and one device 14 in the rear stage. It is possible to introduce the mechanism of the above-described embodiment into the pipeline processing of the preceding and succeeding apparatuses by paying attention to one part of each of the preceding and succeeding apparatuses. That is, the preceding apparatus executes the pipeline stage performed by the subsequent apparatus in place of the subsequent apparatus in accordance with the load status of the subsequent apparatus. In this case, since the apparatus of the preceding stage is one, the effect of C) cannot be obtained, but the effect of A) and B) can be obtained.

<Modification 2>
In yet another modification, a plurality of processes for dynamically allocating between apparatuses are used. In the above-described embodiment described with reference to FIG. 1, it is an object of dynamically determining whether to execute in the upstream (output source) devices 11, 12, and 13 or in the downstream (merging destination) device 14. The process is limited to process-3. However, when the processing executed by the device 14 includes processing-3, processing-4, processing-5, etc., the processing-3 and processing-4 should be dynamically executed by the preceding device (output source). Is also possible. In this case, the information of the completion step attached to the data set in the queue 15 of the apparatus 14 at the subsequent stage (joining destination) is any one of “2”, “3”, and “4”. In response to this completion step information, the device 14
“Execute process-3, process-4, process-5”,
“Execute process-4 and process-5”,
“Execute process-5”
Select one of the following.

  For example, when the system includes one or more preceding devices (processing devices) and one succeeding device (processing devices) as part of the system, the above-described embodiment is suitable for application to load control of the parts. It is said. It is also suitable for application to a communication device that processes a large amount of routine data.

  The disclosure of the above patent document is incorporated herein by reference. Within the scope of the entire disclosure (including claims) of the present invention, the embodiments and examples can be changed and adjusted based on the basic technical concept. Various combinations or selections of various disclosed elements (including each element of each claim, each element of each embodiment, each element of each drawing, etc.) are possible within the scope of the claims of the present invention. . That is, the present invention of course includes various variations and modifications that could be made by those skilled in the art according to the entire disclosure including the claims and the technical idea.

1, 20 System 2 Input channel 3 Output channel 10, 11, 12, 13, 14 Device 15 Queue 21 First device 22 Second device 23 Input data 24, 25 Processing result 111 Receiving unit 112 Queue 113 Processing unit 114 Transmission Unit 115 Control Unit 116 Reception Counter 117 Off-Load Counter 118 Threshold Calculation Unit 119 Queue Length Acquisition Unit 211, 221 Processing 212 Control Means (Control Function)
213 Amount of data waiting to be processed (queue length)
222 Data waiting for processing (queue)

Claims (10)

At least one first device for processing input data and outputting processing result data;
A second device that executes a predetermined process on the processing result data output from the first device;
With
At least one of the first devices is
The load status of the second device is ascertained from the amount of data waiting for processing in the second device, and the predetermined processing performed in the second device is performed according to the load status of the second device. A system comprising: means for controlling to be executed by the first apparatus. The second device includes a queue that receives data processed by one or more of the first devices and holds the data until the processing is executed.
At least one of the first devices is
The number of the predetermined processes to be executed by the first device instead of the second device is determined based on the queue length information of the second device for the input data. The system according to claim 1, wherein: At least one of the first devices is
Performing the first process or the first process and the second process on the input data;
The predetermined process performed in the second device is the second process;
The first device includes:
An offload counter that counts the number of data for which the second processing has been executed by the first device;
Means for obtaining a threshold value that is an upper limit of the number of data to be executed by the first device based on the queue length information of the second device;
With
The first device includes:
After executing the first process on the input data,
When the count value of the offload counter exceeds the threshold value, the processing result data of the first process is transmitted to the second device without executing the second process,
The second device includes:
Executing the second process on the data on which the first process has been executed by the first device;
In the first device,
When the count value of the offload counter is less than or equal to the threshold value, the second process is further executed to increase the offload counter by one, and the first process and the second process are performed. The system according to claim 2, wherein processing result data is transmitted to the second device. Completion step indicating at which one of the first devices the data that has been subjected to the first processing or the processing result data that has been subjected to the first processing and the second processing has been completed. Send to the second device along with the information;
The system according to claim 3, wherein the second device stores the processing result data transmitted from the first device and completion step information in association with each other in the queue.   At least one of the first devices divides the length range into a plurality of sections with respect to the queue length of the second device, and presets the threshold value corresponding to each of the sections. The threshold value corresponding to the queue length is acquired from the queue length of the acquired second device, according to any one of claims 2 to 4. The described system.   The system according to claim 1, further comprising a plurality of the first devices operating in parallel with each other. Means for processing the input data and outputting the processing result data to another device;
From the amount of data waiting for processing in the other device, grasp the load status of the other device,
An apparatus comprising: means for controlling to execute a predetermined process performed in the other apparatus according to a load state of the other apparatus.   Determining the number of processes to be executed on behalf of the other apparatus for the input data based on length information of a queue that holds the data until the process is executed by the other apparatus; 8. A device according to claim 7, characterized in that At least one first device processes input data and outputs processing result data.
The second device executes a predetermined process on the processing result data output from the first device,
At least one of the first devices grasps the load status of the second device from the amount of data waiting for processing in the second device,
A load distribution method comprising: controlling the predetermined processing performed by the second device to be executed by the first device by an amount corresponding to a load situation of the second device. In a computer constituting a device that processes input data and outputs processing result data to other devices,
From the amount of data waiting for processing in the other device, grasp the load status of the other device,
A program for executing a process of controlling to execute a predetermined process performed in the other apparatus according to a load state of the other apparatus.

Images