JP2014164568A - Terminal device, distributed processing method, distributed processing program, and distributed processing system - Google Patents

Abstract

PROBLEM TO BE SOLVED: To increase processing speed of an application in a mobile terminal.SOLUTION: In a mobile terminal 20, a communication state is measured and a throughput is predicted from the measured communication state by using a model created on the basis of information collected from the plurality of mobile terminals 20. Then, the mobile terminal 20 determines whether to cause the own device or a calculation processing server 40 to perform processing related to an application according to the predicted throughput, and controls so as to cause the own device or the calculation processing server 40 to perform the processing.

Description

  The present invention relates to a terminal device, a distributed processing method, a distributed processing program, and a distributed processing system.

  In recent years, with the development and popularization of IT technology such as distributed processing technology, communication technology, and terminal technology, it is becoming possible to use necessary resources in necessary scenes regardless of time and place. For example, in an application that performs heavy processing such as image recognition and video editing, the processing to be executed is distributed according to the characteristics of the execution environment between the server and the terminal, and the processing speed is increased as a result. Can be planned.

  Regarding the processing to be executed in accordance with the characteristics of the execution environment between the server and the terminal, the processing execution environment assignment specification is set and fixed at the time of application development, so the processing speed may be reduced. For example, depending on external factors such as deterioration of the communication environment, it may be better to change the execution environment of the process. As described above, if the process distribution is fixed, the response time becomes longer. In some cases, usability decreased. On the other hand, there is known a technique for performing dynamic switching control by determining whether to switch processing execution to a server based on a measurement result of a communication state such as electric field strength.

Nobu Matsuoka, “Android programming to learn while creating”, pages 152-159, Softbank Creative, Inc., March 28, 2011

  However, in the technology that determines whether to switch the execution of processing to the server based on the measurement result of the communication state such as the electric field strength described above, whether the cause of the response time deterioration is the application execution environment or the communication environment. There is a problem that it is unknown and switching of the appropriate execution environment cannot be performed, and the processing speed of the application in the terminal may decrease.

  Accordingly, the present invention has been made to solve the above-described problems of the prior art, and an object thereof is to improve the processing speed of an application in a terminal.

  In order to solve the above-described problems and achieve the object, a terminal device according to the present invention uses a measurement unit that measures a communication state and a model created based on information collected from a plurality of terminal devices. A prediction unit that predicts throughput from the communication status measured by the measurement unit, and determines whether the device or server is to execute processing related to the application according to the throughput predicted by the prediction unit. And a control unit that controls to execute the processing.

  The distributed processing method according to the present invention is a distributed processing method executed in a terminal device, and includes a measurement process for measuring a communication state and a model created based on information collected from a plurality of terminal devices. Using the prediction process for predicting the throughput from the communication status measured by the measurement process, and determining whether or not the apparatus or server is to execute processing related to the application according to the throughput predicted by the prediction process. Or a control step of controlling the server to execute the process.

  Further, the distributed processing program according to the present invention uses a measurement step for measuring a communication state and a model created based on information collected from a plurality of terminal devices, based on the communication state measured by the measurement step. A prediction step for predicting throughput, and control for determining whether or not the device or server is to execute the process related to the application according to the throughput predicted by the prediction step, and controlling the device or server to execute the process And causing the computer to execute the steps.

  The distributed processing system according to the present invention is a distributed processing system including a terminal device and a server, and the terminal device is based on information collected from a measuring unit that measures a communication state and a plurality of terminal devices. A prediction unit that predicts throughput from the communication status measured by the measurement unit using the model created by the measurement unit, and causes the apparatus or server to execute processing related to the application according to the throughput predicted by the prediction unit A control unit that controls to cause the device or server to execute the process, and the server determines that the terminal device causes the server to execute a process related to an application. It has a processing part which performs this processing and transmits a processing result to the above-mentioned terminal unit.

  The terminal device, the distributed processing method, the distributed processing program, and the distributed processing system according to the embodiment can improve the processing speed of the application in the terminal.

FIG. 1 is a diagram illustrating an example of a configuration of a system according to the first embodiment. FIG. 2 is a diagram for explaining the configuration of the mobile terminal and each server according to the first embodiment. FIG. 3 is a diagram illustrating model generation processing. FIG. 4 is a diagram for explaining the distribution of throughput. FIG. 5 is a diagram for explaining the relationship between the distribution of throughput and the threshold value. FIG. 6 is a diagram for explaining offload determination processing. FIG. 7 is a sequence diagram showing a processing procedure of learning model generation processing of the system according to the first embodiment. FIG. 8 is a flowchart showing a processing procedure of distributed processing of the mobile terminal according to the first embodiment. FIG. 9 is a diagram illustrating a computer that executes a distributed processing program.

  Hereinafter, embodiments of a terminal device, a distributed processing method, a distributed processing program, and a distributed processing system according to the present invention will be described in detail with reference to the accompanying drawings. In addition, this invention is not limited by this embodiment.

[First embodiment]
First, the outline of the system according to the first embodiment will be described with reference to FIG. FIG. 1 is a block diagram showing a configuration of a system according to the present embodiment. As shown in the figure, the system according to the present embodiment includes a network 10, a plurality of mobile terminals 20, a prediction model generation server 30, and a plurality of arithmetic processing servers 40. In addition, each mobile terminal 20, the prediction model generation server 30, and the calculation processing server 40 are connected via the network 10. Here, the network 10 is, for example, 3G (3rd Generation), LTE (Long Term Evolution), WiFi (Wireless Fidelity) (registered trademark), or the like. Communication in each mobile terminal 20 and each server 30 and 40 is not limited to wireless communication, and may be wired communication. Here, the mobile terminal 20 is described as an example of an apparatus that communicates with each of the servers 30 and 40. However, as will be described later, an apparatus that communicates with each of the servers 30 and 40 is limited to the mobile terminal 20. It is not a thing.

  In the following description, a system in which each mobile terminal 20 executes an application in cooperation with each arithmetic processing server 40 having a distributed execution environment will be described. However, this system is limited to such an embodiment. is not. For example, a system that executes applications in cooperation between mobile terminals 20 or between arithmetic processing servers 40 may be used. In this system, the types of execution environments of the mobile terminals 20 and the arithmetic processing servers 40 and the types of applications to be executed may be the same.

  Each mobile terminal 20 is an information processing apparatus that executes various applications. Each mobile terminal 20 executes an application in cooperation with each server 40. For example, the mobile terminal 20 is connected to the Internet such as a mobile phone, a PHS (Personal Handyphone System), a PDA (Personal Digital Assistant), a tablet PC (Personal Computer), etc., and is connected to each of the servers 30 and 40 and other user terminals. Any device can be used as long as it can transmit and receive information between them. In addition, the computer is not limited to a mobile terminal, and a computer installed in a home or office may be used.

  The mobile terminal 20 measures the communication status, and predicts the throughput from the measured communication status using a model created based on information collected from the plurality of mobile terminals 20. In addition, the mobile terminal 20 determines whether or not to offload the arithmetic processing related to the application from the mobile terminal 20 to the arithmetic processing server 40 using the throughput, and causes the own device or the arithmetic processing server 40 to execute the arithmetic processing related to the application. To control.

  The prediction model generation server 30 collects data related to communication from a plurality of mobile terminals 20 and generates a model for predicting throughput. The arithmetic processing server 40 executes arithmetic processing related to the application in the mobile terminal 20 and transmits the execution result to the mobile terminal 20.

  Next, the structure of the mobile terminal 20 and each server 30 and 40 is demonstrated using FIG. FIG. 2 is a diagram for explaining the configuration of the mobile terminal and each server according to the first embodiment.

  First, the configuration of the prediction model generation server 30 will be described. The prediction model generation server 30 includes a data collection unit 31 and a model generation unit 32.

  The data collection unit 31 collects data related to communication from a plurality of mobile terminals 20 via the network 10 such as 3G, LTE, or WiFi. For example, as illustrated in FIG. 3, the data collection unit 31 collects data such as delay and electric field strength, which are wireless communication status parameters, as data related to communication from the plurality of mobile terminals 20.

  The model generation unit 32 generates a model for predicting the throughput using the data related to communication collected by the data collection unit 31. In addition, each time data related to communication is collected by the data collection unit 31, the model generation unit 32 analyzes the data and updates the model. For example, the model generation unit 32 performs analysis / learning by obtaining the average and variance of the communication-related data collected by the data collection unit 31, and generates a model. For example, as a model generated here, if A (for example, electric field strength, RTT (Round-Trip Time)) is known, B (throughput) is known, and a relational expression between A and B. Then, after generating the model, the model generation unit 32 transmits the generated model to the mobile terminal 20 as shown in FIG.

  Here, an example of a learning and prediction method will be described. As an example described here, for example, the learning parameters are NW types (3G, LTE, WiFi, etc.), electric field strength, RTT, and throughput collected from the mobile terminal 20 in advance, and explanatory variables (used for prediction) The NW type (3G, LTE, WiFi, etc.), electric field strength, and RTT of the mobile terminal 20 immediately before executing the application, and the objective variable (prediction result) as the throughput immediately before executing the application. . The learning / prediction method, learning parameters, explanatory variables, and objective variables can be changed as appropriate.

  Each learning parameter is collected, and regression analysis is performed to determine what the objective variable {throughput} is when the explanatory variable {NW type, electric field strength, RTT} is in a state. The result of the regression analysis is obtained as a relational expression (model) of the explanatory variable {NW type, electric field strength, RTT} and the objective variable {throughput}.

  The learning timing may be stream processing, batch processing, or processing that combines stream processing and batch processing. Which one to select can be changed and set as appropriate by the user. This stream processing (real-time analysis using Jubatus (registered trademark) or the like) is effective when the amount of collected data at that time is large because machine learning is performed using data acquired at that time. In addition, batch processing accumulates data in units of days, weeks, seasons, years, etc., and performs machine learning, so it is effective when you can understand trends for each cycle and acquire data in the long term. Analysis is possible even if the amount of data collected during that time is small. In addition, the combined processing of stream processing and batch processing is the collected data of that time that can detect anomalies by grasping the trend of each cycle in batch processing and comparing it with the result of stream processing at that moment Even a small amount can be analyzed with better accuracy.

  Further, the prediction unit 22b of the mobile terminal 20, which will be described later, measures the objective variables {NW type, electric field strength, RTT} on the spot when trying to use the application, and applies these values to the relational expression. As exemplified in FIG. 4, the objective variable {throughput distribution} from the time when the objective variable is measured to several minutes ahead may be predicted. As shown in FIG. 4, after an offload determination is made on whether to offload application processing to the arithmetic processing server 40, the measured throughput changes as time passes while the application is used. To do. For this reason, the model generation unit 32 may generate not only a model that predicts throughput at a certain point of time, but also a model that can predict the distribution of throughput during a predetermined time. Moreover, it can predict with high precision by learning for every crowded time zone.

  Returning to the description of FIG. 2, the mobile terminal 20 will be described. The mobile terminal 20 includes an application 21 and an application execution location switching unit 22 that switches an execution location of the application. The application 21 includes an arithmetic processing unit 21a that executes arithmetic processing. The arithmetic processing unit 21 a is the same as the arithmetic processing unit 41 a of the arithmetic processing server 40. When executing the application, the arithmetic processing unit 21a inquires of the control unit 22c whether to offload application processing to the arithmetic processing server 40.

  The application execution location switching unit 22 determines whether to execute the arithmetic processing for executing the application on the mobile terminal 20 or the arithmetic processing server 40 and switches the processing. The application execution location switching unit 22 includes a measurement unit 22a, a prediction unit 22b, and a control unit 22c that switch execution locations.

  The measurement unit 22a measures the communication status. Specifically, immediately before the application is executed, the measurement unit 22a measures the NW type used immediately before the application execution, the electric field strength and the RTT immediately before the application execution as information on the communication status, and the prediction unit 22b. Notify

  The prediction unit 22b predicts the throughput from the communication status measured by the measurement unit 22a using a model created based on information collected from the plurality of mobile terminals 20. For example, the prediction unit 22b applies A {NW type, electric field strength, RTT} measured by the measurement unit 22a to the model, and predicts B {throughput}.

  Further, as described above, throughput for several minutes while using an application may be predicted as a distribution. By predicting the distribution of throughput in a predetermined time zone instead of the throughput at a certain point in time, it is possible to cope with prediction when the fluctuation of the throughput value is large, such as a crowded time zone. In other words, as illustrated in FIG. 5, whether the distribution of the throughput and the offload are performed so that the determination of the offload is not frequently switched when the fluctuation of the throughput value is large, such as a busy time zone. It is possible to analyze the offload success probability (in the example of FIG. 5, the offload success rate is 80%) by comparing with the throughput threshold for making such a determination. In this way, it is not necessary to predict the throughput many times during application execution by calculating the success rate of offload when the threshold distribution is set by predicting the distribution of throughput in a predetermined time zone. There is an advantage that it is not necessary to learn every time.

  The control unit 22c determines whether the processing related to the application is to be executed by the own device or the arithmetic processing server 40 according to the throughput predicted by the prediction unit 22b, and controls the own device or the arithmetic processing server 40 to execute the processing.

  For example, the control unit 22c uses the predicted throughput to calculate an execution time when the process is executed on the server side, and the execution time required when the process is executed on the server side and the own apparatus side. The execution time required when the process is executed is compared, and control is performed so that the apparatus or the server executes the process according to the comparison result.

  If it demonstrates using the example of FIG. 6, as for the control part 22c, mobile terminal side processing time (EC) will be based on the total time of overhead (OV), server processing time (ES), and NW communication time (up and down). If it is larger, the application processing is offloaded to the arithmetic processing server 40, and if it is smaller, the processing is executed by the own apparatus. Here, the overhead is a time required for obtaining the communication time and a time required for measuring the explanatory variable, and is known in advance. Further, the mobile terminal side processing time (EC) and the server processing time (ES) are times statistically grasped from the past execution time. Moreover, the control part 22c calculates NW communication time (NT) using the following (1) formula.

  In this way, if the execution time is shorter in the arithmetic processing server 40, the arithmetic processing related to the application is offloaded to the arithmetic processing server 40. Therefore, the application execution environment switching control is performed for the purpose of improving the efficiency of the application processing. It is possible to realize.

  Further, for example, when the threshold set based on the processing time required for each application to be executed in advance is set and the throughput predicted by the prediction unit 22b exceeds the preset threshold, The processing server 40 may execute processing related to the application, and if it does not exceed, processing related to the application may be executed by the own apparatus.

  Moreover, in addition to the case where the offload is determined using the processing time as a parameter as described above, the offload is determined using the battery consumption, the accuracy required for the application (for example, the accuracy of image recognition), and the charge for each NW Judgment may be made. For example, for the purpose of power saving, the battery consumption when executed by the arithmetic processing server 40 is calculated from the measured throughput, and the battery consumption when executed by the mobile terminal 20 and the battery when executed by the arithmetic processing server 40 are calculated. You may make it judge offload by comparison with consumption. In other words, when the battery consumption is lower when executed by the arithmetic processing server 40 than when executed by the mobile terminal 20, it is offloaded and the arithmetic processing server 40 executes the process. Information on the battery consumption with respect to the throughput can be obtained by measuring how much battery is consumed in how long the application is executed.

  Further, when importance is placed on the accuracy of the application, the application request accuracy when executed by the arithmetic processing server 40 may be calculated from the measured throughput, and the offload may be determined according to the application request accuracy. In other words, when an application with low application request accuracy is activated, it is offloaded and the arithmetic processing server 40 executes the process. It is assumed that the information related to the application request accuracy is set in advance by the user.

  Further, in the case where the charge is regarded as important, the charge for the NW when executed by the processing server 40 may be calculated from the measured throughput, and the offload may be determined according to the charge for the NW. That is, when an application that requires a small NW fee is activated, it is offloaded and the processing server 40 executes the process. It is assumed that the information related to the NW fee is set in advance by the user. The various parameters described above may be optimized by machine learning.

  Returning to FIG. 2, the arithmetic processing server 40 has an application 41. If it is determined by the mobile terminal 20 to cause the server to execute processing relating to the application, the arithmetic processing unit 41 a of the application 41 executes arithmetic processing relating to the application in the mobile terminal 20 and transmits the processing result to the mobile terminal 20. .

  Note that, for example, when a major earthquake or the like occurs and an abnormal detection of a sudden deterioration in communication status is detected, throughput prediction may not be performed and it may not be determined that offloading is not performed. This makes it possible to refrain from using the NW in an emergency and to prevent the NW from being compressed by an application unrelated to safety confirmation. Moreover, since no abnormal value is used, overlearning of normal data can be prevented.

[Learning model generation processing by the system]
Next, processing by the system according to the first embodiment will be described with reference to FIG. FIG. 7 is a sequence diagram showing a processing procedure of learning model generation processing of the system according to the first embodiment.

  As shown in FIG. 7, the measurement unit 22a of each mobile terminal 20 connects to a wireless network, and measures wireless communication status parameters (delay, electric field strength, etc.) and throughput (step S101). Then, the data collection unit 31 of the prediction model generation server 30 requests each mobile terminal 20 to collect data on wireless communication status parameters and throughput (step S102). For example, the data collection unit 31 requests each mobile terminal 20 for wireless communication status parameters and throughput data by requesting automatic transmission over a NW at regular intervals.

  Then, the data collection unit 31 acquires wireless communication status parameters and throughput data from each mobile terminal 20 (step S103), accumulates the acquired data in a predetermined storage unit (step S104), and the model generation unit 32. Is notified of the acquired data (step S105). And the model production | generation part 32 learns the collected data, produces | generates a model (step S106), and notifies the produced | generated model to the prediction part 22b of the mobile terminal 20 which is using the application (step S107).

[Distributed processing by mobile devices]
Next, an example of mobile terminal distributed processing will be described with reference to FIG. FIG. 8 is a flowchart showing a processing procedure of distributed processing of the mobile terminal according to the first embodiment. As illustrated in FIG. 8, when the measurement unit 22a of the mobile terminal 20 receives a request for using an application from the user (Yes in Step S201), the measurement unit 22a measures the electric field strength and the RTT (Step S202).

  Then, the prediction unit 22b of the mobile terminal 20 predicts the throughput from the electric field strength and the RTT using the model generated by the prediction model generation server 30 (step S203). Subsequently, the control unit 22c of the mobile terminal 20 calculates an execution time when executed on the side of the arithmetic processing server 40 from the throughput (step S204).

  Then, the control unit 22c determines whether the execution on the arithmetic processing server 40 side is faster than the execution on the mobile terminal 20 side (step S205). Specifically, the control unit 22c uses the predicted throughput to calculate an execution time when the process is executed on the server side, and automatically calculates the execution time required when the process is executed on the server side. The execution time required when the process is executed on the apparatus side is compared, and control is performed so that the own apparatus or the server executes the process according to the comparison result.

  As a result, if the control unit 22c is faster to execute on the arithmetic processing server 40 side than the mobile terminal 20 side (Yes in step S205), the control unit 22c executes arithmetic processing related to the application on the arithmetic processing server 40 side ( Step S206). In addition, when the execution on the calculation processing server 40 side is slower than the execution on the mobile terminal 20 side (No at Step S205), the control unit 22c causes the mobile terminal 20 to execute the calculation process on the application (Step S207). ).

[Effect of the first embodiment]
As described above, the mobile terminal 20 measures the communication status, predicts the throughput from the measured communication status using the model created based on the information collected from the plurality of mobile terminals 20, and predicts it. In accordance with the throughput thus determined, it is determined whether the processing related to the application is to be executed by the own device or the arithmetic processing server 40, and control is performed so that the own device or the arithmetic processing server 40 executes the processing. For this reason, the mobile terminal 20 can select an appropriate application execution environment with higher accuracy in consideration of the communication environment including other terminals, etc., and can autonomously execute seamless switching control. The processing speed of the application at 20 can be improved.

  Further, according to the first embodiment, the application can be used in a state in which the processing speed is optimized by switching the place where the application is executed flexibly. Further, when the specifications of the mobile terminal 20 are poor and the communication status is good, the processing can be speeded up by performing processing on the server. In addition, when processing is performed on the arithmetic processing server side even though the specifications of the mobile terminal 20 are good, it is not necessary to use the NW wastefully by executing it on the mobile terminal 20, and the bandwidth is compressed. do not do. In addition, since throughput is predicted using only information that can be acquired by the mobile terminal 20, it is possible for companies and individuals who are not telecommunications carriers to grasp the throughput. In addition, throughput can be predicted using parameters that can be acquired in a short time in the mobile terminal 20, and an application execution location can be flexibly selected based on the predicted result.

  In addition, according to the first embodiment, using the throughput predicted by machine learning using the execution status data of other terminals based on information that can be collected on the mobile terminal 20 side, the communication environment status and the like are also taken into account. Therefore, it is possible to autonomously execute seamless switching control by selecting an appropriate application execution environment. In addition, since data is collected from a plurality of mobile terminals and a model is generated, not only data collected by a single device but also data as collective intelligence is used, so that prediction with higher accuracy is possible.

[System configuration]
In addition, among the processes described in the present embodiment, all or part of the processes described as being automatically performed can be manually performed, or the processes described as being performed manually can be performed. All or a part can be automatically performed by a known method. In addition, the processing procedures, control procedures, specific names, and information including various data and parameters shown in the above-described document and drawings can be arbitrarily changed unless otherwise specified.

  Further, each component of each illustrated apparatus is functionally conceptual, and does not necessarily need to be physically configured as illustrated. That is, the specific form of distribution / integration of each device is not limited to the one shown in the figure, and all or a part of the distribution / integration may be functionally or physically distributed in arbitrary units according to various loads or usage conditions. Can be integrated and configured.

  Further, for example, the mobile terminal 20 may include a data collection unit and a model generation unit without providing the prediction model generation server 30. Further, it is also possible to cause the external server or the like not to execute the process of predicting throughput and determining whether or not to offload, or the process of predicting throughput.

[program]
FIG. 9 is a diagram illustrating that information processing by the distributed processing program according to the disclosed technique is specifically realized using a computer. As illustrated in FIG. 9, the computer 1000 includes, for example, a memory 1010, a CPU (Central Processing Unit) 1020, a hard disk drive interface 1030, a disk drive interface 1040, a serial port interface 1050, a video adapter 1060, A network interface 1070. Each part of the computer 1000 is connected by a bus 1080.

  The memory 1010 includes a ROM 1011 and a RAM 1012 as illustrated in FIG. The ROM 1011 stores a boot program such as BIOS (Basic Input Output System). The hard disk drive interface 1030 is connected to the hard disk drive 1031 as illustrated in FIG. The disk drive interface 1040 is connected to the disk drive 1041 as illustrated in FIG. For example, a removable storage medium such as a magnetic disk or an optical disk is inserted into the disk drive 1041. The serial port interface 1050 is connected to a mouse 1051 and a keyboard 1052, for example, as illustrated in FIG. The video adapter 1060 is connected to a display 1061, for example, as illustrated in FIG.

  Here, as illustrated in FIG. 9, the hard disk drive 1031 stores, for example, an OS 1091, an application program 1092, a program module 1093, and program data 1094. That is, the distributed processing program according to the disclosed technology is stored in, for example, the hard disk drive 1031 as the program module 1093 in which a command to be executed by the computer is described. Specifically, a program module in which a procedure for executing the same information processing as each unit of the control unit 12 described in the above embodiment is described is stored in the hard disk drive 1031.

  Data used for information processing by the distributed processing program is stored as program data 1094 in, for example, the hard disk drive 1031. The CPU 1020 reads out the program module 1093 and the program data 1094 stored in the hard disk drive 1031 to the RAM 1012 as necessary, and executes various procedures.

  Note that the program module 1093 and the program data 1094 related to the distributed processing program are not limited to being stored in the hard disk drive 1031. For example, the program module 1093 and the program data 1094 may be stored in a removable storage medium. In this case, the CPU 1020 reads data via a removable storage medium such as a disk drive. Similarly, the program module 1093 and the program data 1094 related to the distributed processing program may be stored in another computer connected via a network (LAN (Local Area Network), WAN (Wide Area Network), etc.). good. In this case, the CPU 1020 reads various data by accessing another computer via the network interface.

[Others]
Note that the distributed processing program described in the present embodiment can be distributed via a network such as the Internet. The control program can also be executed by being recorded on a computer-readable recording medium such as a hard disk, a flexible disk (FD), a CD-ROM, an MO, and a DVD, and being read from the recording medium by the computer.

DESCRIPTION OF SYMBOLS 10 Network 20 Mobile terminal 21, 41 Application 22a Measurement part 22b Prediction part 22c Control part 30 Prediction model generation server 31 Data collection part 32 Model generation part 40 Arithmetic processing server

Claims (7)

A measurement unit for measuring the communication status;
Using a model created based on information collected from a plurality of terminal devices, a prediction unit that predicts throughput from the communication status measured by the measurement unit,
In accordance with the throughput predicted by the prediction unit, a control unit that determines whether to execute the process related to the application on the own device or the server, and controls the own device or the server to execute the process;
The terminal device characterized by having. A generation unit that generates a model for predicting throughput using data related to communication collected from a plurality of terminal devices;
The terminal device according to claim 1, further comprising:   The terminal device according to claim 2, wherein the generation unit analyzes the data and updates the model every time data related to communication is collected by the collection unit.   The control unit uses the throughput predicted by the prediction unit to calculate an execution time when the process is executed on the server side, and the execution time required when the process is executed on the server side and the own device The execution time required when the process is executed on the side is compared, and control is performed so that the apparatus or the server executes the process according to the comparison result. The terminal device according to any one of the above. A distributed processing method executed in a terminal device,
A measurement process for measuring the communication status;
Using a model created based on information collected from a plurality of terminal devices, a prediction step of predicting throughput from the communication status measured by the measurement step,
In accordance with the throughput predicted by the prediction step, it is determined whether to execute the process related to the application on the own device or the server, and the control step for controlling the own device or the server to execute the process;
The distributed processing method characterized by including. A measurement step for measuring the communication status;
Using a model created based on information collected from a plurality of terminal devices, predicting the throughput from the communication status measured by the measuring step;
A control step of determining whether to execute the process related to the application in the own apparatus or the server according to the throughput predicted in the prediction step, and controlling the own apparatus or the server to execute the process;
A distributed processing program characterized by causing a computer to execute. A distributed processing system comprising a terminal device and a server,
The terminal device
A measurement unit for measuring the communication status;
Using a model created based on information collected from a plurality of terminal devices, a prediction unit that predicts throughput from the communication status measured by the measurement unit,
According to the throughput predicted by the prediction unit, determine whether to execute the process related to the application to the own device or the server, and a control unit to control the own device or the server to execute the process.
The server
A distributed processing system, comprising: a processing unit configured to execute a process related to an application by the terminal device and to transmit the processing result to the terminal device when it is determined to cause the server to execute the process related to the application.

Images