Demand prediction device, demand prediction method, and program

Abstract

A demand prediction device (104) is disposed in a central management data center (100) in a wireless communication network (1000) composed of the central management data center, a plurality of edge locations (201, 202, 203) in each of which a container image of a multi-access edge computing application is distributed from the central management data center, and a network (300) which connects the central management data center and each of the plurality of edge locations. The demand prediction device comprises: a generation unit (1041); a prediction unit (1042); and a disposition unit (1043).

Description

Demand forecasting device, demand forecasting method, and program

[0001] The present invention relates to a demand forecasting device, a demand forecasting method, and a program.

[0002] Multi-access Edge Computing (MEC) applications (hereinafter referred to as "MEC apps") are known (see Patent Literatures 1, 2, and 3). When a user connects to an MEC app running at an edge location close to the user's location, latency is reduced. This provides benefits to the user, such as an improved user experience and smoother operation of game apps that require strict latency.

[0003] Japanese Patent Publication No. 2023-17762 Japanese Patent Publication No. 2022-54417 Japanese Patent Publication No. 2019-62510

[0004] None of Patent Documents 1, 2, and 3 disclose a configuration in which demand for MEC applications is predicted for each edge location based on past usage patterns, and container images of the MEC applications are placed at the corresponding edge locations according to the predicted demand.

[0005] One aspect of the present invention aims to predict demand for MEC applications for each edge location based on past usage patterns, and to deploy container images of the MEC applications at the corresponding edge locations according to the predicted demand.

[0006] In order to solve the above problem, one aspect of the present invention provides a demand prediction device that is located in a central management data center in a wireless communication network that is composed of a central management data center, a plurality of edge locations where a container image of a multi-access edge computing application is deployed from the central management data center, and a network connecting the central management data center and each of the plurality of edge locations, and includes: a generation unit that generates a table showing past usage of the multi-access edge computing application from each of the plurality of edge locations; a prediction unit that predicts the demand for the multi-access edge computing application at each of the plurality of edge locations from the table generated by the generation unit; and a placement unit that places the container image of the multi-access edge computing application at the corresponding edge location in accordance with the demand for the multi-access edge computing application at each of the plurality of edge locations predicted by the prediction unit.

[0007] A demand forecasting method according to one aspect of the present invention is a demand forecasting method executed by a demand forecasting device located in a central management data center in a wireless communication network consisting of a central management data center, a plurality of edge locations where a container image of a multi-access edge computing application is deployed from the central management data center, and a network connecting the central management data center and each of the plurality of edge locations, and includes a generation step of generating a table showing past usage of the multi-access edge computing application from each of the plurality of edge locations, a prediction step of predicting the demand for the multi-access edge computing application at each edge location from the table generated in the generation step, and a deployment step of deploying the container image of the multi-access edge computing application at the corresponding edge location according to the demand for the multi-access edge computing application at each of the plurality of edge locations predicted in the prediction step.

[0008] The demand forecasting device according to each aspect of the present invention may be realized by a computer. In this case, the demand forecasting device program that causes the computer to operate as each part (software element) of the demand forecasting device, thereby realizing the demand forecasting device on the computer, and the computer-readable recording medium on which the program is recorded, also fall within the scope of the present invention.

[0009] According to one aspect of the present invention, the demand for an MEC application can be predicted for each edge location based on past usage, and the container image of the MEC application can be placed at the corresponding edge location according to the predicted demand.

[0010] FIG. 1 is a diagram illustrating an example of a wireless communication network configuration according to an embodiment of the present invention. FIG. 2 is a diagram for explaining an example of a functional configuration of a demand prediction device according to the embodiment. FIG. 3 is a flowchart illustrating a processing procedure of a demand prediction method according to the embodiment. FIG. 4 is a diagram illustrating a specific example of an edge location table according to the embodiment. FIG. 5 is a diagram illustrating a specific example of an MEC application table according to the embodiment. FIG. 6 is a diagram illustrating a specific example of an MEC application launch information table according to the embodiment. FIG. 7 is a diagram illustrating a specific example of an aggregation result table according to the embodiment. FIG. 8 is a diagram illustrating a specific example of an MEC application placement table according to the embodiment.

[0011] Hereinafter, embodiments of the present invention will be described in detail.

[0012] (Wireless Communication Network) Fig. 1 is a schematic diagram showing the general configuration of a wireless communication network 1000 according to this embodiment. As shown in Fig. 1, the wireless communication network 1000 includes a central management data center 100, a first edge location 201, a second edge location 202, a third edge location 203, and a network 300.

[0013] 1 illustrates three edge locations, namely, a first edge location 201, a second edge location 202, and a third edge location 203, as edge locations included in the wireless communication network 1000. However, the number of edge locations is not limited to three, and the wireless communication network 1000 may include a plurality of edge locations (not illustrated) in addition to the first edge location 201, the second edge location 202, and the third edge location 203. The first edge location 201, the second edge location 202, and the third edge location 203 are examples of edge locations included in the wireless communication network 1000. Note that, hereinafter, the edge locations included in the wireless communication network 1000, including the first edge location 201, the second edge location 202, and the third edge location 203, will be collectively referred to as "edge location X."

[0014] (Central Management Data Center) The central management data center 100 is the core of the cloud computing service and provides large-scale data processing, storage, and advanced application services. As shown in FIG. 1, the central management data center 100 includes a 5G core 101, a user DB 102, a container image storage unit 103, and a demand prediction device 104.

[0015] The 5G core 101 is a server equipped with a high-performance processor, large-capacity RAM, and a high-speed storage system to efficiently process large amounts of data and perform complex calculations.

[0016] Each of the user DB 102 and the container image storage unit 103 is configured from a large-capacity storage system such as an HDD or SSD in order to store large amounts of data. As shown in FIG. 1, the user DB 102 and the container image storage unit 103 are illustrated as separate large-capacity storage systems. However, in this embodiment, the user DB 102 and the container image storage unit 103 may be configured from the same large-capacity storage system. Note that, below, this embodiment will be described using an example in which the user DB 102 and the container image storage unit 103 are configured from separate large-capacity storage systems.

[0017] The user DB 102 stores an edge location table, an MEC application table, an MEC application activation information table, a counting result table, and an MEC application placement table, which are shown in Figures 4 to 8 and will be described later. These five tables will be described later.

[0018] The container image storage unit 103 stores container images of MEC applications that can be launched at the edge location X. The same number of container images as the number of MEC applications that can be executed at the edge location X are stored. In other words, one container image is prepared for one MEC application. Note that the container image is a package of code, libraries, setting files, and the like required to execute the MEC application.

[0019] The demand prediction device 104 predicts the demand for the MEC application from edge location X for each edge location based on past usage conditions, and places a container image of the MEC application at the corresponding edge location according to the predicted demand.

[0020] (Edge Location) Edge location X is a facility for performing data processing at a location close to the user device 500, enabling the provision of services with low latency. Below, edge location X will be described using the first edge location 201 as an example. Note that in the following description of the first edge location 201, by replacing the component number assigned to each component, or a part thereof, "201," with "202," the description can be understood as a description of the second edge location 202. Furthermore, in the description of the first edge location 201, by replacing the component number assigned to each component, or a part thereof, "201," with "203," the description can be understood as a description of the third edge location 203.

[0021] As shown in FIG. 1, the first edge location 201 includes an MEC server 2011 , an image repository 2012 , and a base station 2013 .

[0022] The MEC server 2011 provides an execution environment for MEC applications. The MEC server 2011 enables data processing and service provision to be performed in a location close to the user equipment 500, thereby significantly reducing latency and improving communication efficiency. The MEC server 2011 may further have a RAN Intelligent Controller (RIC) function.

[0023] The image repository 2012 is a system for storing and managing container images of MEC applications. The MEC server 2011 obtains container images of MEC applications from the image repository 2012 and generates and executes instances of the MEC applications. This generation and execution enables rapid deployment and update of MEC applications.

[0024] The base station 2013 provides a communication interface with the user equipment 500. The base station 2013 realizes low-latency access to the user equipment 500 and enables efficient data processing. Note that while only one user equipment 500 is illustrated in FIG. 1 , the base station 2013 can provide a communication interface with multiple user equipments, including the user equipment 500. Also, as illustrated in FIG. 1 , one base station 2013 is illustrated as the base station included in the first edge location 201. However, the number of base stations included in the first edge location 201 is not limited to one, and multiple base stations may be included in addition to the base station 2013. Hereinafter, this embodiment will be described using an example in which the base station included in the first edge location 201 is the base station 2013.

[0025] The network 300 connects the centrally managed data center 100 and the edge location X. The network 300 may be, for example, the Internet, a dedicated line independent of the Internet, a cellular network such as 5G, or a combination thereof. The network 300 provides efficient and secure data communication between the centrally managed data center 100 and the edge location X.

[0026] (Demand prediction device) Fig. 2 is a diagram for explaining an example of the functional configuration of the demand prediction device 104. Note that Fig. 2 also illustrates other components of the wireless communication network 1000 that are necessary for explaining the example of the functional configuration of the demand prediction device 104.

[0027] As shown in FIG. 2 , the demand prediction device 104 includes a generation unit 1041 , a prediction unit 1042 , and an arrangement unit 1043 .

[0028] The generation unit 1041 generates a table indicating the past usage status of the MEC application from the edge location X. Details of the table will be described later when describing the demand prediction method executed by the demand prediction device 104.

[0029] The prediction unit 1042 predicts the demand for the MEC application for each edge location X from the table generated by the generation unit 1041.

[0030] The placement unit 1043 places the container image of the MEC application at the corresponding edge location in accordance with the demand for the MEC application at each edge location X predicted by the prediction unit 1042.

[0031] (Demand Forecasting Method) Next, a demand forecasting method executed by the demand forecasting device 104 will be described. Here, first, a procedure for deploying a container image of an MEC application from the central management data center 100 to the first edge location 201 (hereinafter referred to as a "forecast-free deployment procedure") will be described when a request to launch an MEC application is received from a user device 500 connected to the first edge location 201. The forecast-free deployment procedure is a procedure for deploying a container image of an MEC application from the central management data center 100 to the first edge location 201 when the demand forecasting device 104 does not execute the demand forecasting method.

[0032] First, the first edge location 201 receives a request to launch an MEC app from the user equipment 500. Next, based on the received launch request, the first edge location 201 requests a container image of the required MEC app from the container image storage unit 103 in the central management data center 100. Next, after receiving the request from the first edge location 201, the central management data center 100 retrieves the container image of the specified MEC app from the container image storage unit 103 and deploys it to the image repository 2012 of the first edge location 201. Next, the first edge location 201 deploys the container image received from the central management data center 100 and launches an instance of the MEC app. Finally, when the deployment of the container image is complete, the MEC app is launched on the first edge location 201, and the first edge location 201 is ready to provide the MEC app to the user equipment 500. After the MEC app is launched, the first edge location 201 begins providing the MEC app to the user equipment 500.

[0033] As a result of the above, the container image of the MEC application is deployed from the central management data center 100 to the first edge location 201, and the MEC application can be provided based on a startup request from the user device 500.

[0034] In the above-described unpredictable deployment procedure, when an MEC app is launched, the first edge location 201 deploys a container image of the MEC app from the central management data center 100, deploys the container image, and launches the MEC app. Because the container image is deployed at launch, the MEC app may launch slowly, which may impair the user experience when the user launches the MEC app.

[0035] It is also possible to keep the MEC application running at all times on the first edge location 201. However, if the MEC application is kept running at all times, the resources of the first edge location 201 may become strained as the number of types of MEC applications deployed on the first edge location 201 increases.

[0036] In contrast, when the demand prediction device 104 executes the above-described demand prediction method, it predicts the demand for MEC apps from the first edge location 201 based on past usage, and pre-arranges container images of the MEC apps at the first edge location 201 according to the predicted demand. Because container images of MEC apps whose demand is predicted at the first edge location 201 are pre-arranged at the first edge location 201, it is no longer necessary to pre-launch all MEC apps at the first edge location 201, thereby reducing resources at the first edge location 201. Furthermore, it is possible to reduce traffic caused by deploying container images from the central management data center 100 every time a request to launch an MEC app is made. Furthermore, MEC apps stored in the image repository 2012 at the first edge location 201 can be launched quickly.

[0037] The demand forecasting method executed by the demand forecasting device 104 will be specifically described below with reference to FIGS.

[0038] 3, the generation unit 1041 shown in Fig. 2 acquires an edge location table A, an MEC application table B, and an MEC application activation information table C from the user DB 102. Fig. 4 is a diagram showing a specific example of the edge location table A, Fig. 5 is a diagram showing a specific example of the MEC application table B, and Fig. 6 is a diagram showing a specific example of the MEC application activation information table C.

[0039] As shown in FIG. 4 , edge location table A is a table consisting of an "edge ID" item, an "area" item, and an "area characteristic" item. Edge location table A associates the contents of the "edge ID" item, the "area" item, and the "area characteristic" item. The "edge ID" item is an identification number assigned in advance to each edge location X. In the following, edge ID "1" is the edge ID of the first edge location 201 shown in FIG. 1. Edge ID "2" is the edge ID of the second edge location 202 shown in FIG. 1. Edge ID "3" is the edge ID of the third edge location 203 shown in FIG. 1. Edge ID "4" is the edge ID of an edge location X not shown in FIG. 1. The "area" item is the area in which each edge location X is located. The "area characteristic" item is the characteristic of the area in which each edge location X is located.

[0040] As shown in FIG. 5 , MEC app table B is a table consisting of an item "MEC app ID" and an item "MEC app name." MEC app table B links the contents of the item "MEC app ID" with the item "MEC app name." The item "MEC app ID" is an identification number assigned in advance to each MEC app whose container image can be deployed from the central management data center 100. The item "MEC app name" is the name of each of the above-mentioned MEC apps.

[0041] As shown in FIG. 6 , the MEC application launch information table C is a table consisting of the items "launch date and time", "launch user", "launch edge", and "launched MEC application". The MEC application launch information table C links the contents of the items "launch date and time", "launch user", "launch edge", and "launched MEC application". The item "launch date and time" is the date and time when each MEC application that can be deployed as a container image from the central management data center 100 was actually launched in the past. The item "launch user" is the user who launched the MEC application that was actually launched in the past. The item "launch edge" is the edge location X where the MEC application was actually launched in the past. The item "launched MEC application" is the MEC application ID of the MEC application that was actually launched in the past.

[0042] The generation unit 1041 generates a counting result table D using the edge location table A, the MEC application table B, and the MEC application activation information table C acquired from the user DB 102. FIG. 7 is a diagram showing a specific example of the counting result table D.

[0043] 7, the tally result table D is a table including the items “activation date and time,” “activation user,” “activation edge,” and “activation MEC application.” The items in the tally result table D are the same as the items in the MEC application activation information table C shown in FIG. 6, and therefore, description thereof will not be repeated.

[0044] 6 is different from the MEC application launch information table C shown in Fig. 6 in that duplicate MEC application launch information included in the MEC application launch information table C that corresponds to launches by the same user, that correspond to launches on the same edge location X, and that correspond to launches of the same MEC application is eliminated. Specifically, among the MEC application launch information included in the MEC application launch information table C of Fig. 6, MEC application launch information including launch date and time "2023 / 11 / 1 10:35", launching user "B", launch edge "Okinawa", and launched MEC application "app3" is duplicated with MEC application launch information including launch date and time "2023 / 11 / 1 10:38", launching user "B", launch edge "Okinawa", and launched MEC application "app3". 7, the duplicates are eliminated, and the tally result table D does not include MEC application activation information consisting of activation date and time "2023 / 11 / 1 10:38", activation user "B", activation edge "Okinawa", and activation MEC application "application 3". The generation unit 1041 performs the above-mentioned duplicate elimination process by referring to the edge location table A shown in FIG. 4 and the MEC application table B shown in FIG. 5.

[0045] As described above, the generation unit 1041 generates the MEC application launch information table C and the aggregation result table D in a format that enables demand forecast processing of the MEC application from the edge location X.

[0046] 3 , the prediction unit 1042 processes a demand prediction for the MEC application from the edge location X using various statistical methods based on the aggregation result table D generated by the generation unit 1041. The prediction unit 1042 may also process the demand prediction for the MEC application by taking into account the regional characteristics of the edge location X, information and data available on the Internet, such as information about external events held in the area of ​​the edge location X.

[0047] The prediction unit 1042 generates an MEC application arrangement table E, which is a result of processing the above-mentioned demand prediction. FIG. 8 is a diagram showing a specific example of the MEC application arrangement table E.

[0048] As shown in FIG. 8 , the MEC application placement table E is a table consisting of an "edge region", an "placed MEC app", and an "placement period". The MEC application placement table E links the contents of the "edge region", "placed MEC app", and "placement period". The "edge region" item is the region where each edge location X for which demand for an MEC app is predicted is located. The "placed MEC app" item is an MEC app for which demand from the edge location X is predicted. The "placement period" item is the period for which the MEC app is placed at each edge location X for which demand for the MEC app is predicted.

[0049] 3, the placement unit 1043 determines the placement of the container image at the edge location X in accordance with the MEC application placement table E created by the prediction unit 1042. According to the example of FIG. 8, for example, the container image of the placed MEC application "motion detection application" will be placed in the image repository 2012 of the first edge location 201 placed in the edge region "Hokkaido" during the placement period "2023 / 11 / 1-11 / 30."

[0050] As described above, the placement unit 1043 can place container images according to the demand forecast for MEC applications.

[0051] (Summary) A demand prediction device according to aspect 1 of the present invention is a demand prediction device disposed in a central management data center in a wireless communication network made up of a central management data center, a plurality of edge locations where a container image of a multi-access edge computing application is deployed from the central management data center, and a network connecting the central management data center and each of the plurality of edge locations, and comprises: a generation unit that generates a table showing past usage of the multi-access edge computing application from each of the plurality of edge locations; a prediction unit that predicts, for each of the plurality of edge locations, the demand for the multi-access edge computing application at each edge location from the table generated by the generation unit; and a placement unit that places the container image of the multi-access edge computing application at the corresponding edge location according to the demand for the multi-access edge computing application at each of the plurality of edge locations predicted by the prediction unit.

[0052] According to the above configuration, the demand for a multi-access edge computing application can be predicted for each edge location based on past usage, and a container image of the multi-access edge computing application can be placed at the corresponding edge location according to the predicted demand.

[0053] In the demand forecasting device according to aspect 2 of the present invention, in the above-described aspect 1, the table generated by the generation unit is a table linking the date and time when each multi-access edge computing application that can be deployed as a container image from the central management data center was actually launched in the past, the user who launched the multi-access edge computing application that was actually launched in the past, the edge location where the multi-access edge computing application was actually launched in the past, and the multi-access edge computing application that was actually launched in the past, and duplicate information corresponding to launches by the same user, launches at the same edge location, and launches of the same multi-access edge computing application is eliminated.

[0054] According to the above configuration, it is possible to generate a table in a format that enables demand forecast processing for multi-access edge computing applications from edge locations.

[0055] A demand forecasting method according to aspect 3 of the present invention is a demand forecasting method executed by a demand forecasting device located in a central management data center in a wireless communication network consisting of a central management data center, a plurality of edge locations where a container image of a multi-access edge computing application is deployed from the central management data center, and a network connecting the central management data center and each of the plurality of edge locations, and includes a generation step of generating a table showing past usage of the multi-access edge computing application from each of the plurality of edge locations, a prediction step of predicting the demand for the multi-access edge computing application at each edge location from the table generated in the generation step, and a deployment step of deploying the container image of the multi-access edge computing application at the corresponding edge location according to the demand for the multi-access edge computing application at each of the plurality of edge locations predicted in the prediction step.

[0056] According to the above configuration, the same effects as those of the first aspect can be obtained.

[0057] The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the present invention. Furthermore, new technical features can be formed by combining the technical means disclosed in each embodiment.

[0058] REFERENCE SIGNS LIST 100 Central management data center 104 Demand forecasting device 201 First edge location 202 Second edge location 203 Third edge location 300 Network 1000 Wireless communication network 1041 Generation unit 1042 Prediction unit 1043 Placement unit

Claims

1. A demand prediction device located in a central management data center in a wireless communication network consisting of a central management data center, a plurality of edge locations where a container image of a multi-access edge computing application is deployed from the central management data center, and a network connecting the central management data center and each of the plurality of edge locations, the demand prediction device comprising: a generation unit that generates a table showing past usage of the multi-access edge computing application from each of the plurality of edge locations; a prediction unit that predicts the demand for the multi-access edge computing application at each of the plurality of edge locations from the table generated by the generation unit; and a placement unit that places the container image of the multi-access edge computing application at the corresponding edge location in accordance with the demand for the multi-access edge computing application at each of the plurality of edge locations predicted by the prediction unit.

2. The demand forecasting device of claim 1, wherein the table generated by the generation unit is a table linking the date and time when each multi-access edge computing application capable of deploying container images from the central management data center was actually launched in the past, the user who launched the multi-access edge computing application that was actually launched in the past, the edge location where the multi-access edge computing application was actually launched in the past, and the multi-access edge computing application that was actually launched in the past, and wherein duplicate information corresponding to launches by the same user, launches at the same edge location, and launches of the same multi-access edge computing application is eliminated.

3. A demand forecasting method executed by a demand forecasting device located in a central management data center in a wireless communication network consisting of a central management data center, a plurality of edge locations where a container image of a multi-access edge computing application is deployed from the central management data center, and a network connecting the central management data center and each of the plurality of edge locations, the demand forecasting method including: a generation step of generating a table showing past usage of the multi-access edge computing application from each of the plurality of edge locations; a prediction step of predicting, for each of the plurality of edge locations, the demand for the multi-access edge computing application at each edge location from the table generated in the generation step; and a deployment step of deploying the container image of the multi-access edge computing application at the corresponding edge location in accordance with the demand for the multi-access edge computing application at each of the plurality of edge locations predicted in the prediction step.

4. A program for causing a computer to function as the demand forecasting device according to claims 1 and 2, the program causing a computer to function as the generation unit, the forecasting unit and the placement unit.

Images