WO2024095745A1

WO2024095745A1 - Information processing device, and information processing method

Info

Publication number: WO2024095745A1
Application number: PCT/JP2023/037327
Authority: WO
Inventors: 歩小林
Original assignee: ソニーグループ株式会社
Priority date: 2022-11-04
Filing date: 2023-10-16
Publication date: 2024-05-10
Also published as: JP2024067227A

Abstract

The present disclosure relates to an information processing device and an information processing method capable of offloading a more appropriate target. Provided is an information processing device comprising: a management unit for managing a first node, which is a computer at an edge site connected to a mobile communication network, and a second node, which is a computer at a global site connected to the Internet; a determining unit which, if a first resource value indicating a computing resource at the first node exceeds a threshold, determines a movement priority level, with the first node as a movement source node, on the basis of detailed information relating to details of an application container operating at the movement source node, and a second resource value indicating the computing resource; and a control unit for performing control to move the application container, in accordance with the priority level, from the movement source node to a movement destination node, which is the second node serving as a movement destination. The present disclosure is applicable to equipment constituting an MEC management system, for example.

Description

Information processing device and information processing method

This disclosure relates to an information processing device and an information processing method, and in particular to an information processing device and an information processing method that enable offloading of more appropriate targets.

In recent years, research and development into technology related to Multi-access Edge Computing (MEC) has been progressing. MEC is a technology in which edges (edge servers) are distributed close to wireless terminals (UE: User Equipment) and as much processing as possible is carried out on the edge servers.

An example of technology related to MEC is the technology disclosed in Patent Document 1. Patent Document 1 discloses a technology in which a management system predicts the number of UEs using edge applications for the finite resources of an MEC host, and based on the results of the prediction, predicts the degree of congestion of computing resources of an edge server, and switches the connection destination of some of the UEs to a cloud application based on the predicted result of the congestion.

JP 2019-62510 A

However, the technology disclosed in Patent Document 1 treats application programs and UEs uniformly based on the correlation between the number of UEs and computing, but does not take into account the impact that application programs that impose specific loads have on the computing resources of the MEC. As a result, there is a possibility that the targets for offloading are not appropriately determined.

This disclosure was made in light of these circumstances, and makes it possible to offload more appropriate targets.

An information processing device according to one aspect of the present disclosure includes a management unit that manages a first node, which is a computer at an edge site connected to a mobile communication network, and a second node, which is a computer at a global site connected to the Internet; a determination unit that, when a first resource value indicating the computing resources of the first node exceeds a threshold, determines a priority for moving the application container based on detailed information on the details of an application container consisting of multiple containers operating on the source node and a second resource value indicating the computing resources of the application container, with the first node whose first resource value exceeds the threshold as the source node, and a control unit that controls the movement of the application container from the source node to a destination node, which is the second node to which the application container is to be moved, according to the priority.

In one aspect of the information processing method of the present disclosure, an information processing device manages a first node, which is a computer at an edge site connected to a mobile communication network, and a second node, which is a computer at a global site connected to the Internet, and when a first resource value indicating the computing resources of the first node exceeds a threshold, the information processing device determines a priority for moving the application container based on detailed information on the details of an application container consisting of multiple containers running on the source node and a second resource value indicating the computing resources of the application container, and controls the movement of the application container from the source node to a destination node, which is the second node to which the application container is to be moved, according to the priority.

In one aspect of the information processing device and information processing method of the present disclosure, a first node, which is a computer at an edge site connected to a mobile communications network, and a second node, which is a computer at a global site connected to the Internet, are managed, and when a first resource value indicating the computing resources of the first node exceeds a threshold, the first node whose first resource value exceeds the threshold is set as a source node, and a priority for moving the application container is determined based on detailed information regarding the details of an application container consisting of multiple containers operating on the source node and a second resource value indicating the computing resources of the application container, and control is performed to move the application container from the source node to a destination node, which is the second node to which the application container is to be moved, according to the priority.

Note that the information processing device of one aspect of the present disclosure may be an independent device or an internal block constituting a single device.

FIG. 1 is a diagram illustrating an example of the configuration of a system to which the present disclosure is applied. 2 is a block diagram showing a configuration example of the MEC management system of FIG. 1. 3 is a block diagram showing an example of a functional configuration of an MEC control unit in FIG. 2. FIG. 13 is a diagram illustrating an example of a node table. FIG. 13 illustrates an example of a pod movement table. FIG. 13 is a diagram illustrating an example of a UE mobility table. FIG. 13 is a diagram illustrating an example of a priority table. FIG. 11 is a diagram showing a first other example of the priority table. FIG. 11 is a diagram showing a second example of a priority table; FIG. 11 is a diagram showing another third example of the priority table. FIG. 11 is a diagram illustrating an example of a sequence when a migration setting process is executed. FIG. 11 is a diagram illustrating an example of a sequence when a movement control process is executed. 13 is a flowchart illustrating the flow of a movement setting process. 13 is a flowchart illustrating the flow of a movement control process. FIG. 1 is a block diagram illustrating an example of the configuration of a computer.

<System Configuration>
Fig. 1 is a diagram showing an example of the configuration of a system to which the present disclosure is applied. In Fig. 1, the system to which the present disclosure is applied is composed of a MEC (Multi-access Edge Computing) management system 11, an MEC 12, a carrier network 13, a cloud data center 14, a name server 15, and a wireless terminal 16.

The MEC management system 11 includes an information processing device that has a function of controlling the devices that make up the system in FIG. 1. The information processing device is composed of a computer such as a server. The MEC management system 11 may be provided as an on-premise type or as a cloud type.

The MEC management system 11 manages the nodes placed in the MEC 12 and the cloud data center 14, the pods that operate (run) on the nodes, and the exit of the carrier network 13 when connecting to the nodes using DNAI (DN Access Identifier) defined by 3GPP (Third Generation Partnership Project) (registered trademark).

The MEC management system 11 performs management by exchanging control signals (C-plain) between nodes located at at least one edge site and nodes located at at least one global site. For example, a VPN (Virtual Private Network) can be used for this communication path. The nodes at the edge sites are nodes located in the MEC 12, and the nodes at the global sites are nodes located in the cloud data center 14. The detailed configuration of the MEC management system 11 is explained in Figure 2. Details of the pods will also be explained later.

MEC12 is configured with devices that distribute edge servers near wireless terminals 16 and perform as much processing as possible at the edge servers. Edge site node 21A is placed in MEC12. Node 21A is configured with computers such as servers (edge servers).

The carrier network 13 is a mobile communication network constructed by a mobile communication operator (communication carrier) and is composed of a core network, etc. For example, the carrier network 13 includes a 4G network corresponding to the fourth generation mobile communication system (4G) and a 5G network corresponding to the fifth generation mobile communication system (5G). The group of servers that make up the core network of the carrier network 13 includes a NEF (Network Exposure Function) server. Note that while FIG. 1 shows an example in which the MEC 12 is placed outside the carrier network 13, the MEC 12 only needs to be connected to the carrier network 13 and may be placed inside the carrier network 13. The carrier network 13 can be connected to the Internet and is connected to the cloud data center 14 via the Internet.

Cloud data center 14 is a cloud that provides a network environment for service providers to prepare and use servers and equipment, or a data center that provides space and facilities for collecting and managing servers in one place. Cloud data center 14 is connected to carrier network 13 via the Internet. Node 21B of the global site is located in cloud data center 14. Node 21B is composed of computers such as servers (cloud servers). Hereinafter, when there is no need to distinguish between node 21A and node 21B, they will also be referred to as node 21.

The name server 15 is a server that has a DNS (Domain Name System) function that converts domain names and IP addresses to perform name resolution. The name server 15 may be provided in any communication network as long as it can be connected to the MEC management system 11 or the wireless terminal 16, and may be provided as a cloud, for example.

The wireless terminal 16 is a mobile communication terminal such as a smartphone or tablet terminal. The wireless terminal 16 communicates via the carrier network 13. The wireless terminal 16 can also connect to the name server 15 to perform name resolution. Hereinafter, the wireless terminal 16 will also be referred to as UE (User Equipment) as appropriate.

Note that while FIG. 1 shows a case where there is only one MEC 12 and one cloud data center 14, there may be multiple of each. For example, when there are multiple cloud data centers 14, this is the case when there are multiple instances on public clouds, private clouds, or data centers, or when instances (nodes) are located on multiple different public clouds, private clouds, or data centers. In addition, when there are multiple MECs 12, this is the case when there are different edge sites.

In addition, in FIG. 1, a typical configuration would be one in which there are multiple wireless terminals 16 such as smartphones, but to simplify the explanation, a case in which there is one wireless terminal 16 is shown as an example. In FIG. 1, the name server 15 is not a required component and is provided as needed.

FIG. 2 is a block diagram showing an example of the configuration of the MEC management system 11 in FIG. 1. In FIG. 2, the MEC management system 11 is composed of an MEC control unit 111, an API (Application Programming Interface) unit 112, an AF (Application Function) unit 113, and a database 114.

The MEC control unit 111 cooperates with the API unit 112 and the AF unit 113 to perform processes related to the node 21 and the pods 31 operating (running) on the node 21, as well as route control for the wireless terminal 16 (UE). The node 21 is placed in the MEC 12 or the cloud data center 14. The node 21 is configured as hardware, and is capable of operating multiple pods 31 executed as software.

The pod 31 is an application program (hereinafter also referred to as an application container) that is composed of multiple containers. In other words, the pod 31 can be said to be a collection of multiple containers. A container is a unit in which software is packaged, and by using container-based virtualization technology, virtualization is achieved by constructing an application environment in an independent space called a container. For example, a container contains things necessary for software execution, such as libraries, system tools, code, and runtime.

The MEC control unit 111 also works in conjunction with the database 114 to refer to and update tables, perform movement setting processing, movement control processing, and other processing. The MEC control unit 111 has a function of replacing the IP address of the DNS record when the pod 31 moves by controlling a communication I/F (not shown) and connecting to the name server 15. However, this IP address replacement function is not a required function.

For example, if Route53 from AWS (Amazon Web Services) (trademark) is used as the name server 15, the aws route53 change-resource-record-sets command can be rewritten using JSON (JavaScript Object Notation) that describes the contents of the record. AWS is a cloud computing service provided by Amazon Web Services, Inc. JSON is a data description language and a format for text-based data conversion.

The API unit 112 is configured as an API server that uses the API to communicate with the nodes 21 and the pods 31 on the control plane. The API can use the Kubernetes API. Kubernetes is an open source platform for orchestrating containers. Kubernetes defines the smallest deployable unit of computing that can be created and managed as a pod. The pod 31 corresponds to the pod defined in Kubernetes.

The API unit 112 can use the kubectl command in conjunction with the Kubernetes API to, for example, obtain computing resources of the node 21 and the pod 31, obtain detailed information (qosClass) about the pod 31, and specify a manifest file to deploy to or delete from the node 21. The manifest file is a text file that describes resource information about the pod 31.

The qosClass of Pod 31 is classified into three types: Guaranteed, Burstable, and BestEffort. Guaranteed applies when Requests/Limits are set for the CPU (Central Processing Unit) and memory of all containers in the pod, and Requests = Limits. Burstable applies when Requests are set for the CPU or memory of one or more containers in the pod, and Guaranteed is not met. BestEffort applies when Requests/Limits are not set for all containers in the pod.

Requests represents the minimum resource capacity (CPU and memory capacity) required by pod 31. Limits represents the maximum resource capacity (maximum CPU and memory capacity) that can be used when the load on pod 31 increases. In this disclosure, Burstable and BestEffort pods 31 are targets for movement (targets for offloading).

The AF unit 113 performs processing in cooperation with the NEF server 22 using the 3GPP API. The 3GPP API is a function described in 3GPP TS 29.517 that cooperates with an API when a communications carrier provides a core node function of the carrier network 13 to a third party such as a service provider as an API. In this disclosure, the AF unit 113 cooperates with the Traffic Influence API described in 3GPP TS 29.522 5.4 to control the route of a UE accessing a node that has been mapped in advance to a DNAI.

The NEF server 22 exposes the functions of the carrier network 13 (NF: Network Function) so that they can be used by external application programs, allowing applications to grasp information from the network, and also provides an API that allows application programs to control the NF.

The database 114 is recorded in a storage device such as a hard disk drive (HDD). The database 114 may be provided by a database (DB) server or the like. The database 114 manages various tables used in processing by the MEC control unit 111.

FIG. 3 is a block diagram showing an example of the functional configuration of the MEC control unit 111 in FIG. 2. In FIG. 3, the MEC control unit 111 has a management unit 121, a setting unit 122, and a control unit 123.

The management unit 121 manages the nodes 21 and pods 31 using various tables managed in the database 114. The management unit 121 also manages the exit of the carrier network 13 when connecting to the node 21 by mapping the DNAI and the node 21 in advance using the tables managed in the database 114.

The setting unit 122 performs a movement setting process for setting the movement of pods and UEs. In this movement setting process, the priority (movement priority) of the pod movement is determined, and the pods and UEs that are candidates for movement are set with the determined priority. Hereinafter, a pod that is a candidate for movement is also referred to as a candidate pod for movement, and a UE that is a candidate for movement is also referred to as a candidate UE for movement.

The setting unit 122 has a movement setting unit 131 and a priority determination unit 132. The movement setting unit 131 performs settings related to the movement of candidate pods to be moved and candidate UEs to be moved. The priority determination unit 132 determines the priority of the movement of pods.

The control unit 123 performs a movement control process for controlling the movement of pods and UEs. In this movement control process, a pod to be moved and a UE to be moved are determined from among the candidate pods to be moved and the candidate UEs to be moved according to the movement priority (movement priority), and the determined pod to be moved and UE to be moved are moved. Hereinafter, the pod that is actually selected as the movement target among the candidate pods to be moved is also referred to as the movement target pod. The UE that is actually selected as the movement target among the candidate UEs to be moved is also referred to as the movement target UE.

The control unit 123 has a movement control unit 141 and a route control unit 142. The movement control unit 141 controls the API unit 112 to control the movement of the pod to be moved. The route control unit 142 controls the AF unit 113 to control the route of the UE to be moved.

The database 114 manages a node table 151, a pod movement table 152, a UE movement table 153, and a priority table 154.

The node table 151 maps and manages the DNAI defined by 3GPP and the nodes 21 on Kubernetes. In other words, the mapping is performed so that the carrier network 13 can identify the nodes 21 managed by Kubernetes as nodes (DN: DataNet) to which pods 31 can be deployed. In addition, the node table 151 manages information such as site information, identification information as to whether the site is edge or global, latitude and longitude indicating the location, and IP address for each node 21.

The pod movement table 152 manages information on pods that are candidates for movement. The pod movement table 152 manages the movement priority of the candidate pods for movement based on the source node, which is node 21A of the edge site from which the movement originates, the destination node, which is node 21B of the global site to which the movement originates, and the priority. The pod movement table 152 also manages information such as the scheduled movement time, which is determined from the communication volume of the source node, for each candidate pod for movement.

UE movement table 153 manages information on UEs (candidate UEs) that are only accessing candidate pods to be moved. UE movement table 153 manages the movement priority of candidate UEs to be moved based on the source DNAI, which is the DNAI from which the movement will be made, the destination DNAI, which is the DNAI to which the movement will be made, and the priority. UE movement table 153 also manages information such as the scheduled movement time for each candidate UE to be moved. The life cycle of UE movement table 153 is the same as that of pod movement table 152. Hereinafter, when there is no need to distinguish between pod movement table 152 and UE movement table 153, they will also be referred to as movement tables.

The priority table 154 manages detailed information (qosClass) of the pod 31 and a priority according to the utilization rate of the computing resources of the pod 31 as information for determining the priority of candidate pods to be moved. For example, in the priority table 154, the more unrestricted a Besteffort type pod is, the higher its priority as a pod to be moved, and among those, the higher its priority is set for pods with utilization rates of computing resources equal to or above the median. This makes it possible for a pod to be moved with a higher priority as a candidate pod to be moved (a target for offloading).

In addition, since a priority is set for a candidate UE to be moved that corresponds to the candidate pod to be moved that it is accessing, it can be said that by determining the priority of the candidate pod to be moved, the priority of the candidate UE to be moved that is accessing the UE to be moved is also determined.

<Table configuration>
The configuration of a table managed by the database 114 in the MEC management system 11 will be described.

FIG. 4 is a diagram showing an example of the node table 151 in FIG. 3. In FIG. 4, the node table 151 uses the node ID that identifies the node as a key to manage information about the node's site, the DNAI mapped to the node, the latitude and longitude that indicate the node's location, and the node's IP address in association with each other.

For example, a node assigned node ID 1 exists at an edge site, the Sendai Edge Site, is mapped to DNAI1, has latitude and longitude AA,BB, and an IP address aa.aa.aa.aa. A node assigned node ID 5 exists at a global site, the Tokyo Global Site, is mapped to DNAI5, has latitude and longitude II,JJ, and an IP address ee.ee.ee.ee. Here, records with

node IDs

1 and 5 are shown as examples, but other records are managed in the same way.

FIG. 5 is a diagram showing an example of the pod movement table 152 in FIG. 3. In FIG. 5, the pod movement table 152 uses a pod ID that identifies a candidate pod to be moved as a key, and manages the source site (source node), destination site (destination node), scheduled movement time indicating the time of movement, name indicating the domain name, and priority indicating the priority of the movement in association with each other.

For example, a pod assigned pod ID 1 is scheduled to move from a node in the Sendai Edge Site to a node in the Tokyo Global Site at 1:00 on 1 November 2021, and has the highest priority for movement because it has priority 1. Similarly, a pod assigned pod ID 3 is scheduled to move from a node in the Fukuoka Edge Site to a node in the Osaka Global Site at 2:00 on 1 November 2021, and has the highest priority for movement because it has priority 1. Here, records with

pod IDs

1 and 3 are shown as examples, but other records are managed in the same way.

FIG. 6 is a diagram showing an example of the UE movement table 153 in FIG. 3. In FIG. 6, the UE movement table 153 uses the UEID that identifies the candidate UE to be moved as a key, and manages the source DNAI, destination DNAI, scheduled movement time indicating the time of movement, IP address, and priority indicating the priority of the movement in association with each other. Note that instead of the IP address, a Subscription Permanent Identifier (SUPI), which is identification information for identifying the UE in the carrier network 13, may be used.

For example, a UE assigned UEID 1 is scheduled to move from DNAI1 to DNAI5 at 1:00 on 1 November 2021, and has the highest priority for movement because its IP address is aa.aa.aa.1 and its priority is 1. Similarly, a UE assigned UEID 5 is scheduled to move from DNAI4 to DNAI6 at 2:00 on 1 November 2021, and has the highest priority for movement because its IP address is bb.bb.bb.1 and its priority is 1. Here, records for

UEID

1 and 5 are shown as examples, but other records are managed in the same way.

FIG. 7 is a diagram showing an example of the priority table 154 in FIG. 3. In FIG. 7, the priority table 154 is a table used to determine the priority of the movement of pods and UEs that are candidates for movement, and manages the priority according to the detailed information of the pod (qosClass) and the utilization rate of the pod's computing resources (pod resource utilization rate). In the example of FIG. 7, the lower the priority number, the higher the priority. The pod resource utilization rate includes the CPU utilization rate and memory utilization rate of the pod.

For example, in this disclosure, Burstable and BestEffort type pods are targeted for movement, and the more unlimited the Besteffort type pod is, the higher the priority for being targeted for movement. Among these, pods with pod resource utilization rates equal to or greater than the median are given a higher priority, so that BestEffort type pods with pod resource utilization rates equal to or greater than the median are given priority 1, making them the highest priority for movement. Also, BestEffort type pods with pod resource utilization rates less than the median are given priority 2, making them the second highest priority for movement after priority 1.

Furthermore, for burstable pods with pod resource utilization rates equal to or greater than the median, they are assigned a priority of 3, making them the second highest priority for migration after priority 2. For burstable pods with pod resource utilization rates below the median, they are assigned a priority of 4, making them the lowest priority for migration. Note that here, if multiple pods are running on the target node, the median pod resource utilization rates of those pods can be used as the standard.

In this way, the priority of the priority table 154 can be determined based on two axes: detailed information about the pod (qosClass) and the pod resource usage rate. Furthermore, it is not limited to these two axes, and more axes can be added to achieve a more suitable priority determination for the situation. Figures 8 to 10 show other examples of priority tables.

In the priority table 154 in FIG. 7, the priority is determined based on whether the pod resource utilization rate is high or low, based on the median value of the pod resource utilization rate, but cases in which the pod resource utilization rates are concentrated near the median value are also assumed. Therefore, when the variance value of the pod resource utilization rate is less than the threshold value, the priority of a pod with a low number of accesses from UEs, etc., based on the access log for a specified period of time is increased.

Figure 8 shows a priority table 154A that manages the priority according to the detailed information (qosClass) of a pod and the number of accesses to the pod. In the priority table 154A, of the Burstable and BestEffort type pods to be moved, BestEffort type pods with an access count below the median are assigned priority 1, making them the highest priority for movement. Also, BestEffort type pods with an access count equal to or greater than the median are assigned priority 2. Similarly, Burstable type pods with an access count below the median are assigned priority 3, and those with an access count equal to or greater than the median are assigned priority 4.

In this way, when the variance value of the pod resource utilization rate is less than the threshold value, i.e., when the pod resource utilization rate is fixed at the median value, the priority table 154A is used to increase the priority of pods with high resource utilization rates despite a low number of accesses from UEs, etc., and to move them preferentially, thereby reducing the impact of communication interruptions, etc. that occur during the move.

By the way, moving a pod to a global site node tends to increase delays and fluctuations compared to when it is operating on MEC12 (edge site node) due to the increased number of hops on the network in the carrier network 13 configured as a 5G network and the influence of traffic in other areas. Here, in order to prevent application programs used by PDU sessions of Resource Type GBR (Guaranteed Bit Rate) described in "Standardized 5QI to QoS characteristics mapping" in Table 5.7.4-1 of 3GPP TS 23.501 from being affected, it is better to leave the target application programs on MEC12 (edge site node) as much as possible, in order to minimize quality on the network.

Therefore, with regard to the QoS in 3GPP mentioned above, if the UE connected to the application program running on MEC12 (edge site node) knows in advance which QoS will be used to construct a PDU session, or if "AsSessionWithQoS" in 5.14 of 3GPP TS 29.122 is interconnected with the 3GPP API disclosed herein and a pre-configured QoSsubscription (qosReference) can be obtained (if the QoS listed in the Table in 5.7.4-1 above can be recognized via the API), then the priority of Non-GBR will be given a higher priority for movement.

Figure 9 shows priority table 154B, which manages detailed pod information (qosClass) and priority according to the 5G GBR/Non-GBR allocation in the communications used by the pod (UE PDU session). In priority table 154B, among the Burstable and BestEffort pods to be moved, BestEffort pods that are both Non-GBR are assigned priority 1, making them the highest priority for movement. Also, BestEffort pods that are GBR are assigned priority 2. Similarly, Burstable pods that are Non-GBR are assigned priority 3, and GBR pods are assigned priority 4.

In this way, the more GBR access and the pods that use it, the lower the priority, making it more likely that they will remain in MEC12 (the edge site node). This axis for determining priorities is intended to be placed after the axis of pod resource utilization and the axis of number of accesses mentioned above.

The above explanation assumes a pod that uses a CPU, but it may also be the case that the target is a pod that uses other processors such as a GPU (Graphics Processing Unit). Currently, GPU resources cannot be controlled as precisely as CPU resources, so they can be set to two values: limited or non-limited. Limited applies when there are resource restrictions on the pod, and non-limited applies when there are no resource restrictions on the pod.

FIG. 10 shows a priority table 154C that manages priorities according to the utilization rate of computing resources (pod resource utilization rate) of non-limited pods when targeting pods that use GPUs. The pod resource utilization rate here is the GPU utilization rate. In the priority table 154C, among non-limited pods, pods with GPU utilization rates equal to or greater than the median are assigned priority 1, making them the highest priority for migration. Additionally, pods with GPU utilization rates less than the median are assigned priority 2.

In this way, pod resource utilization is not limited to CPU utilization and memory utilization, and resource utilization such as GPU utilization may be used depending on the resources used by the pod. For example, when targeting pods that use GPUs, a similar priority can be determined for non-limited pods based on whether their GPU utilization is above or below the median value.

<Sequence>
Fig. 11 is a diagram showing an example of a sequence when a migration setting process is executed in the MEC management system 11. In Fig. 11, the exchange of data (signals) between devices is indicated by bidirectional arrows. When executing the migration setting process, the MEC management system 11 exchanges data between a node 21A at an edge site arranged in the MEC 12 and a node 21B at a global site arranged in the cloud data center 14, which is a migration source node.

In the MEC management system 11, the API unit 112 executes the kubectl top node command to obtain the computing resources (node resource usage) of node 21A (S11). Then, the MEC control unit 111 performs a threshold check to compare the obtained node resource usage with a threshold (S12). The node resource usage includes the CPU usage and memory usage of node 21A, etc.

If the MEC management system 11 determines in the threshold check (S12) that the node resource usage rate exceeds the threshold, the following process is carried out and the migration priority is determined.

In other words, the API unit 112 executes the kubectl describe pod command to obtain detailed information (qosClass) for each pod 31 running on node 21A (S13), and executes the kubectl top pod command to obtain the computing resources (pod resource utilization) of the pod 31 (S14). The pod resource utilization includes the CPU utilization and memory utilization of the pod 31. The MEC control unit 111 then uses the obtained detailed information (qosClass) and pod resource utilization to determine the migration priority (S15). This creates the priority table 154.

In the MEC management system 11, once the priority of the movement is determined, the pods and UEs that are candidates for movement are set in the pod movement table 152 and the UE movement table 153, respectively, along with their priorities.

In other words, the MEC control unit 111 refers to the node table 151, performs a predetermined distance calculation using latitude and longitude information, and determines the destination node 21B (node at the global site) (S16). The MEC control unit 111 also obtains the communication volume for a predetermined period from the communication log of the source node 21A (S17), and determines the movement time (scheduled movement time) based on the communication volume (S18). The MEC control unit 111 then sets and updates the pod movement table 152 with information about the candidate pod to be moved, such as the determined priority, destination node, and movement time (S19).

The MEC control unit 111 also extracts the IP address (UEIP) of the UE that is accessing only the candidate pod to be moved from the access log in the source node 21A, and sets the UE with the extracted UEIP as a candidate UE to be moved (S20). The MEC control unit 111 then sets and updates the UE movement table 153 with information about the candidate UE to be moved, such as the source DNAI, destination DNAI, movement time, and priority (S21).

In this way, when the mobility setting process is executed, the process is performed according to the sequence flow of FIG. 11, and then the mobility control process is executed based on the information set in the mobility setting process. FIG. 12 is a diagram showing an example of the sequence when the mobility control process is executed in the MEC management system 11. When the mobility control process is executed, the MEC management system 11 also exchanges data between the node 21B placed in the cloud data center 14, the name server 15, and the NEF server 22.

In the MEC management system 11, the MEC control unit 111 checks the time (S31), and when the target time arrives, the subsequent processing is performed. That is, the MEC control unit 111 identifies the record with the highest priority from the pod movement table 152 updated in step S19 of FIG. 11, and determines the candidate pod to be moved of the identified record as the pod to be moved (S32). The MEC control unit 111 also identifies the record with the highest priority from the UE movement table 153 updated in step S21 of FIG. 11, and determines the candidate UE to be moved of the identified record as the UE to be moved (S33).

The API unit 112, under control of the MEC control unit 111, deploys the pod to be moved that was running on the source node 21A to the destination node 21B (S34). The deployment here is performed by specifying a manifest file with the kubectl command. Furthermore, the AF unit 113 uses the TrafficInfluence API to switch the route of the UE to be moved (S36). When using the TrafficInfluence API, the AF unit 113, under control of the MEC control unit 111, specifies the UEIP and destination DNAI of the UE to be moved, and performs route control of the UE in cooperation with the NEF server 22. In this way, the movement of the pod to be moved and the UE to be moved according to priority is realized.

Then, in the MEC management system 11, the MEC control unit 111 updates the DNS record for the name of the pod to be moved in the pod movement table 152 (S37), and notifies the name server 15 of the IP address of the destination node 21B as the new IP address (S38). As a result, if the name of the pod to be moved is registered in the name server 15, it can update the IP address by replacing it with the IP address of the destination node 21B.

Furthermore, the API unit 112, under control of the MEC control unit 111, deletes the pod to be moved from the source node 21A (S39). The deletion here is performed by specifying the manifest file with the kubectl command. Finally, the MEC control unit 111 deletes the record of the pod to be moved from the pod movement table 152 (S40). In addition, the MEC control unit 111 deletes the record of the UE to be moved from the UE movement table 153 (S40).

<Processing flow>
Next, the flow of the migration setting process executed by the MEC control unit 111 of the MEC management system 11 will be described with reference to the flowchart in Fig. 13. The explanation in Fig. 13 illustrates an example in which the utilization rate of computing resources (node resource utilization rate) of the node 21A, which is located at the edge sites Sendai Edge Site and Fukuoka Edge Site among the nodes managed in the node table 151, exceeds a threshold value.

The movement setting unit 131 controls the API unit 112 to acquire the node resource utilization rate of the node 21A placed at the edge site (S111), and determines whether the acquired node resource utilization rate exceeds a threshold value (S112). Here, the API unit 112 executes the kubectl top node command to acquire the CPU utilization rate and memory utilization rate of the node 21A as the node resource utilization rate, which are compared with the threshold value. The node resource utilization rate is not limited to the utilization rate at a specific time (time period), and for example, the load average may be used.

If it is determined that the node resource usage exceeds the threshold (Yes in S112), the migration setting unit 131 checks whether or not there is a migration table in the database 114 (S113). For example, if the node resource usage exceeds the threshold for the first time, neither the pod migration table 152 nor the UE migration table 153 exists (No in S113), and the migration setting unit 131 controls the API unit 112 to acquire detailed information about each pod running on the node 21A (S114). In this example, the API unit 112 executes the kubectl describe pod command to acquire detailed information (qosClass) about each pod running on the node 21A of the Sendai Edge Site and the node 21A of the Fukuoka Edge Site. The migration setting unit 131 then checks the type of qosClass acquired, and performs the processes of steps S116 to S122 for pods determined to be of the BestEffort type or Burstable type (Yes in S115).

In other words, the migration setting unit 131 controls the API unit 112 to obtain the utilization rate of the pod's computing resources (pod resource utilization rate) (S116). Here, the API unit 112 executes the kubectl top pod command to obtain the pod's CPU utilization rate and memory utilization rate as the pod resource utilization rate. The pod resource utilization rate is not limited to the utilization rate at a specific time (time period), and for example, the load average may be used.

The priority determination unit 132 determines the priority of the pod movement based on the qosClass and the pod resource usage rate (S117). This creates a priority table 154. For example, in the priority table 154 of FIG. 7, the priority of unrestricted BestEffort type movement is higher, and among them, pods with pod resource usage rates equal to or higher than the median are assigned the highest movement priority of 1. Note that here, a priority table 154A (FIG. 8) using the number of accesses or a priority table 154B (FIG. 9) according to the GBR/Non-GBR allocation may also be created.

Next, the movement setting unit 131 refers to the node table 151 and determines, from among the nodes of global sites stored in the node table 151, the node with the closest distance from the source node calculated using latitude and longitude as the destination node (S118). For example, in the node table 151 of FIG. 4, the node of the Tokyo Global Site, which is the global site with the closest distance to the source node of the Sendai Edge Site, is determined as the destination. Also, the node of the Osaka Global Site, which is the global site with the closest distance to the source node of the Fukuoka Edge Site, is determined as the destination.

Next, the movement setting unit 131 determines the time period with the least amount of communication on the source node (S119). For example, by referring to the communication log for a specified period (e.g., one week) on the source node, the communication volume for that period is averaged for each day to calculate the time period with the least amount of communication. In this example, 1:00 is calculated as the time period with the least amount of communication on the node at the Sendai Edge Site, and 2:00 is calculated as the time period with the least amount of communication on the node at the Fukuoka Edge Site. Note that the time period to be determined is basically assumed to be at night, but if the global site extends overseas, for example, it may be daytime due to the time difference. The date and time corresponding to the time period determined here becomes the time of movement of the pod (scheduled time of movement).

The movement setting unit 131 sets and updates the information, such as the priority, destination node, and movement time determined in steps S117 to S119, in the pod movement table 152 as information about the pod candidate to be moved (S120). For example, in the pod movement table 152 of FIG. 5, information about the pod to be moved from the Sendai Edge Site to the Tokyo Global Site at the scheduled movement time of 2021/11/1 1:00 is set as records for

pod IDs

1 and 2, and information about the pod to be moved from the Fukuoka Edge Site to the Osaka Global Site at the scheduled movement time of 2021/11/1 2:00 is set as records for

pod IDs

3, 4, 5, 6, and 7. In addition, a priority is set for each pod, and for example, the pods candidate to be moved with

pod IDs

1, 3, and 4 have a priority of 1, and therefore become the pods to be moved first at the corresponding scheduled movement times. The destination node, which is a node in the global site, is mapped to DNAI in node table 151.

Next, the mobility setting unit 131 refers to the access log for a predetermined period (e.g., one week) and extracts the IP addresses (UEIP) of UEs that are accessing only the candidate pods to be moved for each priority (S121). The mobility setting unit 131 sets and updates the UE mobility table 153 with information such as the source DNAI, destination DNAI, mobility time, and priority as the candidate UEs to be moved (S122).

The movement time and priority here correspond to the movement time and priority of the candidate pod to be moved that the candidate UE to be moved is accessing. In addition, the source DNAI and destination DNAI can be set from the DNAI mapped to the node in the node table 151. For example, in the UE movement table 153 of FIG. 6, information on UEs (UEs that only access pods at the Sendai Edge Site) whose routes will be switched from DNAI1 to DNAI5 at the scheduled movement time of 1:00 on November 1, 2021 is set as records for

UEIDs

1, 2, 3, and 4, and information on UEs (UEs that only access pods at the Fukuoka Edge Site) whose routes will be switched from DNAI4 to DNAI6 at the scheduled movement time of 2:00 on November 1, 2021 is set as records for

UEIDs

5, 6, 7, 8, and 9. Additionally, a priority is set for each UE according to the priority of the pod it accesses. For example, candidate UEs for movement with

UEIDs

1, 2, 3, 5, and 6 will have a priority of 1, and will be the first UEs to be switched to at the scheduled movement time.

When step S122 ends, the series of processes ends. If it is determined that the node resource utilization rate is below the threshold (No in S112), and there is a migration table (Yes in S123), the migration table is reset (S124) and the process ends. On the other hand, if there is no migration table (No in S123), the process ends as it is. Also, even if it is determined that the node resource utilization rate exceeds the threshold (Yes in S112), if there is a migration table (Yes in S113), the process ends. Furthermore, if it is determined that the type of qosClass of each node is not BestEffort or Burstable (No in S115), that is, if it is a Guaranteed type pod, it is not a pod to be moved as disclosed herein, so steps S116 to S122 are skipped and the process ends.

The above-mentioned movement setting process updates the pod movement table 152 and the UE movement table 153, and sets the information of the movement target candidate pods and the movement target candidate UEs together with their priorities. Then, based on the information set in the movement setting process, a movement control process is performed to move the movement target pods and the movement target UEs according to the priorities. Next, the flow of the movement control process executed by the MEC control unit 111 of the MEC management system 11 will be described with reference to the flowchart in FIG. 14.

The control unit 123 refers to the pod movement table 152 (S131) and determines whether the current time has reached the target time indicated by the scheduled movement time (S132). If the control unit 123 determines that the target time has reached (Yes in S132), it performs the processes of steps S133 to S139.

In other words, the control unit 123 refers to the pod movement table 152 and sets the candidate pod to be moved of the record with the highest priority as the pod to be moved (S133). The control unit 123 also refers to the UE movement table 153 and sets the candidate UE to be moved of the record with the highest priority as the UE to be moved (S134). In this example, assuming that it is 1:00 on November 1, 2021, the pod with pod ID 1, which has a scheduled movement time of 1:00 on November 1, 2021 and is set to priority 1 in the pod movement table 152 of FIG. 5, becomes the pod to be moved. Also, in the UE movement table 153 of FIG. 6, the UEs with

UE IDs

1, 2, and 3, which have a scheduled movement time of 1:00 on November 1, 2021 and is set to priority 1, become the UEs to be moved.

First, the movement control unit 141 controls the API unit 112 to deploy the pod to be moved to the destination node (S135). In this example, the destination of the pod to be moved with pod ID 1 is a node in the Tokyo Global Site. There are several deployment methods, but a typical one is to specify a manifest file with the kubectl command. The node in the Tokyo Global Site, which is the destination node, is mapped to DNAI, which is DNAI5, in the node table 151 in Figure 4.

Next, the route control unit 142 controls the AF unit 113 to perform route control using the TrafficInfluence API and switch the route of the UE to be moved (S136). When using the TrafficInfluence API, route control is performed in cooperation with the NEF server 22 by specifying the UEIP and destination DNAI of the UE to be moved. For example, in the UE movement table 153 of Figure 6, the routes of the UEs with

UEIDs

1, 2, and 3 are switched by specifying the IP addresses aa.aa.aa.1, aa.aa.aa.2, and aa.aa.aa.3 corresponding to

UEIDs

1, 2, and 3 corresponding to priority 1, and the destination DNAI of DNAI 5.

In addition, although this is not a required process, the control unit 123 updates the IP address of the name of the pod to be moved, which is managed by the name server 15, to the IP address of the destination node (S137). For example, if aaa.co.jp, which is the name of the pod to be moved with pod ID 1 in the pod movement table 152 of Figure 5, is registered in the name server 15, the control unit 123 can replace the IP address of the name of the pod to be moved with the IP address of the destination node by notifying the name server 15 of ee.ee.ee.ee, which is the IP address of the destination node, which is the Tokyo Global Site.

Next, the movement control unit 141 controls the API unit 112 to delete the pod to be moved from the source node (S138). In this example, the pod to be moved with pod ID 1 is deleted from the node of the Sendai Edge Site. There are several ways to delete the pod, but it can be done by specifying a manifest file with the kubectl command.

Finally, the control unit 123 deletes the records to be moved from the pod movement table 152 and the UE movement table 153 (S139). For example, in the pod movement table 152 of FIG. 5, the record of pod ID 1, which is the pod to be moved, is deleted. Also, in the UE movement table 153 of FIG. 6, the records of

UE IDs

1, 2, and 3, which are the UEs to be moved, are deleted. When step S139 ends, the series of processes ends. Note that if it is determined that the target time has not yet arrived (No in S132), steps S133 to S139 are skipped and the process ends.

In the above explanation, a move from the Sendai Edge Site to the Tokyo Global Site was given as an example of a priority 1 move, but then, at 2:00 on November 1st, 2021, a move from the Fukuoka Edge Site to the Osaka Global Site will occur, and the same process will be performed for

pod IDs

3 and 4 and UEIDs 5 and 6.

The following processing is performed after the movement of priority 1. That is, the movement control processing of priority 1 is performed, and the priority 1 records are deleted from both the pod movement table 152 and the UE movement table 153. Then, the movement setting processing of FIG. 13 is executed again, and the node resource utilization threshold is checked (S111, S112). In this example, if it is determined that the node resource utilization rates of the Sendai Edge Site and the Fukuoka Edge Site are below the threshold (No in S112), the pod movement table 152 and the UE movement table 153 become unnecessary, and these movement tables are reset (S124). That is, since the records of priorities 2 to 4 are also deleted, the movement control processing for the remaining candidate pods and candidate UEs to be moved is not performed and ends.

On the other hand, if it is determined that the node resource utilization rate exceeds the threshold (Yes in S112), records with priorities 2 to 4 remain in the pod migration table 152 and the UE migration table 153, so for example, at the scheduled migration time on the next day or later (for example, 1:00 on 2021/11/2, 2:00 on 2021/11/2), it is determined that the target time has arrived in the migration control process of FIG. 14 (Yes in S132), and the migration candidate pod and migration candidate UE with priority 2 are set as the migration target pod and migration target UE (S133, S134), and migration control is performed (S135, S136). Then, the same process is performed on the records with priority 3 and priority 4 until it is determined that the node resource utilization rate is less than the threshold (No in S112). In other words, in this example, the processing will be completed in as little as one day (for example, 2021/11/1) and at most four days (for example, 2021/11/1 - 11/4).

As described above, in the present disclosure, the priority of pod movement can be determined from detailed information about each pod operating on an edge site node where node resource utilization has exceeded a threshold and the pod resource utilization, and the pod can be moved from the edge site node to a global site node according to the determined priority. In addition, in the present disclosure, the route of a UE that is only accessing the pod to be moved can be switched according to the determined priority. As a result, at an edge site node, a pod (application container) with a specific load is preferentially moved to a global site node, and the route of a UE that is only accessing that pod is switched, so that the pod (application container) or UE to be offloaded can be appropriately determined, and more appropriate targets can be offloaded.

In addition, in this disclosure, pods whose computing resources can be controlled tend to gather at MEC12 (edge site node) while minimizing the degradation of the quality of experience that accompanies route changes. This makes it possible to prevent a particular pod running on an edge site node from occupying too many of the node's computing resources, causing the node to shut down. In other words, it makes it possible to prevent a situation in which all application programs (pods, etc.) running on an edge site node shut down completely.

<Modification>
In the present disclosure, a system refers to a logical collection of multiple devices, but the MEC management system 11 may be composed of multiple devices (e.g., servers, etc.). For example, in the MEC management system 11 of FIG. 2, a device having the MEC control unit 111, a device having the API unit 112, a device having the AF unit 113, and a device having the database 114 may be composed of separate devices. Alternatively, the MEC management system 11 may be regarded as an MEC management device. This MEC management device is a device having the MEC control unit 111, the API unit 112, the AF unit 113, and the database 114. Note that the MEC management device only needs to have at least the MEC control unit 111, and the API unit 112, the AF unit 113, and the database 114 may be included in external devices.

In the above explanation, four priority levels from 1 to 4 are used in the priority table 154, but the priority level is not limited to four levels and may be more than four, as long as it is two or more. For example, in this disclosure, Burstable and BestEffort pods are targeted for movement, but other classifications (Guaranteed, etc.) may be targeted for movement, or the pod resource utilization may be further divided into smaller categories using a threshold or the like rather than being above or below the median, thereby increasing the number of priority levels. Also, the median is used for the pod resource utilization, but other indices may be used instead of being limited to the median.

Note that resource utilization rates such as node resource utilization rate and pod resource utilization rate are examples of resource values, and other indices may be used. Furthermore, when multiple resource utilization rates such as CPU utilization rate and memory utilization rate are obtained as node resource utilization rate and pod resource utilization rate, it is sufficient to use at least one of the resource utilization rates, as well as all of the resource utilization rates. When comparing resource utilization rate with a threshold value, median value, etc., the obtained resource utilization rate may be used as is, or a predetermined calculation may be performed before comparing with the threshold value, median value, etc.

In the above explanation, latitude and longitude are used as location information regarding the positions of edge site nodes and global site nodes in node table 151, but latitude and longitude are only one example of location information, and other location information may be used as long as it is information that can identify the node position.

The terms used in the above explanation may differ depending on the standard, and may be replaced with the terms of the relevant standard. For example, a node (node 21A in FIG. 1) consisting of a server etc. in an MEC (MEC 12 in FIG. 1) is also called a worker node in Kubernetes and a data net (DN: DataNet) in 3GPP, but is called a node in this disclosure.

<Computer Configuration>
The above-mentioned series of processes (e.g., the movement setting process, the movement control process, and the like) can be executed by hardware or by software. When the series of processes is executed by software, a program constituting the software is installed in a computer. FIG. 15 is a block diagram showing an example of the hardware configuration of a computer that executes the above-mentioned series of processes by a program.

In a computer, a CPU 1001, a ROM (Read Only Memory) 1002, and a RAM (Random Access Memory) 1003 are interconnected by a bus 1004. An input/output interface 1005 is further connected to the bus 1004. An input unit 1006, an output unit 1007, a memory unit 1008, a communication unit 1009, and a drive 1010 are connected to the input/output interface 1005.

The input unit 1006 includes a keyboard, a mouse, a microphone, etc. The output unit 1007 includes a display, a speaker, etc. The storage unit 1008 includes a hard disk, a non-volatile memory, etc. The communication unit 1009 includes a network interface, etc. The drive 1010 drives a removable recording medium 1011 such as a semiconductor memory, a magnetic disk, an optical disk, or a magneto-optical disk.

In a computer configured as described above, the CPU 1001 loads the programs stored in the ROM 1002 or memory unit 1008 into the RAM 1003 via the input/output interface 1005 and bus 1004, and executes them, thereby carrying out the above-mentioned series of processes.

The program executed by the computer (CPU 1001) can be provided by being recorded on a removable recording medium 1011, such as a package medium, for example. The program can also be provided via a wired or wireless transmission medium, such as a local area network, the Internet, or digital satellite broadcasting.

In a computer, a program can be installed in the storage unit 1008 via the input/output interface 1005 by inserting the removable recording medium 1011 into the drive 1010. The program can also be received by the communication unit 1009 via a wired or wireless transmission medium and installed in the storage unit 1008. Alternatively, the program can be pre-installed in the ROM 1002 or storage unit 1008.

In this specification, the processing performed by a computer according to a program does not necessarily have to be performed chronologically in the order described in the flowchart. In other words, the processing performed by a computer according to a program also includes processing executed in parallel or individually (for example, parallel processing or processing by objects). In addition, a program may be processed by one computer (processor), or may be processed in a distributed manner by multiple computers.

Note that the embodiments of the present disclosure are not limited to the above-described embodiments, and various modifications are possible without departing from the spirit of the present disclosure. Furthermore, the effects described in this specification are merely examples and are not limiting, and other effects may also be present.

Furthermore, this disclosure can be configured as follows:

(1)
a management unit that manages a first node, which is a computer at an edge site connected to a mobile communication network, and a second node, which is a computer at a global site connected to the Internet;
a determination unit that, when a first resource value indicating a computing resource of the first node exceeds a threshold, determines a priority of migration of the application container based on detailed information on details of an application container consisting of a plurality of containers operating on the source node and a second resource value indicating a computing resource of the application container, with the first node whose first resource value exceeds the threshold as a source node;
a control unit that performs control to move the application container from the source node to a destination node, which is the second node that is a destination of the application container, according to the priority.
(2)
The first node and the second node are pre-mapped with DNAI defined in 3GPP,
The information processing device according to (1), wherein the control unit moves the application container to the second node mapped to the DNAI as the destination node.
(3)
the determination unit determines, based on a log related to communication in the source node, a time corresponding to a time period during which communication volume in the source node is the lowest, as a movement time for moving the application container;
The information processing device according to (2), wherein the control unit moves the application container in accordance with the priority when the current time becomes the movement time.
(4)
The information processing device described in (3), wherein the determination unit determines the second node located closest to the source node as the destination node based on location information regarding the locations of the first node and the second node.
(5)
The determination unit determines, as a movement target, a mobile communication terminal that is accessing only the application container to be moved, among the mobile communication terminals that communicate via the mobile communication network, based on a log related to access to the application container;
The information processing device according to (3) or (4), wherein the control unit performs control for switching a route of the mobile communication terminal to be moved according to the priority when the current time becomes the moving time.
(6)
The information processing device described in (1), wherein the detailed information includes information regarding the classification of the application container determined from the minimum resource capacity required by the application container and the maximum resource capacity that can be used when the load on the application container increases.
(7)
The information processing device according to (6), wherein the second resource value is a resource usage rate including a CPU usage rate and a memory usage rate by the application container.
(8)
The detailed information includes a qosClass defined in kubernetes,
The determination unit is
The application container that is BestEffort or Burstable is targeted for migration,
Giving a higher priority to the movement of the BestEffort application container than the Burstable application container;
The information processing device according to (7), wherein a higher priority is given to the movement of the application container whose resource usage rate is equal to or greater than the median than to the application container whose resource usage rate is less than the median.
(9)
The information processing device according to (8), wherein the decision unit assigns a higher priority to the movement of the application container whose access count is less than the median than to the application container whose access count is equal to or greater than the median.
(10)
The information processing device according to (9), wherein the decision unit gives a higher priority to the movement of the application container of Non-GBR than the application container of GBR defined in 3GPP.
(11)
the second resource value is a resource usage rate including a GPU usage rate by the application container;
The information processing device according to (1), wherein the decision unit assigns a higher priority to the movement of application containers with resource usage rates equal to or greater than a median among the application containers with no resource restrictions than to application containers with resource usage rates less than the median.
(12)
The information processing device according to (1), wherein the application container corresponds to a pod defined in Kubernetes.
(13)
An information processing device,
Manage a first node, which is a computer at an edge site connected to a mobile communication network, and a second node, which is a computer at a global site connected to the Internet;
When a first resource value indicating a computing resource of the first node exceeds a threshold, the first node whose first resource value exceeds the threshold is set as a source node, and a priority for migration of the application container is determined based on detailed information on details of an application container consisting of a plurality of containers operating on the source node and a second resource value indicating a computing resource of the application container;
an information processing method for controlling the movement of the application container from the source node to a destination node, which is the second node serving as a destination, according to the priority.

11 MEC management system, 12 MEC, 13 Carrier network, 14 Cloud data center, 15 Name server, 16 Wireless terminal, 21, 21A, 21B Node, 22 NEF server, 111 MEC control unit, 112 API unit, 113 AF unit, 114 Database, 121 Management unit, 122 Setting unit, 123 Control unit, 131 Mobility setting unit, 132 Priority determination unit, 141 Mobility control unit, 142 Route control unit, 151 Node table, 152 Pod mobility table, 153 UE mobility table, 154, 154A, 154B, 154C Priority table, 1001 CPU

Claims

a management unit that manages a first node, which is a computer at an edge site connected to a mobile communication network, and a second node, which is a computer at a global site connected to the Internet;
a determination unit that, when a first resource value indicating a computing resource of the first node exceeds a threshold, determines a priority of migration of the application container based on detailed information on details of an application container consisting of a plurality of containers operating on the source node and a second resource value indicating a computing resource of the application container, with the first node whose first resource value exceeds the threshold as a source node;
a control unit that performs control to move the application container from the source node to a destination node, which is the second node that is a destination of the application container, according to the priority.
The first node and the second node are pre-mapped with DNAI defined in 3GPP,
The information processing device according to claim 1 , wherein the control unit moves the application container to the second node mapped to the DNAI as the destination node.
the determination unit determines, based on a log related to communication in the source node, a time corresponding to a time period during which communication volume in the source node is the lowest, as a movement time for moving the application container;
The information processing device according to claim 2 , wherein the control unit moves the application container in accordance with the priority when the current time becomes the movement time.
The information processing device according to claim 3 , wherein the determination unit determines the second node located closest to the source node as the destination node based on position information regarding the positions of the first node and the second node.
The determination unit determines, as a movement target, a mobile communication terminal that is accessing only the application container to be moved among the mobile communication terminals that communicate via the mobile communication network based on a log related to access to the application container;
The information processing device according to claim 3 , wherein the control unit performs control for switching a route of the mobile communication terminal to be moved in accordance with the priority when the current time becomes the moving time.
The information processing device according to claim 1 , wherein the detailed information includes information regarding a classification of the application container that is identified from a minimum resource capacity required by the application container and a maximum resource capacity that can be used when the load of the application container increases.
The information processing apparatus according to claim 6 , wherein the second resource value is a resource usage rate including a CPU usage rate and a memory usage rate by the application container.
The detailed information includes a qosClass defined in kubernetes,
The determination unit is
The application container that is BestEffort or Burstable is targeted for migration,
Giving a higher priority to the movement of the BestEffort application container than the Burstable application container;
The information processing apparatus according to claim 7 , wherein a higher priority is given to the movement of the application container whose resource usage rate is equal to or greater than the median than to the application container whose resource usage rate is less than the median.
The information processing device according to claim 8 , wherein the decision unit assigns a higher priority to the movement of the application container whose access count is less than a median value than to the application container whose access count is equal to or greater than the median value.
The information processing device according to claim 9 , wherein the decision unit assigns a higher priority to the movement of the non-GBR application container than to the GBR application container defined in 3GPP.
the second resource value is a resource usage rate including a GPU usage rate by the application container;
The information processing device according to claim 1 , wherein the decision unit assigns a higher priority to the movement of application containers with resource usage rates equal to or greater than a median among the application containers with no resource restrictions than to application containers with resource usage rates less than the median.
The information processing device according to claim 1 , wherein the application container corresponds to a pod defined in Kubernetes.
An information processing device,
Manage a first node, which is a computer at an edge site connected to a mobile communication network, and a second node, which is a computer at a global site connected to the Internet;
When a first resource value indicating a computing resource of the first node exceeds a threshold, the first node whose first resource value exceeds the threshold is set as a source node, and a priority for migration of the application container is determined based on detailed information on details of an application container consisting of multiple containers operating on the source node and a second resource value indicating a computing resource of the application container;
an information processing method for controlling the movement of the application container from the source node to a destination node, which is the second node serving as a destination, according to the priority.