Scheduling and resource management system for hard disk-less nodes

Abstract

The present application relates to the field of Internet technology and discloses a scheduling and resource management system for hard disk-free nodes, which is more efficient, reliable and widely applicable. In this system, a cloud scheduling manager is connected to one or more node controllers, and each node controller is connected to one or more cluster nodes; the cloud scheduling manager is used to receive instance applications or release requests, send initialization requests to the node controllers, perform node configuration information and status updates based on feedback, perform node scheduling, send node applications or release requests to the node controllers, and dynamically configure and deploy cluster nodes through the container orchestration system; the node controller checks and reports node status to the cluster nodes based on the initialization request, allocates disk resources through the disk manager, controls nodes based on node applications or release requests, and sends power on / off messages to the cluster nodes; the cluster nodes are deployed with a container orchestration system for automated configuration.

Description

Technical Field

[0001] The present invention relates to the field of Internet technology, and more particularly to a scheduling and resource management system without hard disk nodes. Background Art

[0002] This section is intended to provide a background or context to the embodiments of the present application as recited in the claims. No admission is made that the description herein is prior art as disclosed by virtue of its inclusion in this section.

[0003] The widespread adoption of containerization technology in modern data centers and cloud infrastructure has transformed how applications are deployed and managed. Kubernetes provides powerful tools and extensive ecosystem support for managing the container lifecycle. With the pursuit of cost optimization and operational flexibility, more and more data centers are adopting diskless nodes. These nodes load the operating system and application data from a central server over the network upon startup, without relying on local persistent storage. This architecture reduces hardware investment and simplifies system upgrades and maintenance.

[0004] However, Kubernetes' native scheduling and management features are primarily designed for stateful nodes, and lack sufficient support for the unique scheduling requirements of diskless nodes. These nodes require efficient network startup and frequent data exchange, posing new challenges to the scheduling system's responsiveness and resource allocation strategies.

[0005] Currently, the closest implementation to the present invention uses standard Kubernetes with customized scheduling policies to manage diskless nodes. For example, some systems may extend Kubernetes through Custom Resource Definitions (CRDs) and Operator patterns to accommodate specific hardware configurations and scheduling requirements. These solutions often require significant modifications to Kubernetes, increasing system complexity and maintenance costs.

[0006] However, these approaches often lack scheduling algorithms optimized for diskless environments, suffer from inefficient resource scheduling, fail to automatically handle changes in node hardware status, and fail to effectively leverage existing data analysis and prediction techniques to optimize resource allocation. Furthermore, these solutions often lack advanced fault recovery mechanisms, failing to ensure high application availability and data consistency in the event of network or hardware failures. The systems also lack scalability, making it difficult to effectively dynamically manage the power supply of diskless nodes. Summary of the Invention

[0007] The purpose of the present invention is to provide a scheduling and resource management system for hard disk-free nodes, providing a more efficient and reliable solution for hard disk-free environments.

[0008] The present invention provides a scheduling and resource management system for hard disk-free nodes, including a cloud scheduling manager, a node controller, and cluster nodes:

[0009] The cloud scheduling manager is connected to one or more node controllers, and each node controller is connected to one or more cluster nodes;

[0010] The cloud scheduling manager includes a resource scheduler and a container orchestration system; the cloud scheduling manager is used to receive instance application or instance release requests, initialize network configuration, send a request to initialize the cluster node to the node controller through the resource scheduler, update node configuration information and status based on feedback from the node controller, perform node scheduling, send node application or release requests to the node controller, and dynamically configure and deploy cluster nodes in cooperation with the resource scheduler through the container orchestration system;

[0011] The node controller includes a message distributor and a disk manager; the node controller is used to initialize local resources and process messages between the cloud scheduling manager and the cluster nodes through the message distributor; the node controller checks and reports node status to the cluster nodes based on initialization requests sent by the cloud scheduling manager, allocates disk resources through the disk manager, controls nodes based on node application or release requests sent by the cloud scheduling manager, and sends power on / off messages to the cluster nodes;

[0012] The cluster nodes are deployed with a container orchestration system for automated configuration.

[0013] In a preferred embodiment, the container orchestration system used by the cloud scheduling manager includes Kubernetes and Kubeedge, and the container orchestration system used by the cluster nodes includes Kubeedge; or

[0014] The scheduling and resource management system uses Docker Swarm or Apache Mesos for container orchestration.

[0015] In a preferred embodiment, the system includes a fault detection and recovery mechanism;

[0016] The cloud scheduling manager further includes a fault detection and recovery unit, and the node controller further includes a fault collector. The fault detection and recovery unit is used to monitor and identify node faults in real time and perform fault processing. The fault collector is used to check and report abnormal states of the cluster nodes, including the faults.

[0017] The node controller further includes a power supply controller, which is configured to provide power support to the node controller under the constraint of power saving.

[0018] In a preferred example, the resource scheduler predicts future node usage requirements based on the existing instance application or instance release request, combined with the workload requirements of the node controller and the power saving control requirements, and dynamically adjusts the allocation scheduling of the disk resources and the cluster nodes.

[0019] This application also discloses a scheduling and resource management method for hard disk-free nodes, including:

[0020] When the cloud scheduling manager receives an instance request, it matches the optimal node based on the node information in the database and the scheduling mechanism of the optimal node, and then sends the instance request message to the node controller corresponding to the optimal node;

[0021] The node controller allocates disk resources to the optimal node and wakes up the optimal node according to the request of the instance application;

[0022] The node controller notifies the cloud scheduling manager of information about the allocated disk resources;

[0023] The cloud scheduling manager updates the configuration storage and prepares the scheduling configuration of the K8S of the optimal node; and

[0024] After the optimal node is powered on, the automated configuration software is started and added to the K8S cluster.

[0025] In a preferred embodiment, it also includes:

[0026] When the cloud scheduling manager receives a request to release an instance of a designated node, it notifies the node controller to reclaim the designated node and updates the node status according to the feedback message;

[0027] The node controller sends a shutdown command to the designated node and monitors the status of the designated node;

[0028] After the node controller monitors that the optimal node is shut down, it releases the disk resources used by the designated node.

[0029] In a preferred embodiment, it also includes:

[0030] After the node controller detects that an abnormality occurs in the first node, it reports the abnormality to the cloud scheduler;

[0031] The cloud scheduler starts the standby node according to the standby node scheduling mechanism, determines a second node matching the first node from the current standby nodes through the database, and updates the resource configuration of the first node to the second node;

[0032] The node controller initiates scheduling of the second node, initiates a startup application for the second node, and updates the status of the first node and the second node.

[0033] In a preferred embodiment, the optimal node scheduling mechanism specifically includes:

[0034] Step 801: Check the resources of all current node controllers and find the node controller where the optimal node is located according to the algorithm;

[0035] Step 802: Check the node status of the corresponding node controller. If there is a matching node, determine the matching node. If not, return to step 801.

[0036] Step 803: Check whether the node is in a ready state. If so, check the node usage state. If in an idle state, directly initiate a node application. If not ready or in a non-idle state, proceed to step 802.

[0037] Step 804: If the ready node is scheduled, a standby node application is initiated and the new node is placed in the database as a standby node.

[0038] Step 805: The standby node reaches the threshold value and no new nodes are added.

[0039] In a preferred embodiment, the node status update specifically includes:

[0040] Keep node resources in the ready state and update the node usage status to the idle state.

[0041] In a preferred embodiment, the standby node calling mechanism specifically includes:

[0042] Extract the resource configuration of the existing instance and search for a backup node that matches the first node for scheduling, where:

[0043] The first node is searched first in the backup nodes under the same node controller as the first node. If the backup node is not found, the first node is searched for a backup node across the node controllers.

[0044] Compared with the prior art, the main differences and effects of the embodiments of the present invention are:

[0045] It provides a complete and effective scheduling and management solution in an environment without hard disk nodes, and efficiently and reliably schedules and allocates disk resources according to task arrangements.

[0046] Furthermore, the present application can dynamically adjust resource allocation based on real-time scheduling status monitoring and prediction models of hard disk-free nodes, significantly improving resource utilization and reducing operating costs.

[0047] Furthermore, the present application sets up advanced fault detection and recovery mechanisms, which ensure that the system can recover quickly when facing hardware or network failures, maintaining service continuity and data integrity.

[0048] Furthermore, the present application provides a simplified deployment strategy, including automated management functions to reduce the burden on system administrators and reduce the complexity of system maintenance;

[0049] Furthermore, the present application is scalable and flexible, and can dynamically adjust the cluster size according to the controller load status, thereby achieving seamless expansion and scaling and reducing energy consumption.

[0050] It should be understood that within the scope of the present invention, the above-mentioned technical features of the present invention and the technical features described in detail below (such as in the embodiments) can be combined with each other to form new or preferred technical solutions. Due to space limitations, they will not be listed here one by one. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0052] Figure 1 It is a system block diagram according to one embodiment of the present application.

[0053] Figure 2 It is a schematic diagram of the scheduling process in one embodiment of the present application.

[0054] Figure 3 This is a schematic diagram of the scheduling timing of initialization and instance application in one embodiment of the present application.

[0055] Figure 4 This is a schematic diagram of the scheduling timing of the release according to an example in one embodiment of the present application.

[0056] Figure 5 This is a schematic diagram of the scheduling timing of exception handling in one embodiment of the present application.

[0057] Figure 6 This is a flow chart of the optimal node scheduling mechanism in one embodiment of the present application.

[0058] In the accompanying drawings, the following are marked:

[0059] 1-Cloud scheduling manager;

[0060] 101-Resource Scheduler;

[0061] 102-Fault detection and recovery device;

[0062] 2-node controller;

[0063] 201-message distributor;

[0064] 202-Disk Manager;

[0065] 203-power controller;

[0066] 204-Fault Collector;

[0067] 3-Cluster nodes;

[0068] 4-Container orchestration system;

[0069] 5-Database. DETAILED DESCRIPTION

[0070] In the following description, many technical details are provided to help readers better understand this application. However, those skilled in the art will understand that even without these technical details and various changes and modifications based on the following embodiments, the technical solutions claimed in this application can be implemented.

[0071] Description of some concepts

[0072] A cluster is a group of networked computers that work together as a single system. Clusters are composed of multiple nodes and are typically used to increase computing power, provide higher availability, and achieve load balancing.

[0073] A node is a single computer or device within a cluster, serving as a component of the cluster. Each node carries a portion of the cluster's functionality, potentially processing data, storing information, or performing specific services. A node can be either a physical machine or a virtual machine.

[0074] A Kubernetes (k8s) cluster is a group of collaborative computers, often called nodes, that work together to run containerized applications. A Kubernetes cluster provides a distributed environment for automating the deployment, scaling, and management of containerized applications, ensuring they run efficiently and reliably.

[0075] KubeEdge: Extends native containerized application orchestration capabilities to the enterprise edge. Built on Kubernetes, it enables Kubernetes functionality to be applied in edge computing environments.

[0076] In order to make the objectives, technical solutions and advantages of this application clearer, the implementation methods of this application will be further described in detail below with reference to the accompanying drawings.

[0077] One embodiment of the present application relates to a scheduling and resource management system for hard disk-free nodes, the structure of which is as follows: Figure 1 As shown, the system includes:

[0078] The cloud scheduling manager 1 is connected to one or more node controllers 2, and each node controller 2 is connected to one or more cluster nodes 3;

[0079] The cloud scheduling manager 1 includes a resource scheduler 101 and a container orchestration system 4; the cloud scheduling manager 1 is used to receive instance application or instance release requests, initialize network configuration, send a request to initialize cluster nodes 3 to the node controller 2 through the resource scheduler 101, update node configuration information and status based on feedback from the node controller 2, perform node scheduling, send node application or release requests to the node controller 2, and dynamically configure and deploy cluster nodes 3 in cooperation with the resource scheduler 101 through the container orchestration system 4;

[0080] Node controller 2 includes a message distributor 201 and a disk manager 202. Node controller 2 is used to initialize local resources and process messages between cloud scheduling manager 1 and cluster nodes 3 through message distributor 201. Node controller 2 checks and reports node status to cluster nodes 3 based on initialization requests sent by cloud scheduling manager 1, allocates disk resources through disk manager 202, controls nodes based on node application or release requests sent by cloud scheduling manager 1, and sends power on / off messages to cluster nodes 3.

[0081] Cluster nodes 3 are deployed with a container orchestration system 4 for automated configuration.

[0082] In one embodiment, resource scheduler 101 dynamically adjusts resource allocation based on real-time scheduling data from diskless nodes and the workload requirements of node controller 2. It also dynamically requests and releases resources based on user demand for nodes, including resource power conservation. It also uses machine learning algorithms to predict node demand and optimize task allocation.

[0083] In one embodiment, the container orchestration system 4 used by the cloud scheduling manager 1 includes Kubernetes and Kubeedge, and the container orchestration system 4 used by the cluster nodes 3 also includes Kubeedge; alternatively, Docker Swarm or Apache Mesos is used for container orchestration. The container orchestration system 4 manages and coordinates containerized applications and their dependencies, ensuring they run in a secure, efficient, and reliable environment. In this system, it assists the cloud scheduling manager 1 in rapidly deploying and configuring cluster nodes 3, as well as enabling fast joining and exiting of cluster nodes 3.

[0084] The container orchestration system 4 facilitates simplified node deployment and dynamic cluster scaling. Node deployment simplification specifically provides an automated node configuration mechanism to simplify the deployment of diskless nodes. This implementation leverages Kubernetes' configuration capabilities, automating node configuration and deployment through self-starting services. Dynamic cluster scaling dynamically adjusts cluster size based on controller load status, enabling seamless expansion and scaling. This requires the scheduler's status monitoring of the node controller 2 and Kubernetes's own extension interfaces to enable the rapid addition and release of diskless nodes.

[0085] In one embodiment, the system includes a fault detection and recovery mechanism;

[0086] The cloud scheduling manager 1 also includes a fault detection and recovery device 102, and the node controller 2 also includes a fault collector 204. The fault detection and recovery device 102 is used to monitor and identify node faults in real time and perform fault processing. The fault collector 204 is used to check and report abnormal status of the cluster node 3, which includes faults. The functions of the fault detection and recovery device include real-time monitoring of node health status, automatic recovery of instance scheduling, and ensuring the continuous operation of the system. Specifically, distributed monitoring technology is used to identify and respond to node failures through a health check mechanism.

[0087] The node controller 2 further includes a power controller 203 , which is configured to provide power support to the node controller 2 under the constraint of power saving.

[0088] In one embodiment, the resource scheduler 101 predicts future node usage requirements based on existing instance applications or instance release requests, combined with the workload requirements of the node controller 2 and the requirements of power saving control, and dynamically adjusts the allocation scheduling of disk resources and cluster nodes 3.

[0089] This application also discloses a scheduling and resource management method for hard disk-free nodes, such as Figure 2 and Figure 3 As shown, including:

[0090] When the cloud scheduling manager 1 receives the instance application, it matches the optimal node according to the node information in the database 5 and the scheduling mechanism of the optimal node, and then sends the instance application message to the node controller 2 corresponding to the optimal node;

[0091] Node controller 2 allocates disk resources to the optimal node and wakes up the optimal node according to the instance application request;

[0092] The node controller 2 notifies the cloud scheduling manager 1 of the information of the allocated disk resources;

[0093] The cloud scheduling manager 1 updates the configuration storage and prepares the K8S scheduling configuration for the optimal node; and

[0094] After the optimal node is powered on, start the automated configuration software and add it to the K8S cluster.

[0095] In one embodiment, Figure 4 As shown, it also includes:

[0096] When the cloud scheduling manager 1 receives a request to release the instance of a specified node, it notifies the node controller 2 to reclaim the specified node and updates the node status according to the feedback message;

[0097] Node controller 2 sends a shutdown command to the designated node and monitors the status of the designated node;

[0098] After node controller 2 detects that the optimal node is shut down, it releases the disk resources used by the designated node.

[0099] In one embodiment, Figure 5 As shown, it also includes:

[0100] After node controller 2 detects an abnormality in the first node, it reports the abnormality to the cloud scheduler;

[0101] The cloud scheduler starts the standby node according to the standby node scheduling mechanism, determines the second node that matches the first node from the current standby nodes through the database 5, and updates the resource configuration of the first node to the second node;

[0102] The node controller 2 initiates scheduling of the second node, initiates a startup application for the second node, and updates the status of the first node and the second node.

[0103] In one embodiment, Figure 3 As shown, Figure 6 The scheduling mechanism of the optimal node is explained in detail.

[0104] Step 801: Check the resources of all current node controllers 2 and find the node controller 2 where the optimal node is located according to the algorithm;

[0105] Step 802, check the node status of the corresponding node controller 2, if there is a matching node, determine the matching node, if not, return to step 801;

[0106] Step 803: Check whether the node is in a ready state. If it is ready, check the node usage status. If it is in an idle state, directly initiate a node application; if it is not ready or in a non-idle state, continue with step 802.

[0107] Step 804: If the ready node is scheduled, a standby node application is initiated and the new node is placed in database 5 as a standby node.

[0108] Step 805: The standby node reaches the threshold value and no new nodes are added.

[0109] In one embodiment, Figure 4 As shown, the node status update specifically includes:

[0110] Keep node resources in the ready state and update the node usage status to the idle state.

[0111] In one embodiment, Figure 5 As shown in the figure, the standby node calling mechanism specifically includes:

[0112] Extract the resource configuration of the existing instance and find a backup node that matches the first node for scheduling, where:

[0113] The first node is searched first in the backup node under the same node controller 2 as the first node. If not found, the backup node is searched across node controllers 2.

[0114] It should be noted that in the claims and description of this patent, relational terms such as first and second, etc. are merely used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device that includes a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "including a" does not exclude the presence of other identical elements in the process, method, article or device that includes the element.

[0115] All documents mentioned in this application are incorporated herein by reference, just as if each document were incorporated herein by reference individually. It should also be understood that after reading the above teachings of the present invention, those skilled in the art may make various changes or modifications to the present invention, and that such equivalents also fall within the scope of the claims appended hereto.

Claims

1. A scheduling and resource management system without hard disk nodes, characterized in that: Includes cloud scheduling manager, node controller and cluster nodes: The cloud scheduling manager is connected to one or more node controllers, and each node controller is connected to one or more cluster nodes; The cloud scheduling manager includes a resource scheduler and a container orchestration system; the cloud scheduling manager is used to receive instance application or instance release requests, initialize network configuration, send a request to initialize the cluster node to the node controller through the resource scheduler, update node configuration information and status based on feedback from the node controller, perform node scheduling, send node application or release requests to the node controller, and dynamically configure and deploy cluster nodes in cooperation with the resource scheduler through the container orchestration system; The node controller includes a message distributor and a disk manager; the node controller is used to initialize local resources and process messages between the cloud scheduling manager and the cluster nodes through the message distributor; the node controller checks and reports node status to the cluster nodes based on the initialization request sent by the cloud scheduling manager, allocates disk resources through the disk manager, controls nodes based on node application or release requests sent by the cloud scheduling manager, and sends power on / off messages to the cluster nodes; the node controller also includes a power controller, which is used to provide power support to the node controller under the constraint of saving electricity; The cluster nodes are deployed with a container orchestration system, which is applied in an edge computing environment for automated configuration.

2. The scheduling and resource management system according to claim 1, characterized in that The container orchestration system used by the cloud scheduling manager includes Kubernetes and Kubeedge, and the container orchestration system used by the cluster nodes includes Kubeedge; or The scheduling and resource management system uses Docker Swarm or Apache Mesos for container orchestration.

3. The scheduling and resource management system according to claim 1, characterized in that The system includes a fault detection and recovery mechanism; The cloud scheduling manager also includes a fault detection and recovery device, and the node controller also includes a fault collector. The fault detection and recovery device is used to monitor and identify node faults in real time and perform fault processing. The fault collector is used to check and report abnormal states of the cluster nodes, including the faults.

4. The scheduling and resource management system according to claim 1, characterized in that The resource scheduler predicts the future node usage demand based on the existing instance application or instance release request, combined with the workload demand and power saving control requirements of the node controller, and dynamically adjusts the allocation scheduling of the disk resources and the cluster nodes.

5. A scheduling and resource management method for hard disk-free nodes, characterized in that: include: The method is used for a scheduling and resource management system for hard disk-free nodes, which includes a cloud scheduling manager, a node controller, and cluster nodes: The cloud scheduling manager is connected to one or more node controllers, and each node controller is connected to one or more cluster nodes; The cloud scheduling manager includes a resource scheduler and a container orchestration system; the cloud scheduling manager is used to receive instance application or instance release requests, initialize network configuration, send a request to initialize the cluster node to the node controller through the resource scheduler, update node configuration information and status based on feedback from the node controller, perform node scheduling, send node application or release requests to the node controller, and dynamically configure and deploy cluster nodes in cooperation with the resource scheduler through the container orchestration system; The node controller includes a message distributor and a disk manager; the node controller is used to initialize local resources and process messages between the cloud scheduling manager and the cluster nodes through the message distributor; the node controller checks and reports node status to the cluster nodes based on the initialization request sent by the cloud scheduling manager, allocates disk resources through the disk manager, controls nodes based on node application or release requests sent by the cloud scheduling manager, and sends power on / off messages to the cluster nodes; the node controller also includes a power controller, which is used to provide power support to the node controller under the constraint of saving electricity; The cluster nodes are deployed with a container orchestration system, which is applied in an edge computing environment for automated configuration; The method specifically includes: When the cloud scheduling manager receives an instance request, it matches the optimal node based on the node information in the database and the scheduling mechanism of the optimal node, and then sends the instance request message to the node controller corresponding to the optimal node; The node controller allocates disk resources to the optimal node and wakes up the optimal node according to the request of the instance application; The node controller notifies the cloud scheduling manager of information about the allocated disk resources; The cloud scheduling manager updates the configuration storage and prepares the K8S scheduling configuration for the optimal node; The power controller provides power support to the node controller and dynamically manages power for the hard disk-free node under the constraint of saving power; After the optimal node is powered on, the automated configuration software is started and added to the K8S cluster.

6. The scheduling and resource management method according to claim 5, characterized in that: Also includes: When the cloud scheduling manager receives a request to release an instance of a designated node, it notifies the node controller to reclaim the designated node and updates the node status according to the feedback message; The node controller sends a shutdown command to the designated node and monitors the status of the designated node; After the node controller monitors that the optimal node is shut down, it releases the disk resources used by the designated node.

7. The scheduling and resource management method according to claim 5, characterized in that: Also includes: After the node controller detects that an abnormality occurs in the first node, it reports the abnormality to the cloud scheduling manager; The cloud scheduling manager starts the standby node according to the standby node scheduling mechanism, determines a second node matching the first node from the current standby nodes through the database, and updates the resource configuration of the first node to the second node; The node controller initiates scheduling of the second node, initiates a startup application for the second node, and updates the status of the first node and the second node.

8. The scheduling and resource management method according to claim 5, characterized in that: The optimal node scheduling mechanism specifically includes: Step 801: Check the resources of all current node controllers and find the node controller where the optimal node is located according to the algorithm; Step 802: Check the node status of the corresponding node controller. If there is a matching node, determine the matching node. If not, return to step 801. Step 803: Check whether the node is in a ready state. If so, check the node usage state. If in an idle state, directly initiate a node application. If not ready or in a non-idle state, proceed to step 802. Step 804: If the ready node is scheduled, a standby node application is initiated and the new node is placed in the database as a standby node. Step 805: The standby node reaches the threshold value and no new nodes are added.

9. The scheduling and resource management method according to claim 6, characterized in that: The node status update specifically includes: Keep node resources in the ready state and update the node usage status to the idle state.

10. The scheduling and resource management method according to claim 7, characterized in that: The standby node calling mechanism specifically includes: Extract the resource configuration of the existing instance and search for a backup node that matches the first node for scheduling, where: The first node is searched first in the backup nodes under the same node controller as the first node. If the backup node is not found, the first node is searched for a backup node across the node controllers.

Images