Lightweight scheduling method and device for edge computing, equipment and medium

Abstract

This invention provides a lightweight scheduling method, apparatus, device, and medium for edge computing, comprising: deploying lightweight components on edge devices, including an Edgelet agent and a Containerd lightweight runtime; deploying a global scheduler in the cloud, dividing the edge local resource pool into edge autonomous systems based on the global scheduler, establishing a bidirectional communication link between the global scheduler and the edge autonomous systems, and realizing a cloud-edge collaborative channel based on the communication connection between the global scheduler and the edge autonomous systems; performing hierarchical elastic scheduling based on the edge local resource pool and the cloud-edge collaborative channel, including: edge nodes performing edge-level localized scheduling within the edge local resource pool according to their own load status; and the cloud-based global scheduler performing cross-domain overall scheduling from a global perspective based on the metadata reported by each edge autonomous system, realizing lightweight collaborative scheduling with the cloud managing the global and the edge managing the local.

Description

Technical Field

[0001] This invention relates to the fields of computer technology and Internet communication technology, specifically to a lightweight scheduling method, apparatus, device, and medium for edge computing. Background Technology

[0002] With the widespread adoption of edge computing, a large number of services are being processed locally at network edge nodes to reduce transmission latency and cloud pressure. However, edge nodes in edge computing scenarios generally suffer from limited hardware resources, making traditional centralized scheduling solutions in cloud centers unsuitable. The current mainstream Kubernetes edge scheduling architecture heavily relies on the central management node, posing a significant single point of failure risk; cloud-related anomalies can easily lead to overall scheduling failure. Furthermore, container images typically use full-scale transmission, consuming a large amount of scarce bandwidth resources in the edge environment, further exacerbating the transmission burden.

[0003] Furthermore, edge-cloud links are generally characterized by network instability and large bandwidth fluctuations. Traditional heartbeat detection mechanisms are prone to failure in weak network environments, making it difficult to accurately perceive node status. Existing elastic scaling and load balancing solutions mostly rely on unified monitoring data collection in the cloud, which not only results in high latency during scaling up and down but also fails to quickly respond to scenarios with dynamic changes in edge device topology, leading to unbalanced node load and insufficient stability of business operations.

[0004] Therefore, how to construct a lightweight cloud-edge collaborative scheduling method that supports edge local autonomy, adapts to weak network environments, and achieves efficient image distribution has become an urgent technical problem to be solved in the current edge computing scheduling field. Summary of the Invention

[0005] The main objective of this invention is to solve the problem of limited hardware resources for edge nodes in the prior art.

[0006] The first aspect of this invention provides a lightweight scheduling method for edge computing, comprising: Deploy lightweight components on edge devices, including: Edgelet agent and Containerd lightweight runtime; wherein: the Edgelet agent is used to monitor local network traffic and resource metrics, discover adjacent edge nodes, and build an edge local resource pool; the Containerd lightweight runtime is used to perform lifecycle management and runtime support for containerized business programs on edge devices according to the instructions issued by the Edgelet agent. Deploy a global scheduler in the cloud, divide the edge local resource pool into edge autonomous domains based on the global scheduler, establish a bidirectional communication link between the global scheduler and the edge autonomous domains, and realize a cloud-edge collaborative channel based on the communication connection between the global scheduler and the edge autonomous domains; Layered elastic scheduling is based on edge local resource pools and cloud-edge collaborative channels, including: edge nodes completing edge-level localized scheduling within the edge local resource pool according to their own load status; and the cloud global scheduler performing cross-domain overall scheduling from a global perspective based on the metadata reported by each edge autonomous region, realizing lightweight collaborative scheduling with the cloud managing the global and the edge managing the local.

[0007] Optionally, in a first implementation of the first aspect of the present invention, the deployment of lightweight components on the edge device includes: an Edgelet proxy and a Containerd lightweight runtime, comprising: Environmental adaptation testing is performed on the target edge device to ensure that the hardware architecture of the target edge device supports mainstream edge architectures, including ARM; at the same time, the remaining memory and storage resources of the target edge device meet the preset requirements for lightweight component deployment. Install the Edgelet agent on the target edge device that has passed the adaptation test, and complete the initial configuration of the Edgelet agent program to ensure that it achieves a stable connection with the underlying system of the target edge device that meets the preset requirements, so that the Edgelet agent has the ability to run and collect data. Simultaneously, the Containerd lightweight runtime is deployed on the same target edge device, and the communication link between the Edgelet agent and the Containerd lightweight runtime is established to realize instruction transmission and data interaction between the Edgelet agent and the Containerd lightweight runtime.

[0008] Optionally, in a second implementation of the first aspect of the present invention, the Edgelet proxy is used to monitor local network traffic and resource metrics, discover adjacent edge nodes, and construct an edge local resource pool, including: An eBPF network monitoring module is integrated into the Edgelet agent. Based on the eBPF network monitoring module, resource indicators including network traffic flow of target edge devices, inter-Pod communication data, CPU utilization, and memory usage are captured in real time to form a local monitoring dataset. The Edgelet agent integrates an LLDP protocol module. Based on the LLDP protocol module, it sends topology probe messages to surrounding edge nodes that meet preset requirements according to the local monitoring dataset. At the same time, it receives feedback messages from surrounding edge nodes that meet preset requirements, parses the feedback messages, and extracts node information of surrounding edge nodes, including device identifiers, resource status, and network addresses. Based on the node information parsed by the LLDP protocol module, it constructs an edge node topology structure locally, associates its own device with adjacent collaborating nodes, and integrates the available resource information of all nodes to form a unified edge local resource pool.

[0009] Optionally, in a third implementation of the first aspect of the present invention, during the process of establishing a bidirectional communication link between the global scheduler and the edge autonomous region, the transmission rules and data formats for uplink and downlink communication are configured respectively. Among them, configuring the uplink transmission rules of edge nodes in the edge autonomous domain includes: extracting features from the full resource data of the edge autonomous domain, including metadata such as resource tags, available resource types, and node running status, and encapsulating them according to a preset format, and setting a fixed synchronization period and a trigger-based reporting mechanism. Configure downlink transmission rules for the global scheduler, including: formulating global scheduling constraints based on global resource layout and business needs, standardizing and parsing the constraints to obtain instruction formats recognizable by the edge autonomous region, and realizing targeted or global policy distribution.

[0010] Optionally, in a fourth implementation of the first aspect of the present invention, a QUIC protocol transmission module is deployed on the communication data plane of the communication connection between the global scheduler and the edge autonomous region. The QUIC protocol is configured with parameters through the QUIC protocol transmission module so that the configured QUIC protocol can adapt to weak network and bandwidth fluctuation environments, including edge-cloud links.

[0011] Optionally, in a fifth implementation of the first aspect of the present invention, the method further includes: a cloud-edge image distribution system; The cloud-edge image distribution system includes: container images in the cloud and local images on edge nodes; A block-level differential algorithm module is deployed on the cloud-edge image distribution system. The block-level differential algorithm module splits the container image in the cloud into multiple data blocks and identifies them uniquely. The edge nodes retain the local image basic data blocks. When an image needs to be updated, the block-level differential algorithm module compares the differences between the new image in the cloud and the old image at the edge, extracting only the changed data blocks. An incremental image distribution channel based on the block-level differential algorithm is built, which encapsulates the image change data blocks identified in the cloud and transmits them to the target edge node via the QUIC protocol to complete the incremental update of the edge node image and realize the incremental distribution of container images.

[0012] Optionally, in a sixth implementation of the first aspect of the present invention, the hierarchical elastic scheduling based on the edge local resource pool and the cloud-edge collaborative channel includes: The system acquires local resource metrics data collected by the Edgelet agent in real time, compares the collected local resource metrics data with preset thresholds to determine the current load status of the edge nodes. When the collected local resource metrics data exceeds the preset thresholds, it is determined that the current edge node is in an overloaded state. In this case, the system retrieves node information from the local edge resource pool, filters out adjacent edge nodes that meet preset requirements for resource availability and network distance, formulates Pod migration paths and execution plans, and completes the migration operation of overloaded Pods locally on the edge according to the formulated migration plan. This achieves edge-level local autonomous scheduling and enables rapid adjustment of single-node load. The cloud-based global scheduler collects metadata uploaded by each edge autonomous system at a preset period. Based on the collected metadata, it constructs a global edge resource layout view. Based on the global edge resource layout view, it analyzes the load status and resource supply and demand of each edge domain and formulates a cross-regional global load balancing strategy. The cloud-based global scheduler then distributes the global load balancing strategy to the corresponding edge autonomous systems. Each edge autonomous system executes cross-domain resource allocation and task scheduling according to the global load balancing strategy, achieving cloud-level global resource balance.

[0013] A second aspect of the present invention provides a lightweight scheduling device for edge computing, comprising: The edge local resource pool construction module is used to deploy lightweight components on edge devices, including: Edgelet agent and Containerd lightweight runtime; wherein: the Edgelet agent is used to monitor local network traffic and resource indicators, discover adjacent edge nodes, and build an edge local resource pool; the Containerd lightweight runtime is used to perform lifecycle management and runtime support for containerized business programs on edge devices according to the instructions issued by the Edgelet agent. The cloud-edge collaboration channel construction module is used to deploy a global scheduler in the cloud, divide the edge local resource pool into edge autonomous domains based on the global scheduler, build a bidirectional communication link between the global scheduler and the edge autonomous domains, and realize the cloud-edge collaboration channel based on the communication connection between the global scheduler and the edge autonomous domains. The scheduling module is used for hierarchical elastic scheduling based on the edge local resource pool and cloud-edge collaborative channel. It includes: edge nodes completing edge-level localized scheduling within the edge local resource pool according to their own load status; and the cloud global scheduler performing cross-domain overall scheduling from a global perspective based on the metadata reported by each edge autonomous region, realizing lightweight collaborative scheduling with the cloud managing the global and the edge managing the local.

[0014] Optionally, in a first implementation of the second aspect of the present invention, the edge local resource pool construction module deploys lightweight components on the edge device, including: an Edgelet proxy and a Containerd lightweight runtime, comprising: Environmental adaptation testing is performed on the target edge device to ensure that the hardware architecture of the target edge device supports mainstream edge architectures, including ARM; at the same time, the remaining memory and storage resources of the target edge device meet the preset requirements for lightweight component deployment. Install the Edgelet agent on the target edge device that has passed the adaptation test, and complete the initial configuration of the Edgelet agent program to ensure that it achieves a stable connection with the underlying system of the target edge device that meets the preset requirements, so that the Edgelet agent has the ability to run and collect data. Simultaneously, the Containerd lightweight runtime is deployed on the same target edge device, and the communication link between the Edgelet agent and the Containerd lightweight runtime is established to realize instruction transmission and data interaction between the Edgelet agent and the Containerd lightweight runtime.

[0015] Optionally, in a second implementation of the second aspect of the present invention, the Edgelet proxy in the edge local resource pool construction module is used to monitor local network traffic and resource indicators, discover adjacent edge nodes, and construct an edge local resource pool, including: An eBPF network monitoring module is integrated into the Edgelet agent. Based on the eBPF network monitoring module, resource indicators including network traffic flow of target edge devices, inter-Pod communication data, CPU utilization, and memory usage are captured in real time to form a local monitoring dataset. The Edgelet agent integrates an LLDP protocol module. Based on the LLDP protocol module, it sends topology probe messages to surrounding edge nodes that meet preset requirements according to the local monitoring dataset. At the same time, it receives feedback messages from surrounding edge nodes that meet preset requirements, parses the feedback messages, and extracts node information of surrounding edge nodes, including device identifiers, resource status, and network addresses. Based on the node information parsed by the LLDP protocol module, it constructs an edge node topology structure locally, associates its own device with adjacent collaborating nodes, and integrates the available resource information of all nodes to form a unified edge local resource pool.

[0016] Optionally, in a third implementation of the second aspect of the present invention, during the process of building a bidirectional communication link between the global scheduler and the edge autonomous region in the cloud-edge collaborative channel construction module, the transmission rules and data formats for uplink and downlink communication are configured respectively. Among them, configuring the uplink transmission rules of edge nodes in the edge autonomous domain includes: extracting features from the full resource data of the edge autonomous domain, including metadata such as resource tags, available resource types, and node running status, and encapsulating them according to a preset format, and setting a fixed synchronization period and a trigger-based reporting mechanism. Configure downlink transmission rules for the global scheduler, including: formulating global scheduling constraints based on global resource layout and business needs, standardizing and parsing the constraints to obtain instruction formats recognizable by the edge autonomous region, and realizing targeted or global policy distribution.

[0017] Optionally, in a fourth implementation of the second aspect of the present invention, a QUIC protocol transmission module is deployed on the communication data plane of the communication connection between the global scheduler and the edge autonomous region. The QUIC protocol is configured with parameters through the QUIC protocol transmission module so that the configured QUIC protocol can adapt to weak network and bandwidth fluctuation environments, including edge-cloud links.

[0018] Optionally, in a fifth implementation of the second aspect of the present invention, the system further includes: a cloud-edge image distribution module, comprising: a cloud-edge image distribution system; The cloud-edge image distribution system includes: container images in the cloud and local images on edge nodes; A block-level differential algorithm module is deployed on the cloud-edge image distribution system. The block-level differential algorithm module splits the container image in the cloud into multiple data blocks and identifies them uniquely. The edge nodes retain the local image basic data blocks. When an image needs to be updated, the block-level differential algorithm module compares the differences between the new image in the cloud and the old image at the edge, extracting only the changed data blocks. An incremental image distribution channel based on the block-level differential algorithm is built, which encapsulates the image change data blocks identified in the cloud and transmits them to the target edge node via the QUIC protocol to complete the incremental update of the edge node image and realize the incremental distribution of container images.

[0019] Optionally, in a sixth implementation of the second aspect of the present invention, the scheduling module includes: The system acquires local resource metrics data collected by the Edgelet agent in real time, compares the collected local resource metrics data with preset thresholds to determine the current load status of the edge nodes. When the collected local resource metrics data exceeds the preset thresholds, it is determined that the current edge node is in an overloaded state. In this case, the system retrieves node information from the local edge resource pool, filters out adjacent edge nodes that meet preset requirements for resource availability and network distance, formulates Pod migration paths and execution plans, and completes the migration operation of overloaded Pods locally on the edge according to the formulated migration plan. This achieves edge-level local autonomous scheduling and enables rapid adjustment of single-node load. The cloud-based global scheduler collects metadata uploaded by each edge autonomous system at a preset period. Based on the collected metadata, it constructs a global edge resource layout view. Based on the global edge resource layout view, it analyzes the load status and resource supply and demand of each edge domain and formulates a cross-regional global load balancing strategy. The cloud-based global scheduler then distributes the global load balancing strategy to the corresponding edge autonomous systems. Each edge autonomous system executes cross-domain resource allocation and task scheduling according to the global load balancing strategy, achieving cloud-level global resource balance.

[0020] A third aspect of the present invention provides an electronic device, the electronic device comprising a memory and at least one processor, wherein the memory stores instructions; The at least one processor invokes the instructions in the memory to cause the electronic device to perform the steps of the lightweight scheduling method for edge computing as described above.

[0021] A fourth aspect of the present invention provides a computer-readable storage medium storing instructions that, when executed by a processor, implement the steps of the lightweight scheduling method for edge computing as described above.

[0022] Compared with the prior art, the present invention has the following beneficial effects: 1. This invention achieves lightweight deployment of edge components by deploying Edgelet agents and Containerd lightweight runtime on edge devices, solves the problem of limited hardware resources of edge nodes, reduces the resource consumption of edge devices, and ensures stable operation of components in resource-constrained edge environments; 2. This invention divides the edge local resource pool into edge autonomous domains, constructs an architecture of cloud-coordinated and edge-autonomous management, realizes the autonomous scheduling capability of edge nodes, avoids the single point of failure risk of traditional centralized scheduling, and can maintain local service operation even if the cloud is abnormal. 3. This invention enables stable cloud-edge communication in weak network and bandwidth fluctuation environments by deploying a QUIC protocol transmission module and configuring its parameters on the cloud-edge communication data plane, thereby enabling connectionless and packet loss-resistant characteristics. This solves the problem of unstable transmission of traditional transmission protocols in the edge-cloud link and reduces data loss rate and transmission latency. 4. This invention achieves incremental distribution of container images by deploying a block-level differential algorithm module in the cloud-edge image distribution system. Only the changed data blocks are transmitted, which reduces the amount of data by more than 90% compared with the traditional full image transmission, greatly saving scarce edge bandwidth and improving image update efficiency. 5. This invention achieves a collaborative scheduling mode of cloud-based global management and edge-based local management through hierarchical elastic scheduling. This not only solves the problems of cloud-based scheduling and scaling delays in traditional scheduling, but also solves the problems of isolated edge scheduling and waste of global resources, thereby improving the stability of edge service operation and the utilization rate of global resources. 6. By deploying an offline detection module and a local caching strategy, this invention enables continuous service operation when the edge is offline and completes incremental state synchronization after the network is restored, further improving the reliability and anti-interference capability of the scheduling system and adapting to scenarios where the edge-cloud link is frequently interrupted. Attached Figure Description

[0023] Other features, objects, and advantages of the present invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a first flowchart of a lightweight scheduling method for edge computing provided in an embodiment of the present invention.

[0024] Figure 2 This is a second flowchart of a lightweight scheduling method for edge computing provided in an embodiment of the present invention.

[0025] Figure 3 This is a schematic diagram of a lightweight scheduling device for edge computing provided in an embodiment of the present invention.

[0026] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0027] This invention provides a lightweight scheduling method, apparatus, device, and medium for edge computing, comprising: deploying lightweight components on an edge device, including an Edgelet agent and a Containerd lightweight runtime; wherein: the Edgelet agent is used to monitor local network traffic and resource indicators, discover adjacent edge nodes, and build an edge local resource pool; the Containerd lightweight runtime is used to perform lifecycle management and runtime support for containerized business programs on the edge device according to instructions issued by the Edgelet agent; deploying a global scheduler in the cloud, dividing the edge local resource pool into edge autonomous regions based on the global scheduler, establishing a bidirectional communication link between the global scheduler and the edge autonomous regions, and realizing a cloud-edge collaborative channel based on the communication connection between the global scheduler and the edge autonomous regions; performing hierarchical elastic scheduling based on the edge local resource pool and the cloud-edge collaborative channel, including: edge nodes completing edge-level localized scheduling within the edge local resource pool according to their own load status; and the cloud-based global scheduler performing cross-domain overall scheduling from a global perspective based on the metadata reported by each edge autonomous region, realizing lightweight collaborative scheduling with the cloud managing the global and the edge managing the local. This invention solves the problem of limited hardware resources for edge nodes in the prior art.

[0028] The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0029] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 The first embodiment of the lightweight scheduling method for edge computing in this invention includes: 101. Deploy lightweight components on edge devices, including: Edgelet agent and Containerd lightweight runtime; wherein: the Edgelet agent is used to monitor local network traffic and resource indicators, discover adjacent edge nodes, and build an edge local resource pool; the Containerd lightweight runtime is used to perform lifecycle management and operation support for containerized business programs on edge devices according to the instructions issued by the Edgelet agent. In this embodiment, environmental adaptation detection is performed on the target edge device to ensure that the hardware architecture of the target edge device supports mainstream edge architectures including ARM and adapts to the diverse hardware environments of the edge device. Meanwhile, the remaining memory and storage resources of the target edge device meet the preset requirements for deployment of Edgelet agent and Containerd lightweight runtime, avoiding component lag and crashes due to insufficient resources, and subsequent component deployment is only carried out on target edge devices that have passed the adaptation test.

[0030] This embodiment achieves compatibility and feasibility of edge component deployment by performing hardware architecture and resource adaptation testing on the target edge device, avoiding component deployment failures or running lag and crashes caused by hardware incompatibility and insufficient resources, and improving the deployment success rate.

[0031] In the edge device that has passed the adaptation test, install an Edgelet agent program with a memory footprint of less than 50MB and complete the initial configuration of the Edgelet agent program to ensure that it achieves a stable connection with the underlying system of the target edge device that meets the preset requirements, so that the Edgelet agent has the ability to run and collect data. Simultaneously, the Containerd lightweight runtime is deployed on the same target edge device, and the communication link between the Edgelet agent and the Containerd lightweight runtime is established to realize instruction transmission and data interaction between the Edgelet agent and the Containerd lightweight runtime, ensuring the basic operational support for containerized tasks at the edge.

[0032] An eBPF network monitoring module is integrated into the Edgelet agent. Leveraging the efficiency and low overhead of eBPF, real-time capture of network traffic and resource metrics for edge devices is achieved. Based on this eBPF network monitoring module, resource metrics including network traffic flow of target edge devices, inter-Pod communication data, CPU utilization, and memory usage are captured in real-time and preliminarily processed to form a local monitoring dataset. The preliminary processing includes data cleansing and deduplication. This embodiment, by integrating an eBPF network monitoring module into the Edgelet agent, achieves efficient and low-overhead real-time capture of local network traffic and resource metrics for edge nodes, forming an accurate local monitoring dataset. This provides reliable data support for subsequent load assessment and scheduling decisions, while simultaneously reducing resource consumption during the monitoring process.

[0033] An LLDP protocol module is integrated into the Edgelet agent, and the topology discovery capability of the LLDP protocol module is used to realize the automatic discovery of adjacent edge nodes. Based on the LLDP protocol module, topology probe messages are sent to surrounding edge nodes that meet preset requirements according to the local monitoring dataset. Simultaneously, feedback messages from these nodes are received, parsed, and node information including device identifiers, resource status, and network addresses is extracted. Based on the node information parsed by the LLDP protocol module, an edge node topology structure is constructed locally, associating the device with adjacent collaborating nodes. Available resource information from all nodes is integrated to form a unified edge local resource pool, completing the initial construction and information synchronization of the resource pool. This embodiment, by integrating the LLDP protocol module into the Edgelet agent, achieves automatic discovery of adjacent edge nodes and topology construction without manual configuration, adapting to scenarios with dynamic changes in edge node topology. Simultaneously, the unified edge local resource pool provides ample resource support for edge-localized scheduling.

[0034] Specifically, the Edgelet agent integrates its own available resource information with that of neighboring nodes, including resource types, total resources, idle resources, and resource tags for each node. This integrated node resource information is then standardized to form a unified edge-local resource pool. This pool contains all available resources of the current node and its neighboring collaborating nodes, and features resource query and filtering capabilities. During subsequent edge-local scheduling, available node resources can be directly retrieved from this resource pool.

[0035] The Edgelet agent updates the resource information of the local resource pool at a preset period, and synchronizes the resource status changes of each node in real time to ensure the accuracy of the resource pool information and provide reliable support for scheduling decisions.

[0036] 102. Deploy a global scheduler in the cloud, divide the edge local resource pool into edge autonomous domains based on the global scheduler, establish a bidirectional communication link between the global scheduler and the edge autonomous domains, and realize a cloud-edge collaborative channel based on the communication connection between the global scheduler and the edge autonomous domains; In this embodiment, during the process of establishing a bidirectional communication link between the global scheduler and the edge autonomous region, the transmission rules and data formats for uplink and downlink communication are configured respectively to ensure that standardized and normalized data interaction can be achieved after the communication link is established, thus avoiding subsequent data transmission chaos and format incompatibility issues.

[0037] The configuration includes uplink transmission rules for edge nodes within the edge autonomous system, including: extracting features from all resource data at the edge autonomous system, including metadata such as resource tags, available resource types, and node operating status, to ensure lightweight reporting and avoid uploading all data; and encapsulating the data according to a preset format, setting a mechanism that combines a fixed synchronization period with triggered reporting to ensure the timeliness and rationality of data reporting; wherein, the fixed synchronization period is when the edge autonomous system reports metadata to the cloud once according to a preset period, ensuring that the cloud has real-time knowledge of the overall status of the edge; triggered reporting is triggered immediately when an edge node experiences an abnormal state or a drastic change in resource status, without waiting for the fixed period, ensuring that the cloud can respond to abnormal situations in a timely manner.

[0038] Configure downlink transmission rules for the global scheduler, including: formulating global scheduling constraints based on global resource layout and business needs, standardizing and parsing the constraints to obtain instruction formats recognizable by the edge autonomous region, and realizing targeted or global policy distribution.

[0039] A QUIC protocol transmission module is deployed on the "communication data plane" of the communication connection between the global scheduler and the edge autonomous system. This module directly acts on the data transmission link of the cloud-edge bidirectional communication, and is responsible for the transmission and processing of all cloud-edge interactive data to ensure the stability of data transmission.

[0040] By configuring the parameters of the QUIC protocol through the QUIC protocol transmission module, its connectionless and packet loss-resistant transmission characteristics are enabled, replacing the traditional transmission protocol and adapting to the weak network and bandwidth fluctuation environment of the edge-cloud link, thus ensuring the stability of data transmission.

[0041] At the same time, the block-level differential algorithm module is deployed to the cloud-edge image distribution system to achieve efficient and lightweight distribution of container images. The cloud-side stores the complete container image as the basis for image distribution, and also deploys an image management module responsible for image version updates, block-level splitting, and other operations. The edge-side storage stores the local images that have been downloaded and retained by the edge nodes, serving as the basis for incremental updates. At the same time, it deploys an image receiving and update module, which is responsible for receiving image difference blocks transmitted from the cloud and completing image updates.

[0042] More specifically, a block-level differential algorithm module is deployed in the cloud-edge image distribution system. This module performs block-level splitting of container images in the cloud, dividing the image into multiple data blocks and assigning them unique identifiers. The basic data blocks of the image are retained at the edge nodes. When an image needs to be updated, the block-level differential algorithm compares the differences between the new image in the cloud and the old image at the edge, extracting only the changed data blocks. An incremental image distribution channel based on the block-level differential algorithm is built, which encapsulates the image change data blocks identified in the cloud and transmits them to the target edge nodes via the QUIC protocol, completing the incremental update of the image at the edge nodes and realizing the incremental distribution of container images.

[0043] 103. Based on edge local resource pools and cloud-edge collaborative channels, hierarchical elastic scheduling is implemented, including: edge nodes completing edge-level localized scheduling within the edge local resource pool according to their own load status; the cloud global scheduler performs cross-domain overall scheduling from a global perspective based on the metadata reported by each edge autonomous region, realizing lightweight collaborative scheduling with cloud managing the global and edge managing the local.

[0044] In this embodiment, the triggering rules for edge-level autonomous scheduling are configured, the threshold ranges of local resource indicators such as CPU, memory, and bandwidth are set, the judgment criteria for different states such as edge node overload and low load are clarified, and the corresponding scheduling operation types and execution priorities are defined to start the edge-level localized autonomous scheduling monitoring program. Specifically, the system acquires local resource metrics data collected by the Edgelet agent in real time, continuously compares the real-time data with preset thresholds to determine the current load status of edge nodes. When the monitoring program determines that an edge node is overloaded, it retrieves node information from the local edge resource pool, filters out adjacent edge nodes with idle resources and the closest network distance, formulates the Pod migration path and execution plan, and completes the migration operation of overloaded Pods locally on the edge according to the formulated migration plan. The entire process does not require cloud intervention, achieving millisecond-level edge-level local autonomous scheduling and completing rapid adjustment of single-node load. Configure the triggering conditions and scheduling cycle for cloud-level global scheduling, set the startup rules for cloud-level scheduling in scenarios such as multi-edge domain load imbalance and cross-regional resource demand allocation, and determine the global resource scanning and balancing cycle in minutes; the cloud scheduler collects the core metadata uploaded by each edge autonomous region according to the preset cycle, integrates it to form a global edge resource layout view, analyzes the load status and resource supply and demand of each edge domain, and formulates a cross-regional global load balancing strategy; the cloud scheduler distributes the global load balancing strategy to the corresponding edge autonomous regions, and each edge autonomous region executes cross-domain resource allocation and task scheduling according to the cloud strategy to achieve cloud-level global resource balancing in minutes.

[0045] This embodiment comprehensively addresses the pain points of existing technologies, such as limited edge resources, single point of failure in the center, bandwidth waste, poor adaptability to weak networks, and insufficient adaptability to dynamic topology changes, by constructing a lightweight scheduling system. It achieves lightweight, efficient, and stable scheduling in edge computing scenarios, thereby improving the overall performance and practicality of the edge computing system.

[0046] Please see Figure 2 The second embodiment of the lightweight scheduling method for edge computing in this invention includes: 201. Deploy lightweight components on edge devices, including: Edgelet agent and Containerd lightweight runtime; wherein: the Edgelet agent is used to monitor local network traffic and resource indicators, discover adjacent edge nodes, and build an edge local resource pool; the Containerd lightweight runtime is used to perform lifecycle management and operation support for containerized business programs on edge devices according to the instructions issued by the Edgelet agent. 202. Deploy a global scheduler in the cloud, divide the edge local resource pool into edge autonomous domains based on the global scheduler, establish a bidirectional communication link between the global scheduler and the edge autonomous domains, and realize a cloud-edge collaborative channel based on the communication connection between the global scheduler and the edge autonomous domains; 203. Based on edge local resource pools and cloud-edge collaborative channels, hierarchical elastic scheduling is implemented, including: edge nodes perform edge-level localized scheduling within the edge local resource pool according to their own load status; the cloud global scheduler performs cross-domain overall scheduling from a global perspective based on the metadata reported by each edge autonomous region, realizing lightweight collaborative scheduling with cloud managing the global and edge managing the local. 204. When the edge device goes offline, a local caching strategy is triggered to maintain service, and incremental state synchronization is completed after the network is restored; In this embodiment, an offline detection module is deployed on each edge node to monitor the connectivity status of the cloud-edge communication link in real time. When a link interruption is detected and the edge is offline, the local caching policy activation program is automatically triggered. The edge node autonomously manages local container tasks and resource allocation according to the locally pre-stored caching scheduling policy to maintain basic service operation, while recording key data such as node running status and task execution information during the offline period. When the offline detection module detects that the cloud-edge communication link is restored, the edge node's status incremental synchronization program is started. The key data recorded during the offline period is encapsulated according to the incremental synchronization rules and uploaded to the cloud to complete the alignment of the cloud and edge status data and restore the normal cloud-edge collaborative scheduling mode.

[0047] The lightweight scheduling method for edge computing in the embodiments of the present invention has been described above. The lightweight scheduling device for edge computing in the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 3 One embodiment of the lightweight scheduling device for edge computing in this invention includes: The edge local resource pool construction module 301 is used to deploy lightweight components on edge devices, including: Edgelet agent and Containerd lightweight runtime; wherein: the Edgelet agent is used to monitor local network traffic and resource indicators, discover adjacent edge nodes, and build an edge local resource pool; the Containerd lightweight runtime is used to perform lifecycle management and operation support for containerized business programs on edge devices according to the instructions issued by the Edgelet agent. In this embodiment, the edge local resource pool construction module 301 includes: The environment adaptation detection submodule 3011 is used to perform environment adaptation detection on the target edge device to ensure that the hardware architecture of the target edge device supports mainstream edge architectures including ARM; at the same time, the remaining memory and storage resources of the target edge device meet the preset requirements for lightweight component deployment. Edgelet agent deployment submodule 3012 is used to install the Edgelet agent in the target edge device that has passed the adaptation test, and complete the initial configuration of the Edgelet agent program to ensure that it achieves stable docking with the underlying system of the target edge device to meet the preset requirements, so that the Edgelet agent has the ability to run and collect data. The Containerd Lightweight Runtime Deployment Submodule 3013 is used to deploy the Containerd Lightweight Runtime in the same target edge device; The communication link establishment submodule 3013 is used to establish a communication link between the Edgelet agent and the Containerd lightweight runtime, so as to realize instruction transmission and data interaction between the Edgelet agent and the Containerd lightweight runtime. eBPF network monitoring submodule 3014 is used to integrate the eBPF network monitoring module in the Edgelet agent and capture resource indicators including network traffic flow of target edge devices, inter-Pod communication data, CPU utilization, and memory usage in real time based on the eBPF network monitoring module to form a local monitoring dataset. The LLDP protocol submodule 3015 is used to integrate the LLDP protocol module into the Edgelet agent. Based on the LLDP protocol module, it sends topology probe messages to surrounding edge nodes that meet preset requirements according to the local monitoring dataset, and simultaneously receives feedback messages from surrounding edge nodes that meet preset requirements. It parses the feedback messages to extract node information of surrounding edge nodes, including device identifiers, resource status, and network addresses. Based on the node information parsed by the LLDP protocol module, it constructs an edge node topology structure locally, associates its own device with adjacent collaborating nodes, and integrates the available resource information of all nodes to form a unified edge local resource pool.

[0048] The cloud-edge collaborative channel construction module 302 is used to deploy a global scheduler in the cloud, divide the edge local resource pool into edge autonomous domains based on the global scheduler, build a bidirectional communication link between the global scheduler and the edge autonomous domains, and realize a cloud-edge collaborative channel based on the communication connection between the global scheduler and the edge autonomous domains. In this embodiment, the cloud-edge collaborative channel construction module 302 includes: The communication configuration submodule 3021 is used to configure the transmission rules and data formats for uplink and downlink communication respectively during the process of establishing a bidirectional communication link between the global scheduler and the edge autonomous region; Among them, configuring the uplink transmission rules of edge nodes in the edge autonomous domain includes: extracting features from the full resource data of the edge autonomous domain, including metadata such as resource tags, available resource types, and node running status, and encapsulating them according to a preset format, and setting a fixed synchronization period and a trigger-based reporting mechanism. Configure downlink transmission rules for the global scheduler, including: formulating global scheduling constraints based on global resource layout and business requirements, standardizing and parsing the constraints to obtain instruction formats recognizable by the edge autonomous region, and realizing targeted or global policy distribution; QUIC protocol submodule 3022 is used to deploy a QUIC protocol transmission module on the communication data plane of the communication connection between the global scheduler and the edge autonomous region. The QUIC protocol transmission module is used to configure the parameters of the QUIC protocol so that the configured QUIC protocol can adapt to weak network and bandwidth fluctuation environments, including edge-cloud links. The block-level differential module 3023 is used to deploy a block-level differential algorithm module on the cloud-edge image distribution system. The block-level differential algorithm module splits the container image in the cloud into multiple data blocks and identifies them uniquely. The edge nodes retain the local image basic data blocks. When an image needs to be updated, the block-level differential algorithm module compares the differences between the new image in the cloud and the old image at the edge, extracting only the changed data blocks. An incremental image distribution channel based on the block-level differential algorithm is built, which encapsulates the image change data blocks identified in the cloud and transmits them to the target edge node via the QUIC protocol to complete the incremental update of the edge node image and realize the incremental distribution of container images.

[0049] The scheduling module 303 is used for hierarchical elastic scheduling based on the edge local resource pool and the cloud-edge collaborative channel. It includes: edge nodes completing edge-level localized scheduling within the edge local resource pool according to their own load status; and the cloud global scheduler performing cross-domain overall scheduling from a global perspective based on the metadata reported by each edge autonomous region, realizing lightweight collaborative scheduling with the cloud managing the global and the edge managing the local.

[0050] In this embodiment, the scheduling module 303 includes: The local scheduling submodule 3031 is used to acquire local resource indicator data collected by the Edgelet agent in real time, compare the collected local resource indicator data with preset thresholds to determine the current load status of the edge node; when the collected local resource indicator data exceeds the preset threshold, it is determined that the current edge node is in an overload state. Then, it retrieves the node information of the edge local resource pool, filters out the adjacent edge node with the nearest available resources and network distance that meets the preset requirements, formulates the Pod migration path and execution plan, and completes the migration operation of the overloaded Pod locally on the edge according to the formulated migration plan, realizing edge-level local autonomous scheduling and completing the rapid adjustment of single node load; The global scheduling submodule 3032 is used to collect metadata uploaded by each edge autonomous system according to a preset period through the cloud global scheduler. Based on the collected metadata uploaded by each edge autonomous system, a global edge resource layout view is constructed. Based on the global edge resource layout view, the load status and resource supply and demand of each edge domain are analyzed, and a cross-regional global load balancing strategy is formulated. The cloud global scheduler distributes the global load balancing strategy to the corresponding edge autonomous systems. Each edge autonomous system performs cross-domain resource allocation and task scheduling according to the global load balancing strategy to achieve cloud-level global resource balance.

[0051] The offline protection module 304 is used to trigger a local caching strategy to maintain service when the edge device is offline, and to complete incremental state synchronization after the network is restored. In this embodiment, the offline protection module 304 includes: deploying an offline detection module on each edge node to monitor the connectivity status of the cloud-edge communication link in real time; automatically triggering a local caching policy activation program when a link interruption is detected or the edge is offline; the edge node autonomously manages local container tasks and resource allocation according to the locally pre-stored caching scheduling policy to maintain basic service operation, while recording key data such as node running status and task execution information during the offline period; when the offline detection module detects that the cloud-edge communication link is restored, it starts the edge node's status incremental synchronization program, encapsulates the key data recorded during the offline period according to the incremental synchronization rules and uploads it to the cloud, completing the alignment of the cloud and edge status data and restoring the normal cloud-edge collaborative scheduling mode.

[0052] above Figure 3 The lightweight scheduling device for edge computing in this embodiment of the invention will be described in detail from the perspective of modular functional entities. The electronic device in this embodiment of the invention will be described in detail from the perspective of hardware processing.

[0053] Figure 4 This is a schematic diagram of the structure of an electronic device 700 provided in an embodiment of the present invention. The electronic device 700 can vary significantly due to different configurations or performance characteristics. It may include one or more central processing units (CPUs) 710 (e.g., one or more processors) and a memory 720, and one or more storage media 730 (e.g., one or more mass storage devices) for storing application programs 733 or data 732. The memory 720 and storage media 730 can be temporary or persistent storage. The program stored in the storage media 730 may include one or more modules (not shown in the diagram), each module including a series of instruction operations on the electronic device 700. Furthermore, the processor 710 may be configured to communicate with the storage media 730 and execute the series of instruction operations in the storage media 730 on the electronic device 700.

[0054] Electronic device 700 may also include one or more power supplies 740, one or more wired or wireless network interfaces 750, one or more input / output interfaces 750, and / or one or more operating systems 731, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will understand that... Figure 4 The illustrated electronic device structure does not constitute a limitation on electronic devices and may include more or fewer components than illustrated, or combine certain components, or have different component arrangements.

[0055] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when executed on a computer, cause the computer to perform the steps of a lightweight scheduling method for edge computing.

[0056] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the system, device, or unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0057] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0058] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An edge computing oriented lightweight scheduling method, characterized in that, include: Deploy lightweight components on edge devices, including: Edgelet agent and Containerd lightweight runtime; wherein: the Edgelet agent is used to monitor local network traffic and resource metrics, discover adjacent edge nodes, and build an edge local resource pool; the Containerd lightweight runtime is used to perform lifecycle management and runtime support for containerized business programs on edge devices according to the instructions issued by the Edgelet agent. Deploy a global scheduler in the cloud, divide the edge local resource pool into edge autonomous domains based on the global scheduler, establish a bidirectional communication link between the global scheduler and the edge autonomous domains, and realize a cloud-edge collaborative channel based on the communication connection between the global scheduler and the edge autonomous domains; Layered elastic scheduling is based on edge local resource pools and cloud-edge collaborative channels, including: edge nodes completing edge-level localized scheduling within the edge local resource pool according to their own load status; and the cloud global scheduler performing cross-domain overall scheduling from a global perspective based on the metadata reported by each edge autonomous region, realizing lightweight collaborative scheduling with the cloud managing the global and the edge managing the local.

2. The edge computing oriented lightweight scheduling method according to claim 1, characterized in that, The deployment of lightweight components on edge devices includes: Edgelet proxy and Containerd lightweight runtime, including: Environmental adaptation testing is performed on the target edge device to ensure that the hardware architecture of the target edge device supports mainstream edge architectures, including ARM; at the same time, the remaining memory and storage resources of the target edge device meet the preset requirements for lightweight component deployment. Install the Edgelet agent on the target edge device that has passed the adaptation test, and complete the initial configuration of the Edgelet agent program to ensure that it achieves a stable connection with the underlying system of the target edge device that meets the preset requirements, so that the Edgelet agent has the ability to run and collect data. Simultaneously, the Containerd lightweight runtime is deployed on the same target edge device, and the communication link between the Edgelet agent and the Containerd lightweight runtime is established to realize instruction transmission and data interaction between the Edgelet agent and the Containerd lightweight runtime.

3. The edge computing oriented lightweight scheduling method according to claim 1, characterized in that, The Edgelet proxy is used to monitor local network traffic and resource metrics, discover adjacent edge nodes, and build an edge-local resource pool, including: An eBPF network monitoring module is integrated into the Edgelet agent. Based on the eBPF network monitoring module, resource indicators including network traffic flow of target edge devices, inter-Pod communication data, CPU utilization, and memory usage are captured in real time to form a local monitoring dataset. The Edgelet agent integrates an LLDP protocol module. Based on the LLDP protocol module, it sends topology probe messages to surrounding edge nodes that meet preset requirements according to the local monitoring dataset. At the same time, it receives feedback messages from surrounding edge nodes that meet preset requirements, parses the feedback messages, and extracts node information of surrounding edge nodes, including device identifiers, resource status, and network addresses. Based on the node information parsed by the LLDP protocol module, it constructs an edge node topology structure locally, associates its own device with adjacent collaborating nodes, and integrates the available resource information of all nodes to form a unified edge local resource pool.

4. The lightweight scheduling method for edge computing according to claim 1, characterized in that, During the process of establishing a bidirectional communication link between the global scheduler and the edge autonomous region, the transmission rules and data formats for uplink and downlink communication are configured respectively; Among them, configuring the uplink transmission rules of edge nodes in the edge autonomous domain includes: extracting features from the full resource data of the edge autonomous domain, including metadata such as resource tags, available resource types, and node running status, and encapsulating them according to a preset format, and setting a fixed synchronization period and a trigger-based reporting mechanism. Configure downlink transmission rules for the global scheduler, including: formulating global scheduling constraints based on global resource layout and business needs, standardizing and parsing the constraints to obtain instruction formats recognizable by the edge autonomous region, and realizing targeted or global policy distribution.

5. The lightweight scheduling method for edge computing according to claim 1, characterized in that, A QUIC protocol transmission module is deployed on the communication data plane of the communication connection between the global scheduler and the edge autonomous region. The QUIC protocol is configured with parameters through the QUIC protocol transmission module so that the configured QUIC protocol can adapt to weak network and bandwidth fluctuation environments, including edge-cloud links.

6. The lightweight scheduling method for edge computing according to claim 1, characterized in that, The method also includes: a cloud-edge image distribution system; The cloud-edge image distribution system includes: container images in the cloud and local images on edge nodes; A block-level differential algorithm module is deployed on the cloud-edge image distribution system. The block-level differential algorithm module splits the container image in the cloud into multiple data blocks and identifies them uniquely. The edge nodes retain the local image basic data blocks. When an image needs to be updated, the block-level differential algorithm module compares the differences between the new image in the cloud and the old image at the edge, extracting only the changed data blocks. An incremental image distribution channel based on the block-level differential algorithm is built, which encapsulates the image change data blocks identified in the cloud and transmits them to the target edge node via the QUIC protocol to complete the incremental update of the edge node image and realize the incremental distribution of container images.

7. The lightweight scheduling method for edge computing according to claim 1, characterized in that, The hierarchical elastic scheduling based on edge local resource pools and cloud-edge collaborative channels includes: edge nodes performing edge-level localized scheduling within the edge local resource pool according to their own load status; and the cloud-based global scheduler performing cross-domain coordinated scheduling from a global perspective based on the metadata reported by each edge autonomous region, achieving lightweight collaborative scheduling with the cloud managing the global picture and the edge managing the local picture, including: The system acquires local resource metrics data collected by the Edgelet agent in real time, compares the collected local resource metrics data with preset thresholds to determine the current load status of the edge nodes. When the collected local resource metrics data exceeds the preset thresholds, it is determined that the current edge node is in an overloaded state. In this case, the system retrieves node information from the local edge resource pool, filters out adjacent edge nodes that meet preset requirements for resource availability and network distance, formulates Pod migration paths and execution plans, and completes the migration operation of overloaded Pods locally on the edge according to the formulated migration plan. This achieves edge-level local autonomous scheduling and enables rapid adjustment of single-node load. The cloud-based global scheduler collects metadata uploaded by each edge autonomous system at a preset period. Based on the collected metadata, it constructs a global edge resource layout view. Based on the global edge resource layout view, it analyzes the load status and resource supply and demand of each edge domain and formulates a cross-regional global load balancing strategy. The cloud-based global scheduler then distributes the global load balancing strategy to the corresponding edge autonomous systems. Each edge autonomous system executes cross-domain resource allocation and task scheduling according to the global load balancing strategy, achieving cloud-level global resource balance.

8. A lightweight scheduling device for edge computing, characterized in that, include: The edge local resource pool construction module is used to deploy lightweight components on edge devices, including: Edgelet agent and Containerd lightweight runtime; wherein: the Edgelet agent is used to monitor local network traffic and resource indicators, discover adjacent edge nodes, and build an edge local resource pool; the Containerd lightweight runtime is used to perform lifecycle management and runtime support for containerized business programs on edge devices according to the instructions issued by the Edgelet agent. The cloud-edge collaboration channel construction module is used to deploy a global scheduler in the cloud, divide the edge local resource pool into edge autonomous domains based on the global scheduler, build a bidirectional communication link between the global scheduler and the edge autonomous domains, and realize the cloud-edge collaboration channel based on the communication connection between the global scheduler and the edge autonomous domains. The scheduling module is used for hierarchical elastic scheduling based on the edge local resource pool and cloud-edge collaborative channel. It includes: edge nodes completing edge-level localized scheduling within the edge local resource pool according to their own load status; and the cloud global scheduler performing cross-domain overall scheduling from a global perspective based on the metadata reported by each edge autonomous region, realizing lightweight collaborative scheduling with the cloud managing the global and the edge managing the local.

9. An electronic device comprising a memory and at least one processor, wherein the memory stores instructions; The at least one processor invokes the instructions in the memory to cause the electronic device to perform the steps of the lightweight scheduling method for edge computing as described in any one of claims 1-7.

10. A computer-readable storage medium storing instructions thereon, characterized in that, When the instructions are executed by the processor, they implement the steps of the lightweight scheduling method for edge computing as described in any one of claims 1-7.

Images