A cluster state synchronization method and system for smart grid

Abstract

The present invention discloses a cluster state synchronization method and system for smart grids, which relate to the field of edge computing technology. The present invention obtains network delay information between all nodes, divides autonomous regions and determines seed nodes based on the network delay information; establishes a synchronization path between the management node, the seed node and its neighboring nodes; monitors the connection mode between each autonomous region and determines the connection window; obtains the status information of other nodes through the seed node, and transmits all the status information of the autonomous region to the management node along the synchronization path within the connection window to complete cluster state synchronization; and outputs all the obtained status information through the management node. The present invention can solve the performance and availability problems of cluster state synchronization caused by network connection mode and delay mutation in edge network environment.

Description

Technical Field

[0001] The present invention relates to the field of edge computing technology, and in particular to a cluster state synchronization method and system for smart grids. Background Art

[0002] With the increasing popularity of edge computing concepts and smart terminal devices, the collaborative integration of "cloud-edge-end" has become a hot topic and is widely used to support typical application scenarios such as smart grids. Currently, smart grid scenarios include a large number of fixed-site edge computing servers and a vast number of sensors such as smart meters. There are also a large number of devices such as patrol drones and operation and maintenance robots that require periodic movement to support applications such as routine inspections and condition monitoring and alarms. In these applications, task allocation and scheduling rely on the status information of the edge cluster to efficiently utilize edge computing resources. For example, information such as the available computing power, available memory, and available disk space of each node in the edge cluster needs to be aggregated to certain decision nodes in the cluster. Data such as the calculation results of the edge computing nodes needs to be synchronized within a specific range to support task scheduling and allocation.

[0003] However, in the smart grid edge network environment, network connection modes and network latency may change frequently, and cluster state synchronization faces challenges. Edge environments usually do not have stable wired or wireless network connections, and periodic and intermittent network connection interruptions may occur during the execution of some tasks, making network partitioning the norm. For example, when performing maintenance tasks such as line patrol, operation and maintenance drones cannot always maintain connection with the ground cluster. In each cycle, only part of the connection window may be able to exchange data with the ground cluster. However, existing methods such as data distribution, state synchronization, and data consistency assurance usually regard network partitioning as an abnormal situation equivalent to node failure, and use sequential or eventual consistency assurance mechanisms to deal with this failure scenario, introducing unnecessary failure recovery overhead; on the other hand, because existing methods fail to fully consider the periodic characteristics of network interruptions, they may introduce frequent and unnecessary node online and offline, cluster topology maintenance and other node life cycle related operations, which may have a significant impact on cluster performance and availability.

[0004] In summary, there is an urgent need for a cluster state synchronization method and system for smart grids. According to the characteristics of the smart grid edge network, a corresponding node, cluster topology and network connection mode discovery and maintenance mechanism should be designed. The characteristics of the periodic network connection of nodes in the network should be perceived, and the cluster should be guided to perform state synchronization operations at the appropriate time to reduce the impact of network delay and connection mode mutations, and improve the performance and availability of cluster state synchronization. Summary of the Invention

[0005] The purpose of the present invention is to propose a cluster state synchronization method and system for smart grids to solve the performance and availability problems of cluster state synchronization caused by network connection mode and delay mutation in edge network environment.

[0006] The technical solutions adopted by the present invention to achieve the above-mentioned purpose are as follows:

[0007] A cluster state synchronization method for a smart grid, wherein the cluster includes a management node and multiple working nodes, and the method includes the following steps:

[0008] Get network delay information between all nodes;

[0009] Based on the acquired network delay information, multiple nodes whose network delays are less than a preset threshold are divided into the same autonomous region. For any two autonomous regions, the two nodes with the smallest network delays belonging to the two autonomous regions are used as seed nodes.

[0010] Establish synchronization paths between the management node, the seed node, and the neighboring nodes in the same autonomous area of ​​the seed node;

[0011] Monitor the connection patterns between autonomous regions and determine the connection windows between the autonomous region where the management node is located and other autonomous regions;

[0012] The seed node in each autonomous region obtains the status information of other nodes in the same autonomous region. Within the above connection window, the seed node transmits all the status information of the autonomous region to the management node along the synchronization path to complete the cluster status synchronization;

[0013] Outputs all status information obtained through the management node.

[0014] Furthermore, the management node collects the heartbeat information of all nodes in the autonomous area through the seed node, extracts features from the heartbeat information, obtains the interval and period of each heartbeat information, and judges the node survival and connection periodicity; then maintains a bitmap of heartbeat information for each autonomous area, and uses a binary bit to mark the survival of each node; for disconnected autonomous areas, use a binary bit to mark the survival of the entire area.

[0015] Furthermore, for nodes that fail within the connection window, the connectivity from the neighboring nodes of the failed node to other nodes in the autonomous region where the failed node is located is first determined to confirm the failure range; for single-node failure, its neighboring nodes use a fixed interval to send heartbeat detection packets to the failed node, and the management node uses an exponential back-off mechanism to detect the health status of the failed node and obtain the connection window of the management node to the autonomous region; for multiple-node failure, the connection window size of the autonomous region where the node is located is obtained, with 0.1 times the connection window size as the initial interval. When the time to the next connection window is less than 1 / 3 of the interval between the two connection windows, the detection interval is reduced to 0.02 times the connection window size until a connection window is detected to restore the connection with the failed node.

[0016] Furthermore, the same mark is set for nodes whose network delay is less than a preset threshold, and the autonomous regions are divided according to the set mark; for failed nodes, they are added to the autonomous region divided last time. If the autonomous region divided last time does not exist, they are added to the autonomous region to which the random neighbor node belongs, and the autonomous region allocation of all nodes is completed.

[0017] Furthermore, for every two autonomous regions, the seed node connectivity data is obtained every fixed time window, and the average delay and delay variance in the past time window are continuously recorded; when the average delay and delay variance deviate from the average value of the past time window by more than a preset threshold, the time period when the network delay was last stable is recorded as the connection window of the two autonomous regions.

[0018] Furthermore, a data synchronization topology is created for each autonomous region, and a spanning tree is constructed with the management node that initiates state synchronization as the root node, and the seed node of the autonomous region as its child node; the neighbor nodes of the seed node are used as the lower-level child nodes of the seed node in the spanning tree to form a complete synchronization path.

[0019] Furthermore, when two autonomous regions are located in a connection window, the two autonomous regions are regarded as the same autonomous region and a synchronization path is generated to perform data exchange within the connection window.

[0020] Furthermore, the seed node saves the latest state information snapshot and records the synchronization progress; when the state synchronization is performed within the connection window, the synchronization progress is updated.

[0021] Furthermore, when it is impossible to connect to the corresponding autonomous area within the connection window, all unsent state information is merged and state synchronization is attempted again in the next connection window.

[0022] A cluster state synchronization system for smart grids, wherein the cluster includes a management node and multiple working nodes. The system includes: an agent module deployed on the working node, and a node management module, a cluster management module, a topology management module, and a state management module deployed on the management node; wherein:

[0023] The proxy module is responsible for obtaining network delay information between all nodes;

[0024] The node management module is responsible for monitoring the survival status and network delay information of all nodes;

[0025] The cluster management module is responsible for grouping multiple nodes whose network delays between them are less than a preset threshold into the same autonomous region based on the acquired network delay information. For any two autonomous regions, the two nodes with the lowest network delays belonging to these two autonomous regions are used as seed nodes. The module also monitors the connection patterns between the autonomous regions and determines the connection windows between the autonomous region where the management node is located and other autonomous regions.

[0026] The topology management module is responsible for establishing synchronization paths between the management node, seed node, and neighboring nodes in the same autonomous area of ​​the seed node;

[0027] The state management module is responsible for enabling seed nodes within each autonomous region to obtain the state information of their neighbor nodes. Within the above-mentioned connection window, the seed node, as a working node, transmits all state information of itself and its neighbor nodes to the management node along the synchronization path to complete cluster state synchronization, and outputs all obtained state information through the management node.

[0028] Compared with the prior art, the present invention has the following advantages:

[0029] 1. For scenarios where edge nodes and small clusters (i.e., autonomous regions) in smart grids are regularly disconnected, this approach can effectively reduce fault recovery overhead and improve cluster state synchronization performance.

[0030] 2. When there are many states to be synchronized and they accumulate, the present invention can effectively reduce the space occupied when maintaining the synchronization state through state information snapshots and an improved fault tolerance mechanism, thereby improving cluster state synchronization performance;

[0031] 3. When the connection status changes, the method proposed in the present invention can quickly perceive and adapt to the new cluster topology and connection mode, thereby improving cluster availability. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 1 is a module architecture diagram of a cluster state synchronization system for a smart grid in an embodiment;

[0033] Figure 24 is a deployment diagram of a cluster state synchronization system for a smart grid in an embodiment. DETAILED DESCRIPTION

[0034] In order to make the various technical features and advantages or technical effects of the above technical solutions of the present invention more obvious and easy to understand, they are described in detail below with reference to the accompanying drawings.

[0035] The embodiment of the present invention proposes a cluster state synchronization method and system for smart grid, the module composition of the system is as follows: Figure 1 As shown in the figure, it mainly includes cluster management module, node management module, topology management module, state management module and agent module. The deployment, main responsibilities and interaction methods of each module mainly include:

[0036] The cluster consists of management nodes and worker nodes. The cluster status is synchronized between worker nodes. The cluster management module, node management module, topology management module, and state management module are deployed on the management node, and the agent module is deployed on the worker node.

[0037] The proxy module interacts with the state management module and the node management module, is responsible for reporting network status information such as network latency of each node in the cluster, and performs specific cluster state synchronization tasks;

[0038] The node management module interacts with the cluster management module, topology management module, and agent module, and is responsible for maintaining a global view of the nodes in the cluster and performing node lifecycle management.

[0039] The cluster management module interacts with the node management module and the topology management module. It is responsible for determining autonomous regions based on the current and historical network connection status of the cluster and maintaining basic information such as the connection window of each autonomous region in the cluster. An autonomous region is a small cluster in the cluster with a stable network topology and interconnected nodes.

[0040] The topology management module interacts with the cluster management module and the node management module, and is responsible for recording and maintaining the connectivity of each node in the cluster, and determining the node status synchronization path based on the node connection mode;

[0041] The state management module maintains the latest state within the cluster and interacts with the topology management module and the proxy module to trigger cluster state synchronization based on the current network connectivity and mode.

[0042] Users or upper-layer applications interact with the state management module to define the state information that needs to be monitored and synchronized, and obtain the latest state of the cluster.

[0043] As a preferred implementation, the specific steps performed by the agent module are as follows:

[0044] 1) Regularly detect the network delay between the current node and neighboring nodes and report it to the node management module;

[0045] 2) Continuously try to establish a connection with the management node. When the connection window with the management node is entered and the connection is successfully established, it interacts with the cluster management module and the node management module to report the network delay and obtain the autonomous area where the current node is located;

[0046] 3) If it is a seed node, it is also responsible for collecting status, merging status, and forwarding status to other autonomous regions;

[0047] 4) If it is not a seed node, it obtains the status information that needs to be synchronized from the seed node and sends its own status;

[0048] 5) Continue to execute steps 1) to 4) to complete state synchronization.

[0049] As a preferred implementation, the specific steps performed by the node management module are as follows:

[0050] 1) Interact with the topology management module to obtain current node connectivity;

[0051] 2) Interact with the cluster management module to obtain the current autonomous area division;

[0052] 3) Interact with the agent module on the autonomous region seed node to continuously collect the heartbeat information of all nodes in each autonomous region;

[0053] 4) Extract features from the heartbeat information, obtain the interval and period of each heartbeat information, and judge the node viability and connection periodicity;

[0054] 5) A bitmap of heartbeat information is maintained for each autonomous region, and a disconnected autonomous region is marked with one binary bit to reduce space usage;

[0055] 6) For nodes that are within the connection window but fail due to a fault or network partition, first determine the connectivity of the failed node's neighbor nodes to other nodes in the failed node's autonomous region to confirm the failure scope;

[0056] 7) In the event of a single node failure, its neighboring nodes send heartbeat detection packets to the failed node at fixed intervals; the management node uses an improved exponential backoff mechanism to detect the health status of the failed node, interacts with the cluster management module to obtain the connection window from the management node to the autonomous area, and sends heartbeat detection packets to the failed node starting at 100 millisecond intervals. When a timeout is detected, the next retry interval is twice the previous one, and the maximum retry interval is 0.5 times the connection window size.

[0057] 8) For multiple node failures, the system first interacts with the cluster management module to obtain the connection window size of the autonomous region where the node is located. 0.1 times the connection window size is used as the initial interval. When the time until the next connection window is less than one-third of the interval between two connection windows, the detection interval is reduced to 0.02 times the connection window to restore the connection with the failed node in the next connection window as much as possible.

[0058] 9. Repeat steps 1) to 8) to continuously record the global view and lifecycle information of the node for use by other modules.

[0059] As a preferred implementation, the cluster management module performs the following specific steps:

[0060] 1) Interact with the node management module to obtain the node list and the inter-node delay data reported by the proxy module;

[0061] 2) Interact with the topology management module to obtain node connectivity data;

[0062] 3) For each surviving node in the node list, determine the delay from the node to its neighboring nodes;

[0063] 4) When the delay between two nodes is less than a certain threshold, that is, when the physical distance between the nodes is close, they are considered to be in the same autonomous region. If the two nodes do not have autonomous region tags, the autonomous region tags are set for the two nodes. Otherwise, the two nodes and the nodes with the same autonomous region tags as the two nodes are set to the same tag;

[0064] 5) Repeat steps 3) to 4) to set an autonomous region mark for each node, and preliminarily complete the autonomous region division;

[0065] 6) For failed nodes, add them to the autonomous region of the previous round. If there is no autonomous region in the previous round, add them to the autonomous region of a random neighbor node. This completes the autonomous region allocation for all nodes.

[0066] 7) For each autonomous region, compare the delays from surviving nodes in the autonomous region to nodes outside the autonomous region, and select the two nodes with the smaller delay as the seed nodes between the two autonomous regions;

[0067] 8) Traverse all autonomous regions found in step 5) in sequence and determine the seed nodes from each autonomous region to other autonomous regions;

[0068] 9) For every two autonomous regions, obtain seed node connectivity data from the topology management module every fixed time window, and continuously record the average delay and delay variance in the past time window;

[0069] 10) When the average delay and delay variance change suddenly, record the time period when the delay was last stable as the connection window of the two autonomous regions;

[0070] 11) Repeat steps 1) to 10) to continuously update the connection mode and autonomous area for use by other modules.

[0071] As a preferred implementation, the specific steps performed by the topology management module are as follows:

[0072] 1) Interact with the cluster management module to obtain current autonomous area information and connection mode information;

[0073] 2) Interact with the node management module to obtain the current node global view and life cycle;

[0074] 3) Create a data synchronization topology for each autonomous region collected in step 1). Build a spanning tree with the node that initiated state synchronization as the root and the seed node of the autonomous region as its child nodes. Then, for each node in the spanning tree, add its neighbor nodes as its child nodes in the spanning tree, thus forming a complete synchronization path.

[0075] 4) Based on the connection mode information collected in step 1), when two autonomous regions are within a connection window, their synchronization paths are merged, i.e., the two autonomous regions are considered as one and a synchronization path is generated to support fast data exchange within the connection window;

[0076] 5) Repeat steps 1) to 4) to continuously update node connectivity and synchronization paths for use by other modules.

[0077] As a preferred implementation, the specific steps performed by the state management module are as follows:

[0078] 1) Interact with the topology management module to obtain the current connectivity and synchronization path;

[0079] 2) Interact with the agent modules on the seed nodes in each autonomous region, send connectivity and synchronization paths, and trigger state synchronization;

[0080] 3) The seed nodes in the autonomous region interact with other nodes to determine the status information and synchronization scope that need to be synchronized. Each node regularly synchronizes status information to the seed nodes.

[0081] 4) The seed node saves the most recent state information snapshot and records the synchronization progress. When it is in the connection window, it performs state synchronization, sends the state to the management node and the seed nodes in the target autonomous area, and updates the synchronization progress.

[0082] 5) When the connection window fails to connect to the corresponding autonomous region, all unsent state information will be merged and state synchronization will be attempted again in the next connection window;

[0083] 6) When a seed node receives status information from other seed nodes, it synchronizes within its own autonomous region.

[0084] A specific example is provided below:

[0085] 1) If Figure 2 As shown in the figure, the cluster consists of a ground cluster and two small drone clusters (i.e., small clusters 1 and 2). The ground small cluster includes a management node (i.e., node 1) and two computing nodes (i.e., nodes 2 and 3, which are working nodes). The two small drone clusters each include three working nodes (i.e., nodes 4 to 6 and nodes 7 to 9).

[0086] 2) Two small drone clusters collect and process image information from different areas. Based on the currently available resources of the cluster, they decide and schedule nodes to complete data processing and transmit the results back.

[0087] 3) The state to be synchronized includes the resource information of each node in each drone cluster and the task processing result information. Among them, the resource information needs to be reported to node 1, and the task processing result information needs to be reported to node 1 and all nodes 2 to 9 in each drone cluster;

[0088] 4) The proxy modules on all nodes 2 to 9 first monitor the network delay between themselves and their neighboring nodes, and report to the node management module on node 1 when they can connect to node 1;

[0089] 5) The cluster management module on node 1 collects the latency of all nodes 1-9, sets a threshold of 2 milliseconds, and finds that the latency between nodes 1-3 is 1 millisecond, the latency between nodes 4-6 is 1 millisecond, and the latency between nodes 7-9 is 1 millisecond. Therefore, nodes 1-3 are marked as autonomous region 1, nodes 4-6 are marked as autonomous region 2, and nodes 7-9 are marked as autonomous region 3.

[0090] 6) The cluster management module collects the information that the delay from node 1 to node 4 is the smallest between autonomous regions 1 and 2, and marks nodes 1 and 4 as seed nodes; the delay from node 1 to node 8 is the smallest between autonomous regions 1 and 3, and marks nodes 1 and 8 as seed nodes; the delay from node 5 to node 7 is the smallest between autonomous regions 2 and 3, and marks nodes 5 and 7 as seed nodes;

[0091] 7) The cluster management module sets the monitoring time window to 10 seconds and detects that the connection pattern between autonomous region 1 and autonomous region 2 is connected for 50 seconds and disconnected for 50 seconds, the connection pattern between autonomous region 1 and autonomous region 3 is connected for 100 seconds and disconnected for 30 seconds, and the connection pattern between autonomous region 2 and autonomous region 3 is connected for 20 seconds and disconnected for 60 seconds. The cluster management module saves this connection pattern and connection window size (i.e., 50 seconds, 100 seconds, and 20 seconds).

[0092] 8) The topology management module generates synchronization paths. Taking autonomous region 2 as an example, when reporting resource status, the agent modules on nodes 5 and 6 transmit the status information to node 4. Node 4 then transmits the status information to node 1 when it enters the connection window. When in the connection window, nodes 4 and 6 transmit the status information directly to node 1.

[0093] 9) The state management module continuously collects state information from autonomous regions 2 and 3, and sends the state and topology information to be collected to the agent module of each node; the nodes in autonomous regions 2 and 3 follow the current topology and complete data synchronization;

[0094] 10) The node management module continuously monitors the survival status and delay information of all nodes;

[0095] 11) When node 6 in autonomous region 2 fails, nodes 4-5 detect the failure and report it to the node management module. Nodes 4-5 use a fixed interval of 50 milliseconds to detect the health status of node 6, while node 1 uses a modified exponential back-off method to detect the health status of node 6. The maximum detection interval is half the connection window between autonomous regions 1 and 2, that is, 25 seconds.

[0096] 12) When node 9 in autonomous region 3 switches to connecting to small cluster 1, the cluster management module detects a sudden change in node 9's connection mode in the next window and re-divides the autonomous region, placing nodes 4-6 and node 9 into the same autonomous region.

[0097] 13) The cluster management module continuously monitors the changes in node latency and completes the autonomous region division.

[0098] Although the present invention has been disclosed as above by way of embodiments, they are not intended to limit the present invention. Any appropriate modification or equivalent substitution of the technical solution of the present invention by a person skilled in the art should be included in the protection scope of the present invention. The protection scope of the present invention shall be based on that defined in the claims.

Claims

1. A cluster state synchronization method for smart grid, characterized in that: The cluster includes a management node and multiple worker nodes. The method includes the following steps: Get network delay information between all nodes; Based on the acquired network delay information, multiple nodes whose network delays are less than a preset threshold are divided into the same autonomous region. For any two autonomous regions, the two nodes with the smallest network delays belonging to the two autonomous regions are used as seed nodes. Establish synchronization paths between the management node, the seed node, and the neighboring nodes in the same autonomous area of ​​the seed node; Monitor the connection patterns between autonomous regions and determine the connection windows between the autonomous region where the management node is located and other autonomous regions; The seed node in each autonomous region obtains the status information of other nodes in the same autonomous region. Within the above connection window, the seed node transmits all the status information of the autonomous region to the management node along the synchronization path to complete the cluster status synchronization; Outputs all status information obtained through the management node.

2. The method according to claim 1, wherein The management node collects heartbeat information from all nodes in the autonomous area through the seed node, extracts features from the heartbeat information, obtains the interval and period of each heartbeat information, and determines the node survival and connection periodicity; then, a bitmap of heartbeat information is maintained for each autonomous area, and the survival of each node is marked with a binary bit; for disconnected autonomous areas, a binary bit is used to mark the survival of the entire area.

3. The method according to claim 1, wherein For nodes that fail within the connection window, the connectivity of the failed node's neighboring nodes to other nodes in the autonomous region where the failed node is located is first determined to confirm the failure range; for single-node failure, its neighboring nodes use a fixed interval to send heartbeat detection packets to the failed node, and the management node uses an exponential back-off mechanism to detect the health status of the failed node and obtain the connection window of the management node to the autonomous region; for multiple-node failure, the connection window size of the autonomous region where the node is located is obtained, with 0.1 times the connection window size as the initial interval. When the time to the next connection window is less than 1 / 3 of the interval between the two connection windows, the detection interval is reduced to 0.02 times the connection window size until a connection window is detected to restore the connection with the failed node.

4. The method according to claim 1, wherein The same mark is set for nodes whose network delay is less than the preset threshold, and the autonomous regions are divided according to the set marks. For failed nodes, they are added to the autonomous region divided last time. If the autonomous region divided last time does not exist, they are added to the autonomous region to which a random neighbor node belongs, completing the autonomous region allocation for all nodes.

5. The method according to claim 1, wherein For every two autonomous regions, seed node connectivity data is obtained every fixed time window, and the average delay and delay variance in the past time window are continuously recorded; when the average delay and delay variance deviate from the average value of the past time window by more than the preset threshold, the time period when the network delay was last stable is recorded as the connection window of the two autonomous regions.

6. The method according to claim 1, wherein Create a data synchronization topology for each autonomous region, build a spanning tree with the management node that initiates state synchronization as the root node, and the seed node of the autonomous region as its child nodes; The neighbor nodes of the seed node are used as the lower-level child nodes of the seed node in the spanning tree to form a complete synchronization path.

7. The method according to claim 1, wherein When two autonomous areas are located in a connection window, the two autonomous areas are regarded as the same autonomous area and a synchronization path is generated to perform data exchange within the connection window.

8. The method according to claim 1, wherein The seed node saves the latest status information snapshot and records the synchronization progress; when the status synchronization is completed within the connection window, the synchronization progress is updated.

9. The method according to claim 1, wherein When it is impossible to connect to the corresponding autonomous area within the connection window, all unsent state information is merged and state synchronization is attempted again in the next connection window.

10. A cluster state synchronization system for smart grids, used to implement the method according to any one of claims 1 to 9, characterized in that: The cluster includes a management node and multiple working nodes. The system includes: an agent module deployed on the working node, and a node management module, a cluster management module, a topology management module, and a state management module deployed on the management node; wherein: The proxy module is responsible for obtaining network delay information between all nodes; The node management module is responsible for monitoring the survival status and network delay information of all nodes; The cluster management module is responsible for grouping multiple nodes whose network delays are less than a preset threshold into the same autonomous region based on the acquired network delay information. For any two autonomous regions, the two nodes with the lowest network delays in these two autonomous regions are used as seed nodes. The module also monitors the connection patterns between autonomous regions and determines the connection windows between the autonomous region where the management node is located and other autonomous regions. The topology management module is responsible for establishing synchronization paths between the management node, seed node, and neighboring nodes in the same autonomous area of ​​the seed node; The state management module is responsible for enabling seed nodes within each autonomous region to obtain the state information of their neighbor nodes. Within the connection window mentioned above, the seed node, as a working node, transmits all state information of itself and its neighbor nodes to the management node along the synchronization path to complete cluster state synchronization, and outputs all obtained state information through the management node.

Images