8 results about "Cluster Node" patented technology

The embodiment of the invention discloses a method and device for clearing a mirror image in a cluster node, and is applied to the cluster node. The method comprises the following steps: when a target application deployed on the cluster node of a server is upgraded, recording the name of the newly downloaded mirror image used for upgrading, the current latest version information of the mirror image, and the version information of the previous old version mirror image of the current latest version; when the task of clearing the mirror image is executed, keeping the current latest version mirror image and the previous old version mirror image of the current latest version in all the found locally stored mirror images according to the currently recorded name of the mirror image, the latest version information, and the version information of the previous old version mirror image of the current latest version, and clearing the mirror images of the other versions. By using the method provided by the embodiment of the invention, the probability of deleting by accident the mirror image probably reused in the future is reduced.

The invention discloses a container arrangement method and system for a CPU heterogeneous cluster in a cloud edge environment and a medium, and the method comprises the steps of screening cluster nodes according to a CPU architecture supported by a container for a scheduled container unit, and carrying out the sorting according to the CPU performance of the cluster nodes; based on the sorted cluster nodes, performing a scheduling process irrelevant to a CPU architecture, and determining a corresponding target node; updating a container mirror name in the configuration of the scheduled container unit according to the CPU architecture of the target node, and updating the CPU resource configuration in the configuration of the scheduled container unit according to the CPU performance of the target node; and performing node binding on the scheduled container unit and the target node. According to the invention, the container mirror image matched with the node CPU architecture can be correctly pulled in the deployment process of the deployment container, the same container has similar performance when being deployed on different nodes, and the node CPU architecture and the transparent performance are realized when the container is deployed under a heterogeneous cluster.

The embodiment of the invention discloses a distributed storage system file lock characteristic test method and device. According to the method, an automated test framework can be pre-constructed. Theautomated test framework comprises an automated test node and a task machine pool. At least a task queue is arranged on the automated test node. The task machine pool includes a plurality of task machines. The target task in the task queue can be automatically read by the automated test node. The target task machines are determined from the task pool according to an automated test configuration file corresponding to the target task, wherein the target task machines include a plurality of task machines. The target test script is obtained from the target task machines and the target test scriptis executed so as to complete the distributed storage system file lock characteristic test task. The automated test framework is automatically constructed and then the automated test node automatically tests the file lock characteristics of the distributed storage system so as to improve the test efficiency, avoid missing the test of certain cluster nodes and improve the test quality.

The invention relates to a high-precision AOA clustering and locating method. According to the method, a clustering algorithm is used for clustering node groups, a cluster head is selected from clusters, a newly-added node automatically selects a nearest cluster to join, and angle measurement is conducted in each node in the signal emitting direction; after the last step is completed, the distance between every two nodes is calculated according to an AOA ranging algorithm, then relative coordinates are obtained and sent to a total cluster head from each cluster head to be gathered, and true coordinates are worked out. By means of the high-precision AOA clustering and locating method, each node of the clustering algorithm has self-detecting performance, and the coordinates of each node relative to the corresponding cluster head are worked out though a coordinate algorithm; clustering is only conducted once at the beginning, when no cluster head exists in each cluster or energy is low in each cluster, the cluster head is selected again, information is sent to the total cluster later, and a range-based clustering and locating algorithm is achieved. According to experimental results, the algorithm maintains the advantages of fast deployment, high energy efficiency and high precision, and the algorithm is suitable for wireless networks which are compactly deployed.

The invention provides a cluster capacity management method and system. The cluster capacity management method comprises the steps: acquiring a pod scheduling state of a kubernetes cluster; determining a preset high-consumption resource pod, and pressurizing the kubernetes cluster on the basis of the preset high-consumption resource pod; if it is judged that a plurality of pending pods exist in the first preset period range, obtaining a new virtual machine node from a node resource pool; and loading the plurality of pending pods to the new virtual machine node until the kubernetes cluster does not have the pending pods within the preset operation time. According to the method, dynamic capacity expansion and capacity shrinkage are realized through the cluster nodes, the service tide phenomenon is effectively dealt with, and the node resource occupation state of the cluster is accurately identified according to the state and the number of the pending pods.

The present invention provides a method for optimization of the configuration of a hierarchical network having a first hierarchical level (LEVEL(j-1)), a second hierarchical level (LEVEL(j) ), and a third hierarchical level (LEVEL(j+1)) immediately below the second hierarchical level, each hierarchical level including nodes (RNC, HUB, RBS) at a level different from the lowest level (eg LEVEL(j+1) Each node in a hierarchical level of 1)) (eg, LEVEL(j-1), LEVEL(j)) receives traffic from a cluster of nodes (HUB, RBS) at the hierarchical level immediately below. The method involves first determining the initial arrangement of nodes (RNC, HUB, RBS) in each hierarchical level different from the highest level, and in each hierarchical level different from the highest level (eg LEVEL(j-1)), Each initial cluster of nodes (RNC, HUB) at that level is assigned to the immediately higher hierarchical level of nodes (RNC, HUB), and optimization iterations are performed until they do not result in a node ( RNC, HUB, RBS) number, arrangement and distribution change.