23 results about "Cluster systems" patented technology

This application discloses a flow control method and related equipment. The method proposed in this application can acquire flow pressure data of at least one functional module in a cluster system; based on the flow pressure data of each functional module, overload detection is performed on the flow of each functional module; when overload is detected in at least one functional module, a target functional module for flow control is identified; based on the flow control reference information and flow control history information of the target functional module, a target flow control policy corresponding to the target functional module is determined; based on the target flow control policy, a target generation rate for the access permission credential information of the target functional module is determined; based on the target generation rate, flow control is performed on the target functional module, thereby improving the reliability of flow control and achieving flexible and precise flow control.

The present application relates to a kind of multi autonomous guide vehicle (AGV) cluster security path tracking control method of resisting false data injection attack, belong to control engineering technical field.For the problem that system stability and safety are threatened by false data injection attack in unreliable network, the present application introduces dynamic event triggering mechanism, only updates control input when necessary, reduces communication burden and energy consumption;Design adaptive state estimator, recover normal system signal from tampered sensor and actuator signal;Adaptive attack compensation mechanism is built, to inhibit the negative influence of attack on system performance.The present application fully considers the application requirement of multi-AGV cluster system in unreliable network environment, by introducing dynamic event triggered control, designing adaptive state estimator and building adaptive attack compensation mechanism, false data injection attack can be effectively resisted, and strong guarantee is provided for multi-AGV cluster system security path tracking control.

Embodiments of the present disclosure provide a graphics processing unit (GPU) cluster system and a server. The system comprises: a computing node and a switching node; a first board where the computing node is located and a second board where the switching node is located are orthogonally arranged in a cabinet, and each GPU of the computing node is interconnected with each switching chip of the switching node respectively by means of an orthogonal OD connector.

Systems and methods for Wi-Fi sensing are provided. A networking device is configured to transmit a sensing trigger message including information about a plurality of fast sounding sensing signals to be transmitted. The networking device receives the plurality of fast sounding sensing signals transmitted by a sensing transmitter in response to the sensing trigger message within a transmission opportunity. The networking device performs a plurality of sensing measurements on the plurality of fast sounding sensing signals wherein the plurality of sensing measurements include measurements of Doppler frequency information associated with motion in a sensing space.

The application discloses a cluster arbitration method, a network device and a system. A first network device in a cluster acquires a master stealing request message of a first node and determines a master node of the first cluster. In this way, the network device in the cluster is used to realize arbitration master selection, and other shared storage spaces need not be additionally arranged in the cluster, so that the cost of the distributed cluster system can be reduced on the basis of ensuring a certain degree of reliability.

The application discloses a data processing method, device, equipment and storage medium, and belongs to the technical field of computers. The method specifically comprises the following steps: obtaining a hardware resource usage and a draft history acceptance rate; wherein the hardware resource usage is determined based on the running condition of a GPU cluster system performing inference of a large language model, and the draft history acceptance rate is determined based on an inference result of the large language model; based on the hardware resource usage and the draft history acceptance rate, an inference acceleration strategy prediction model is used to obtain a speculation parameter adjustment strategy of a speculation decoding algorithm; and based on the speculation parameter adjustment strategy, the inference of the large language model is optimized.

This specification discloses a data synchronization method, apparatus, device, medium, and program product. The method is applied to a distributed cluster system comprising multiple single-machine replicas, each running an independent virtual machine. The method includes: responding to a timed scheduling instruction to control each single-machine replica to start a data loading and parsing task at the beginning of a synchronization time window; the duration of the synchronization time window is less than the scheduling cycle of the timed scheduling instruction; during the execution of the data loading and parsing task, when it is detected that a target single-machine replica has completed the data loading and parsing task and generated data to be deployed, triggering a deployment blocking mechanism for the target single-machine replica; when the end of the synchronization time window is reached, releasing the blocking state of all single-machine replicas under the deployment blocking mechanism, controlling multiple single-machine replicas to perform data deployment operations simultaneously, synchronizing their respective data to be deployed to their respective virtual machines.

This invention discloses a collaborative control method for a solid-state transformer cluster system. The method includes: detecting whether a new SST (Solid State Transformer) is powered on and connected to the common AC bus; if so, first triggering the new SST to broadcast a "Hello" data packet containing its ID and rated capacity to all online SSTs; then triggering the first online SST to receive the data packet to send a "Challenge" data packet containing a random number to the new SST; next, signing the random number using the new SST's pre-stored private key and returning it to the first online SST to receive the "Hello" data packet; then verifying the signature using the SST's pre-stored public key; after successful verification, synchronizing the new SST's ID and rated capacity to all other online SSTs in the cluster. This invention solves the problem in the current collaborative control system of solid-state transformer clusters where, when a new solid-state transformer is connected to the system, the system cannot automatically identify its device identity and other parameters, requiring the entire system to be shut down and undergoing cumbersome software configuration and parameter tuning operations.

This application discloses a model routing method, system, and device based on decoupled model representation, belonging to the field of artificial intelligence technology. The method employs a pre-trained query text encoder to encode user query text into query features, achieving effective representation of the user query text. Furthermore, it introduces prior information and non-prior information at the model representation level to constitute decoupled model representation information, achieving effective representation of large language models. Then, a routing model based on a structured attention mechanism selects a large language model to generate an answer for the input user query text, according to the query features and the decoupled model representation information of each large language model. This enhances the cluster system's responsiveness to different inputs, allowing the model routing process to fully consider the diversity and complexity of inputs. Thus, when processing different types of inputs, it can flexibly adjust routing decisions to accurately select the large language model suitable for processing the current input user query text.

The application discloses a kind of unmanned aerial vehicle cluster multi-task time slot allocation and role switching method, solve the problem of low resource utilization, poor coordination, insufficient synchronization accuracy.Information collector, information discriminator, decision maker, communication coordinator and action executor are abstracted as task roles, and each role is mapped to a specific hardware module or software algorithm.The optimal task role allocation table is generated by improving the genetic algorithm using role reuse rate and task timeliness as the fitness function, enabling single unmanned aerial vehicle to reuse multiple roles.Based on performance quantization parameters, sub-slots are divided and their lengths are determined.≤20 level periodic synchronization is achieved through PTP or enhanced radio synchronization mechanism.The application can achieve multi-task parallelism without expanding the size of unmanned aerial vehicle, the role reuse rate is ≥80%, the task completion efficiency is improved by more than 30%, the synchronization confusion rate is reduced to less than 1%, and is suitable for complex scenarios such as military reconnaissance and environmental monitoring.

Systems and methods are provided for providing an on-demand distributed computing cluster. The system comprises: a memory for storing a set of software instructions, and (ii) one or more processors configured to execute the set of software instructions to: (a) track one or more computing environments at one or more time points by a version controlled file system; (b) reproduce one or more components of a tracked environment for an execution; and (c) receive, via a graphical user interface (GUI), a user input to synchronize the one or more components across a plurality of nodes in a distributed computing cluster.

The application discloses an envelope constraint cooperative filtering method and device for a nonlinear cluster system under communication restriction, comprising the following steps: approximating unknown dynamics of a nonlinear system by using a neural network; for the problem of limited communication bit rate of a cluster topology, designing a double-description coding scheme to cope with possible data packet loss in information interaction; establishing a sufficient condition for ensuring that a filtering error system simultaneously satisfies preset average performance and envelope constraints in probability; converting a filter design problem into a convex optimization problem, and obtaining the filter by solving a set of recursive linear matrix inequalities. The application solves the problem that existing filtering methods are difficult to simultaneously process unknown nonlinear dynamics and guarantee envelope constraint performance under limited communication bit rate, and can be applied to the field of multi-agent cluster cooperative situation awareness and estimation with unknown dynamics and communication constraints.

This invention discloses a highly reliable, highly self-healing, and highly available cluster system and a fault switching and service self-healing method. By introducing an independent arbitration node, an odd-numbered sentinel cluster, a middleware status synchronization proxy, and a dynamic priority adjustment mechanism, it achieves second-level fault switching and seamless business recovery. The highly available cluster system includes: an arbitration server and at least two business servers (the number of business servers is even), with communication connections between the business servers and the arbitration server; the arbitration server is equipped with monitoring components to monitor the operational status of the business servers, including a Redis sentinel instance; the business servers are equipped with business applications, Redis sentinel instances, a virtual IP management module, and a middleware status synchronization proxy module, and business data is synchronized between the business servers.

This invention discloses a cross-domain heterogeneous unmanned swarm target fusion situation assessment method. The method includes constructing a heterogeneous unmanned swarm system and completing task initialization; mapping the environmental state space to a reward function, and using a multi-agent reinforcement learning algorithm to generate search strategies for each node; an unmanned underwater vehicle compressing key spectrum information and its own state information and transmitting it back to a surface unmanned surface vessel (SUV), which generates a spatiotemporal graph tensor; performing feature extraction and message passing on the spatiotemporal graph tensor, with each node updating its own situation awareness vector by aggregating information from neighboring nodes, and obtaining a fused situation vector; using the UAV's situation assessment results and the fused situation vector as state input, and using a deep reinforcement learning strategy to output situation assessment results and formation adjustment actions, dynamically optimizing the topology structure to adapt to dynamic situation assessment requirements. This invention achieves high-precision, all-temporal and spatiotemporal perception of targets in wide-area, complex, and dynamic environments.

Embodiments of the present application provide a GPU cluster system and a server, the system comprises: a computing node and a switching node; a first board card on which the computing node is located and a second board card on which the switching node is located are arranged in an orthogonal manner in a cabinet, and each GPU of the computing node is interconnected with each switching chip of the switching node through an orthogonal OD connector. Therefore, through the embodiments of the present application, the problem that the number of GPU interconnections is limited in the related art, resulting in low GPU cluster computing scale and performance, can be solved, and the effect of improving the number of GPU interconnections and computing performance is achieved.

The disclosure provides a fault processing method, system, device, equipment, medium and program of a cluster system, relates to the technical field of computers, in particular to cloud computing technology, artificial intelligence, big data and large model technology, and includes: acquiring fault sensing data of the AI computing power equipment in the cluster system; performing fault positioning on the AI computing power equipment according to the fault sensing data; in the case where it is determined that the target AI computing power equipment has a target fault, performing fault isolation processing on the target AI computing power equipment; and performing automatic fault self-recovery processing on the target AI computing power equipment after fault isolation. The embodiment of the disclosure can realize the automatic fault detection and processing flow of the cluster system in the AI scenario, and improve the automation, intelligence and processing efficiency of the fault processing of the cluster system in the AI scenario.

A multi-node, multi-container cluster system that generates, aggregates, and manages log files from services and components to be used for audit logs and to debug and perform other serviceability tasks provided by a vendor of the cluster system. Logs are collected from all components of the system and aggregated into a consistent format for user analysis and debugging. Embodiments provide a comprehensive way to parse and index vast numbers of log files that can then be packaged and displayed to a user in a way that facilitates analysis and debugging and / or efficient input to appropriate debugging programs.

A multi-node, multi-container cluster system that generates, aggregates, and manages log files from services and components to be used for audit logs used by an auditor. Logs are collected from all components of the system and aggregated using a central persistent volume for non-critical applications and additional dedicated persistent volumes for critical tasks. Audit logs are formatted to provide a consistent logging format in conformance with relevant corporate and government requirements. The audit logs are transferred to an audit log management system that is configured to securely store the audit logs in accordance with security requirements for availability, protection against tampering, and secure access.

The application discloses a second-order cluster increment encirclement control method and device and equipment, and relates to the field of artificial intelligence, and is used for solving the problems that the number of intelligent agents in a cluster system is invariable, and the efficiency and success rate of task execution are low in the prior art. The method comprises the following steps: a communication topology of an original cluster system is established, and initial state information of the original cluster system is obtained; a second-order communication topology corresponding to a target cluster system obtained by adding an increment intelligent agent set and state information of the intelligent agents are determined based on an intelligent agent increment instruction; control amounts of each intelligent agent in the target cluster system under a current encirclement target are calculated; and the state information of each intelligent agent in the target cluster system under the current encirclement target is updated according to the control amounts, so that the target increment encirclement control is realized. The method overcomes the difficulty that the encirclement range of the cluster system is limited in the encirclement task, expands the task range of the cluster system, and improves the efficiency and success rate of task execution.

The present invention relates to authorization of service requests in a multi-cluster system. A method comprises receiving a dependency chain matrix from each of a plurality of service endpoints in a multi-cluster system, fusing the received dependency chain matrices into a fused data matrix; sharing the fused data matrix as a sidecar module to each of the plurality of service endpoints, receiving a service request for at least one of the plurality of service endpoints by an external user or service, verifying authorization for the service request based on the fused data matrix, and generating a label attached to the service request and passed downstream to the plurality of service endpoints.

The application relates to a network security defense method based on dynamic camouflage, and belongs to the cross technical field of network information confrontation and security protection. The application can construct a dynamically changed camouflage network environment, simulate real network behavior characteristics, utilize a frequency-adjustable address hopping mechanism to improve the confusion of the network, and effectively enhance the resistance of a cluster system to various network attacks.

Systems and methods for predicting future risk for a target entity are provided. A risk assessment system receives historical risk assessment data of the target entity and identifies a target cluster that matches the historical risk assessment data. The target cluster is identified from a group of clusters determined using high dimensional clustering based on risk assessment data of a set of entities. The risk assessment system identifies a set of nearest neighbors of the target cluster and determines a prediction of future risk for the target entity based on the target cluster and the set of nearest neighbors. The risk assessment system transmits a responsive message, which can include the prediction of future risk, to a remote computing device for use in controlling access of the target entity to one or more interactive computing environments.