55 results about "Autoscaling" patented technology

The present invention relates to a system and computer implemented method for cost-aware autoscaling of artificial intelligence workloads using predictive queueing models, designed to achieve proactive and economically optimized scaling of computational resources across cloud and edge environments. The invention introduces a predictive queueing-based technique that anticipates future workload congestion by modeling dynamic task arrivals and service times using a stochastic queueing process. A cost estimation unit computes the total projected operational cost of potential scaling actions by integrating real-time infrastructure pricing data, predicted delay penalties derived from service-level objectives, and estimated energy consumption. A scaling decision unit applies reinforcement learning-based optimization to select the scaling action that minimizes total cost while ensuring compliance with latency and throughput constraints. The system includes a hardware-integrated autoscaling controller device comprising a predictive computation processor, cost-decision processor, and scaling actuation interface configured for real-time execution of predictive and scaling operations.

The invention relates to a resource elastic scaling decision-making method, system and device and a medium. The method comprises the following steps: collecting real-time operation data of a security service node, and performing multi-dimensional security index analysis according to the real-time operation data to obtain a portrait data packet; predicting the security service weight value to obtain a prediction result, performing dynamic error compensation on the prediction result to generate a corrected weight prediction value, and generating a control instruction based on the corrected weight prediction value and the active session state; and when the instruction is a migration instruction, analyzing a session state snapshot of the instruction, calling a preset kernel state locking function to lock a memory session block of a source node, obtaining incremental state change data to generate a migration snapshot packet, and performing block verification injection operation on a target node. According to the method, by integrating multi-dimensional safety index analysis, prediction error compensation and stateful transition verification mechanisms, the accuracy and response efficiency of resource elastic scaling decision making are improved, and the continuity of stateful service transition and the consistency of safety strategies are enhanced.

The invention relates to the technical field of cloud computing resource management, and discloses a storage and computing resource dynamic recombination management method, system, device and medium, which comprises the following steps: constructing a hierarchical management architecture comprising a local control center and a global network control center, constructing a uniform resource topology view of the whole network through a heartbeat mechanism and metadata reporting, automatic registration and state monitoring of calculation and storage resources are realized; multi-target task scheduling is carried out by adopting an improved genetic algorithm fused with a greedy strategy, and on the premise of meeting service quality constraints, localization processing of calculation tasks is preferentially realized to reduce network overhead; an elastic telescoping mechanism based on an ARIMA-LSTM mixed time sequence prediction model is introduced, and through deep learning of historical load data, a load peak value is pre-judged in advance and the scale of a resource pool is dynamically adjusted. According to the method, the global utilization rate of heterogeneous resources is effectively improved, and lossless rapid expansion and dynamic recombination of storage and calculation resources are realized.

The invention provides an intelligent scheduling management system based on a carbon footprint model, and the system achieves the dynamic coupling of calculation task scheduling and the real-time carbon intensity of a power grid through the cooperation of a carbon footprint modeling engine, a dynamic scheduling controller, a resource state sensing layer and a feedback execution module. The problem that a traditional data center resource scheduling system is lack of carbon sensing capacity is effectively solved. By establishing the dynamic association between the power source and the geographic position and fusing the multi-dimensional equipment energy consumption characteristics and the task demand labels, the scheduling decision automatically responds to the fluctuation of the regional renewable energy sources, and the overall carbon footprint is remarkably reduced on the premise of guaranteeing the calculation performance and the service level agreement. Meanwhile, the improved multi-objective optimization algorithm can efficiently search for an optimal balance point in a solution space formed by carbon emission, default rate and economic cost, the resource energy efficiency utilization rate is further improved through an elastic expansion mechanism based on real-time state perception, and finally unification of data center operation environment benefits and economic benefits is achieved.

This invention discloses a method and apparatus for implementing a high-availability cloud firewall cluster, relating to the field of cloud firewall technology. The method includes session-based traffic splitting of data packets using a Session Flow Table (SLB); simplified processing of data streams for ALG-related protocols; and real-time monitoring of the health status and concurrent session count of each firewall node via SLB to achieve dynamic scaling of firewall nodes. This invention supports automatic elastic scaling of cloud firewalls, as well as multi-active high-availability clusters for NAT and ALG scenarios, while avoiding hot session standby between cluster nodes.

This invention discloses a cloud-native method for dynamic resource scheduling and elastic scaling management of mini-programs, specifically relating to the fields of cloud-native and microservice resource scheduling. The method involves: First, collecting multi-dimensional operational metrics in a non-intrusive bypass manner via a sidecar proxy; second, inputting the filtered and filled metrics into a hybrid predictor to predict resource demand and calculate a burst factor; third, designing a two-layer collaborative scheduler that combines horizontal elastic scaling with vertical dynamic adjustment based on reinforcement learning; and fourth, introducing a multi-tenant fair arbitration mechanism based on resource claims, prioritizing the expansion requests of less-used tenants by sorting them in ascending order of claim value; finally, encapsulating the decision into a custom resource object for execution, and performing incremental model training and closed-loop optimization based on the feedback error sequence.

The invention relates to the technical field of financial data intelligence, and discloses a financial personalized management classification system based on digital economy. According to the system, a dynamic mapping network is formed through weaving by constructing a user composite intention profile. And injecting a probe into a network node, performing compliance deconstruction and verification on a response sequence, driving behavior partition structure recombination, and performing re-calibration on the network according to the behavior partition structure recombination. And capturing an intention drift trajectory through a real-time data stream, dynamically and elastically telescoping network link weight, and finally fusing a strategy to complete management classification. According to the system, online driving optimization of data partitions in a compliance state and real-time dynamic shaping of a decision network by intention drift are realized, a closed-loop adaptive mechanism from bottom data to upper rules is formed, and classification accuracy, timeliness and compliance adaptability of financial personalized management in a dynamic environment are improved.

The invention provides a method for optimizing streaming DAG engine task operation based on K3S. The independent namespace and Pod resource group are created for each task, so that strong isolation among the tasks is realized, resource competition and mutual interference among multiple tasks are avoided, and the operation stability of key tasks is guaranteed. A K3S native Pod automatic expansion piece (HPA) is used for managing a resource group, so that the system can dynamically and elastically expand and contract according to a real-time load, the utilization rate of cluster resources is remarkably improved, and service fluctuation is effectively coped with; the operation and maintenance management is carried out by depending on a high availability (HA) mechanism built in the K3S, the operation and maintenance complexity of the cluster is greatly simplified, rapid fault detection and recovery can be realized without depending on external components such as ZooKeeper and the like, and the reliability and maintainability of the system are improved.

The application relates to the technical field of distributed simulation based on containerization deployment and elasticity, and discloses a distributed simulation method and system based on containerization deployment and elasticity, which comprises the following steps: collecting real-time resource monitoring and historical load trend data of a simulation task, and combining simulation calculation complexity parameters and I / O density parameters to perform unified load evaluation, so that fine modeling of the simulation task and resource demand prediction are realized, and the accuracy of task scheduling and distribution is improved; through subtask segmentation based on the minimum executable unit and a standard containerization packaging mechanism, and in cooperation with a container arrangement platform, modular deployment of the simulation task is realized, and then concurrent execution and elastic scheduling in a multi-node environment are supported.

The application provides a Presto scheduling-based elastic scaling method and device. The elastic scaling method comprises the following steps: in the case that a query request needs to be independently run and is a core service, obtaining a target performance index of an initial cluster and target query information of the query request, wherein the target performance index at least comprises memory usage and CPU occupancy of the initial cluster, and the target query information at least comprises metadata information and context information; and performing elastic scaling on the initial cluster according to at least the target performance index and the target query information. The application realizes dynamic arrangement and control of resources, alleviates the queuing problem caused by resource competition, avoids the problem that the number of PODs is uncontrollable caused by the elastic scaling taking only the CPU and memory usage (i.e. the target performance index corresponding to the initial cluster) as the monitoring index, and achieves the purpose of improving the stability and resource utilization of the service.

The invention relates to the technical field of computers, in particular to a Kubernetes container vertical stretching adjustment method and device, and the method comprises the steps: collecting business data, system data and scene data; determining a message priority coefficient according to the message delay requirement, and quantizing and generating a load index according to the message priority coefficient, the request queue length and the push error rate; diagnosing a performance state according to the average push delay, the memory page change rate and the network bandwidth utilization rate; constructing a resource demand index by combining the load index, the performance state and the active connection number, and judging a resource demand state according to the resource demand index; and determining a scaling direction based on the CPU utilization rate, the memory utilization rate and the network bandwidth utilization rate, and determining a CPU adjustment amount and a memory adjustment amount in combination with the resource demand state, the scaling direction and the context switching times of the process in the container. According to the method, the utilization efficiency of cluster resources is maximized, and a higher-level intelligent solution is provided for elastic scaling of cloud native applications.

The application relates to the technical field of application server middleware performance test, and particularly discloses an application server middleware performance test system. The system comprises a business scene intelligent modeling module, an adaptive load generation and scheduling module, a full-stack performance data acquisition and correlation analysis module and a test environment automatic management module. Through the cooperative work of the above modules, the test script is automatically generated from natural language or logs, the load pressure is dynamically adjusted based on real-time feedback, the full-stack performance data correlation analysis and root cause positioning are realized, and the test environment is automatically constructed and elastically expanded, so that the efficiency, authenticity and automation level of the performance test are improved.

The invention relates to the technical field of artificial intelligence security and cloud computing resource management, and discloses an AI reasoning service elastic scaling method and system for an edge cluster, and the method comprises the steps: constructing a dynamic risk file for each hardware computing unit in a computing cluster; distributing the reasoning service request to a target hardware calculation unit of which the risk state is matched with the security demand level; instantiating an independent logic calculation container for each allocated reasoning service request, and destroying the logic calculation container after the reasoning task is executed; and monitoring the resource configuration change of the computing cluster, and updating the risk state of the computing cluster in the dynamic risk file. According to the method, a mixed protection mechanism with logic isolation as a main part and physical reset as an auxiliary part is adopted, and hardware-level state pollution attacks triggered by antagonistic input are effectively defended in a dynamic elastic edge AI reasoning scene on the premise that the elastic scalability is not obviously sacrificed.

The invention provides a distributed identification generation method and system and a distributed service cluster, and relates to the field of distributed cluster processing, the method combines distributed member management based on a Gossip protocol with a Hash ring, and on the premise of ensuring that the generated snowflake identification is globally unique, gradually increased in trend and high in performance, the distributed service cluster is generated. Decentralization and automatic process management of the distributed service cluster on the Worker ID are realized, strong dependence on external coordination service is eliminated, and reliability, expandability and elastic expansion and contraction capability of the distributed service cluster are improved.

The application discloses a multi-agent collaborative resource elastic scaling decision system for intelligent operation and maintenance, a multi-agent collaborative sensing module, which completes multi-dimensional data collection and standardized preprocessing, and provides standardized data input for subsequent processes; an intelligent decision and prediction module, which carries out load prediction and scaling strategy generation based on the preprocessed data; an execution scheduling and feedback optimization module, which executes the scaling strategy, and realizes continuous iteration of the prediction model and the decision model based on the execution effect, forming a closed loop of 'execution-evaluation-optimization'. The multi-agent collaborative resource elastic scaling decision system for intelligent operation and maintenance can reduce operation and maintenance costs while ensuring stable operation of services, and effectively improves the intelligent level of the operation and maintenance system.

The invention discloses a game activity resource dynamic pre-expansion and scheduling method and system, and the method specifically comprises the steps: obtaining activity configuration information in response to an activity creation request, and carrying out the associated collection of multi-modal data; based on the multi-modal data, performing resource demand estimation by using a pre-trained multi-modal resource prediction model to obtain a resource prediction result; inputting the resource prediction result into a reinforcement learning model deployed in a virtual simulation environment, and generating a resource configuration and elastic scaling plan; pushing and auditing the resource configuration and the elastic scaling plan, and performing verification and correction in a virtual simulation environment in response to an auditing instruction to form an executable pre-expansion scheme; in accordance with an executable pre-capacity expansion scheme, a multi-level resource pre-allocation operation is triggered prior to initiation of an activity. According to the method and the device, the accuracy of game activity resource allocation and the timeliness of response are realized, the user experience is effectively improved, meanwhile, the safety risk is reduced, and a powerful guarantee is provided for smooth development of game activities.

The invention discloses a GPU resource optimization method and system based on dynamic prediction and a topological thermodynamic diagram. The method comprises the steps that a server thermodynamic diagram is constructed, and a thermodynamic value of the server thermodynamic diagram is subjected to weighted calculation based on historical task characteristics and the number of node topology connection edges; on the basis of the to-be-scheduled task queue and the historical task data, predicting a future integrated GPU resource demand to generate a resource demand popularity map; based on the resource demand heat map and a dynamic reference value calculated based on a server thermodynamic diagram, GPU resource fragment risks are quantitatively identified; in response to the fragment risk, selecting an optimal optimization strategy; and executing the selected strategy, including dynamically migrating the tasks supporting the check points and / or performing resource scaling on the tasks supporting the elastic scaling, so as to realize lossless reconstruction of GPU resources. Through intelligent closed loop of prediction-decision-execution, resource fragments are actively resolved, and the resource utilization rate and task scheduling efficiency of a large-scale GPU cluster are remarkably improved.

The invention relates to the technical field of electric power computing power scheduling, and discloses an electric power cross-domain computing power elastic scaling scheduling system supporting a creative GPU (Graphics Processing Unit). The system receives a power cross-domain computing power scheduling request, and searches a matched historical computing power use record set from a historical computing power database according to the request. And carrying out load category division on the historical record set to generate historical computing power record subsets of a plurality of load categories. Computing power feature core calculation is performed on the subsets to obtain a plurality of computing power feature core points and associated time intervals; and using the associated time intervals as the time range of a plurality of scheduling analysis layers, performing multi-scale scheduling feature analysis on the computing power monitoring data sequence of the current power cross-domain within the specified time range through the plurality of set scheduling analysis layers, and outputting a plurality of computing power scheduling feature sets. And carrying out cross-scale fusion calculation on the feature sets to obtain a fusion computing power scheduling feature set.

The invention discloses an instance scheduling method and system, a cluster and a computer readable storage medium. Relates to the technical field of computers. In the embodiment of the invention, the cloud management platform runs the first type of instances with fixed specifications. In the implementation of the elastic scaling capacity, the cloud management platform obtains a first operation parameter of a first type of instance for operating a target service. And under the condition that the first operation parameter meets the capacity reduction requirement and meets the instance type adjustment condition, changing the first type of instance into a second type of instance, and operating the target service by the second type of instance. Thus, in the capacity reduction processing, instance type adjustment condition judgment is added to realize instance type change, and the first type of instances with fixed specifications are changed into the second type of instances with adjustable specifications, so that the specifications of the instances can be adjusted based on the actual load of the service, and the resource utilization rate is improved.

Embodiments relate to container resource autoscaling by a control plane. According to an aspect, a computer-implemented method includes receiving and intercepting a request from a software application by a proxy, the request for a service provided by a backend service of one or more control plane components. A processing device of the proxy determines, based on the intercepted request, an amount of resources to be assigned to or updated in the backend service. The processing device causes a control plane scaler coupled to the one or more control plane components to request the determined amount of resources for the backend service. Upon receiving a confirmation that the determined amount of resources is available in the backend service, the processing device forwards from the proxy, the intercepted request to the backend service.

The invention discloses a clouded test upper computer system and a test method. The clouded test upper computer system comprises a cloud upper computer and at least one edge execution unit, the cloud upper computer is deployed at the cloud and comprises a user interaction module, an artificial intelligence module, a central control and task management module and a cloud database, the edge execution unit is deployed at a test site close to a tested object, and the cloud upper computer is connected with the edge execution unit through a network. And the centralized management, distributed execution and result summarization of the test tasks are cooperatively completed. According to the invention, tight coupling of traditional upper computer functions and hardware is broken through, non-real-time and compute-intensive functions are migrated to the cloud, and elastic expansion and centralized management of resources are realized.

The invention discloses a dynamic optimization method and device of computing resources, electronic equipment and a storage medium, and relates to the technical field of computers, comprising the steps of performing hierarchical management on computing nodes to adapt to different load requirements and priorities, and combining real-time service load and node state data to obtain a dynamic optimization result of the computing resources; the load change is pre-judged through the prediction model, the resource quota is dynamically adjusted, and meanwhile, the dynamic migration of the computing node between the resource pools is realized; the technical problems that in an existing elastic scaling method, due to the fact that a single cache pool or a static resource allocation strategy is adopted, load multi-dimensional characteristics and resource state dynamic changes are not fully considered, resource allocation delay is high, the utilization rate is unbalanced, and resources under burst loads are insufficient, and therefore the system performance and the operation cost are affected can be solved. The technical effects of reducing the resource allocation delay, balancing the resource utilization rate, effectively coping with the burst load, improving the system performance and optimizing the operation cost are achieved.

This invention relates to the fields of edge computing and distributed machine learning technology, specifically disclosing a heterogeneous edge large-scale model collaborative training method based on nested manifold alignment. By introducing a nested heterogeneous compatible architecture guided by singular value decomposition, an ordered spectral basis is constructed in the parameter space, enabling in-vehicle intelligent terminals with different computing power to train different spectral slices of the large model separately. This achieves elastic scaling for inference across the entire spectrum of in-vehicle intelligent terminals with a single global collaborative training. Secondly, to address gradient conflicts in multiple tasks, an energy and density-aware parameter-semantic dual-stream spectral aggregation mechanism is designed. This mechanism uses orthogonal spectral projection to decouple complex mixed gradients into independent feature components, eliminating destructive interference at the microscopic level. Simultaneously, it uses singular value energy density as a physical prior to robustly calibrate long-tail noise, breaking down the dual barriers of heterogeneity and task conflicts in in-vehicle intelligent terminals.

The application provides a distributed identity generation method and system and a distributed service cluster, relates to the field of distributed cluster processing, and combines distributed member management based on a Gossip protocol with a hash ring, realizes the decentralization and automatic process management of WorkerID of the distributed service cluster under the premise of guaranteeing that the generated snowflake identity is globally unique, trend-increasing and high-performance, eliminates the strong dependence on external coordination services, and improves the reliability, scalability and elastic scaling capability of the distributed service cluster.

The application provides an implementation method and device for application atomization fusion arrangement based on OPENAPI, relates to the technical field of computer application, and covers service disintegration and interface definition, service registration and discovery, arrangement and adaptive execution, data processing and mapping, security control, integrated deployment and operation and maintenance steps; the application splits complex business into atomized services, defines interfaces according to the OpenAPI specification, facilitates service management, reduces integration cost, and improves system flexibility and expansibility; load balancing is realized by constructing a service registration center and combining a weighted round robin algorithm to guarantee efficient and stable service calling; API gateways, containerization technology and automatic operation and maintenance systems are used to realize rapid service deployment, elastic scaling and intelligent operation and maintenance; meanwhile, data is processed through semantic analysis and format conversion, and security is guaranteed through multi-level identity authentication and encryption.

The present application relates to the technical field of computers, and discloses an instance scheduling method, a system, a cluster, and a computer-readable storage medium. In embodiments of the present application, a cloud management platform runs a first-type instance having a fixed specification. During implementation of elastic scaling, the cloud management platform obtains a first running parameter for running the first-type instance of a target service. When the first running parameter satisfies a scaling-down requirement and satisfies an instance type adjustment condition, the first-type instance is changed to a second-type instance, and the target service is run by the second-type instance. In this way, in a scaling-down process, the instance type adjustment condition is added to determine and implement an instance type change, and the first-type instance having a fixed specification is changed to the second-type instance having an adjustable specification, so that the specification of the instance can be adjusted on the basis of an actual service load, thereby improving the resource utilization rate.

An embodiment analyzes an inference request to determine a set of parameters of execution corresponding to the inference request. For a Large Language Model (LLM), a first amount of a computing resource is computed, that amount of computing resource being estimated to be needed to produce a set of intermediate results while processing the inference request by executing the LLM using a set of multi-instance Graphical Processing Units (GPUs) (MIGs), a MIG in the set of MIGs comprising a set of slices of a corresponding GPU (set of MIG slices). A set of instructions is sent to a controller associated with the MIG, to cause the controller to modify a second amount of the computing resource available to a MIG slice in the set of MIG slices. The inference request is scheduled to execute using the first amount of computing resource at the MIG slice.

The invention discloses a data synchronization method based on dynamic adaptation of AST (Accelerated State Transfer) drive and Kubernetes. The method comprises the following steps: acquiring an application code of a to-be-synchronized service and performing grammar analysis to generate an abstract syntax tree AST; identifying a write operation point interacting with a target database based on AST, inserting batch processing logic into at least part of single write on the premise of not changing existing business semantics, realizing parameter aggregation and batch submission, and returning to single execution when constraint verification is not passed; a synchronization task is operated in a Kubernetes environment, an elastic scaling event is monitored, and access parameters such as a database connection pool and the like are dynamically calculated and subjected to hot update under the non-stop condition according to the number change of instances; and executing writing after rewriting and outputting a structured synchronization result. Through the method, the synchronization throughput is improved, and the stability in a scaling scene is enhanced.

Techniques are disclosed for automatically adjusting the number of application instances in a distributed software-defined network based on dynamic resource usage. A distributed resource service (dRS) monitors usage metrics such as CPU utilization, memory usage, or request rate associated with fxDeviceApp and fxCloudApp components of a distributed application. When usage exceeds or falls below defined thresholds, the dRS initiates deployment or removal of execution environments—such as containers or virtual machines—hosting the respective application components. Secure communication tunnels are established between the newly deployed instances and other components using a virtual messaging fabric, and routing data in a flow information base is updated accordingly. The system supports zone-based and component-specific scaling, policy-governed traffic assignment through switch controllers, and enforcement of security and access control through cryptographically signed identities and validated certificates. These capabilities enable adaptive, policy-driven autoscaling of applications across distributed network and cloud infrastructure.

The application discloses a kind of computing resource scheduling methods based on user's demand and task priority, it is related to resource scheduling technical field, including, receiving the computing task request submitted by user, explicit demand parameter and implicit demand parameter are parsed and verified, and standardization demand description object is generated;Real-time acquisition cluster state data and external environment parameter, construct user-task-environment three-dimensional feature tensor, output standardization feature vector group;The performance data flow of container instance group is collected, and the pre-trained LSTM prediction model is triggered to trigger elastic scaling decision, and dynamically adjusts cluster resource configuration and executes abnormal task rescheduling.The application constructs user-task-environment three-dimensional feature tensor, and generates dynamic mixed weighted priority score in combination with reinforcement learning model, realizes the space alignment of multidimensional feature and time sequence cumulative effect fusion.