305 results about "Low resource" patented technology

The invention discloses a computing power scheduling method and system based on dynamic load prediction and resource priority ranking. The computing power scheduling method comprises the following steps: collecting historical load data, task submission data and resource state data of each node in a computing power cluster; on the basis of the preprocessed multi-dimensional load feature data set, constructing an improved hybrid prediction model, optimizing model parameters through training, and predicting the load change trend of each computing power node in a future preset time period by using the trained model to obtain a node load prediction result; extracting a service level protocol parameter, a resource demand type and historical execution efficiency data of a to-be-scheduled task, and establishing a multi-dimensional resource priority evaluation index system; according to the computing power scheduling method, the problems of low resource utilization rate and high task response delay caused by low load prediction precision and mismatching of resource allocation and task priority in a traditional computing power scheduling method are solved, and the overall operation efficiency and service quality of a computing power cluster are improved.

The invention relates to the field of distributed machine learning, provides a big data-oriented distributed model training optimization method, and solves the problems of load imbalance, low resource utilization rate, large communication overhead, insufficient fault-tolerant efficiency and the like caused by data fragmentation staticization in the prior art. Load balancing is realized through intelligent clustering and overlapping control; the multi-dimensional heterogeneous resource evaluation model monitors calculation / storage / network indexes in real time, and realizes adaptive scheduling in combination with a task prediction and optimization algorithm; the hierarchical gradient synchronization mechanism adopts a tree-shaped parameter server and a dynamic compression technology, so that the communication traffic is reduced by 50%, and the precision loss is less than 0.8%; the incremental checkpoint system uses erasure code coding and parallel recovery to shorten the fault recovery time from 15 minutes to within 2 minutes, the resource utilization rate reaches 85% or above, the convergence speed is improved by 30%-40% in ResNet, BERT and other model training, and the large-scale training efficiency and the system stability are remarkably optimized.

The invention provides a resource scheduling control method and system for a big data server, and the method comprises the steps: constructing a multi-dimensional resource portrait module, collecting the CPU, memory, network, storage I / O load and task queue length of each node in real time, and predicting a resource demand trend through a time sequence algorithm; extracting characteristics such as calculation intensity, data dependence, memory requirements, network transmission quantity and the like; adjusting the weight coefficients of the resource utilization rate, the task completion time and the energy consumption efficiency according to the system load and the historical effect; establishing a bipartite graph model by taking a resource trend as a node feature and a task vector as an edge feature, and calculating a matching score through graph convolution and a multi-objective optimization function; the scheduling scheme is synchronized by adopting a consistency algorithm; automatic rollback and reallocation are carried out when resources are detected to be insufficient; and optimizing a weight coefficient and a network parameter through reinforcement learning. Through the method, the system resource utilization rate can be improved, the task execution efficiency is improved, the overall scheduling effect stability is improved, and the system fault recovery time is shortened.

The invention relates to the technical field of digital computing and resource scheduling, in particular to a computing power resource intelligent scheduling and load balancing method. The method comprises the following steps: collecting performance indexes of a computing power node on different resource dimensions, and constructing a resource capability vector of the computing power node; extracting demand features of the calculation task on different resource dimensions, and constructing a task feature vector; matching the resource capability vector of the computing power node with the task feature vector to obtain an adaptation degree score; operation indexes of the computing power nodes are collected, comprehensive load scoring is carried out on the computing power nodes, and the current load states of the computing power nodes are obtained; and distributing calculation tasks based on the current load state of the computing power node and the adaptation degree score, and updating the current load state of the computing power node in real time. The problems that the resource utilization rate is low, the task response delay is high and the node load is unbalanced due to the fact that the computing power resources cannot be dynamically matched according to the task characteristics and the computing power node states in a traditional computing power resource scheduling method are solved.

The invention relates to a computing power scheduling method and device based on resource affinity, equipment and a medium, and the method comprises the following steps: collecting the load state data of each computing node through a computing power node monitoring module; then demand analysis is carried out on a task submitted by a user, a task resource demand vector is generated, affinity matching is carried out on the task resource demand vector and node load data, and a resource affinity matching value is obtained; then analyzing the dependency relationship between the nodes, calculating a resource dependency coefficient, and weighting the resource dependency coefficient with the affinity matching value to obtain a comprehensive affinity value; sorting the comprehensive affinity values in a descending order to form a node priority sequence; and finally, computing power mapping scheduling is carried out based on the sequence, and a final scheduling scheme is generated. According to the method, the resource matching degree and the node cooperation relation are comprehensively considered, the scheduling efficiency and the resource utilization rate are improved, and the technical problems that the resource utilization rate is low, task response delay is high and loads among the nodes are uneven due to a traditional scheduling method are solved.

The invention discloses a dynamic batch processing method and device for reasoning requests, electronic equipment and a storage medium, relates to the technical field of computers, and aims to learn and obtain an optimal batch processing strategy by utilizing a dynamic batch processing reward algorithm of reinforcement learning, dynamically adjust the batch processing amount of the reasoning requests and improve the efficiency of batch processing of the reasoning requests. The reasoning request processing efficiency and the resource utilization rate are remarkably improved, the training and reasoning progress is accelerated, the resource utilization rate and the reasoning response speed are improved, the real-time performance index is monitored and obtained to dynamically adjust the optimal batch processing strategy in the process that the reasoning node executes the reasoning request, data changes are adapted, and the reasoning efficiency and the reasoning response speed are improved. The method is suitable for reasoning scenes requiring high throughput, low delay and high resource utilization rate, so that the technical problem of low resource utilization rate or response delay increase caused by static batch processing in the face of request quantity fluctuation or data distribution change can be solved, and resource management and scheduling are more efficient and unified.

The embodiment of the invention discloses a distributed computing power resource scheduling method and system, a storage medium and a program product. The method comprises the following steps: decomposing a composite computing task into a plurality of sub-tasks and constructing a dependency graph; obtaining demand resource information of each subtask and constructing a demand resource matrix according to the demand resource information; obtaining performance data of the global computing power node and constructing a global dynamic resource portrait according to the performance data; dividing global computing power nodes into different areas based on the global dynamic resource portrait; based on the demand resource matrix and the performance data, determining an adaptive area of each sub-task from all the areas, and then determining adaptive computing power nodes of the sub-tasks from the adaptive areas of the sub-tasks; and sequentially executing each subtask based on the dependency graph and the adaptive computing power node of each subtask. According to the method, the problems of low cross-data center resource utilization rate, high burst task response delay and poor heterogeneous environment scheduling efficiency can be solved, the overall computing power resource utilization rate is improved, and the task average delay is reduced.

The invention relates to the technical field of network slice resource allocation, in particular to a 6G intelligent load balancing and fault self-healing method based on AI and network slices, which comprises the step of building an AI decision-making layer model architecture, a self-adaptive optimization mechanism, a resource dynamic scheduling mechanism and a strategy execution guarantee architecture. According to the method, the network load trend is predicted in real time through the AI technology, the resource allocation between the slices is dynamically adjusted, and the problem of low resource utilization rate caused by static allocation and periodic adjustment is solved. Meanwhile, an automatic fault detection and recovery mechanism is designed, when a node fault is detected, standby resource takeover can be quickly triggered, session continuity can be kept, the service interruption time is remarkably shortened, and the requirements of a 6G network for high reliability and low time delay are met.

The invention discloses a method and a device for supporting elastic expansion capability by computing power resources of an intelligent computing center. The method comprises the following steps: monitoring a QPS index of a Kubernetes cluster service in real time, and dynamically dividing service state grades after smoothing data through a sliding window; dynamically adjusting a QPS threshold value based on the historical data and the cluster load; the HF service dynamically expands and shrinks the capacity according to the QPS, the MF service keeps a single instance, and the LF service releases the instance and pre-caches resources to the SSD node; the service life cycle is managed through queue state transition, and cache is optimized in combination with an LRU-LFU mixed elimination strategy and a disk pressure trigger mechanism; the device comprises a monitor, a scheduler, a router, a three-level cache system, a fault-tolerant processing unit, a processor and a memory. And a three-level cache system is adopted to accelerate cold start, so that the fault recovery time is effectively shortened, and the problems of high loading delay, low resource utilization rate and cross-node scheduling of the AI large model are solved.

The invention discloses a cross-regional computing power resource collaborative allocation method based on computing power centers, and relates to the technical field of computing power resource scheduling and optimizing.The cross-regional computing power resource collaborative allocation method comprises the steps that real-time load data, historical task execution data and network delay data of all computing power centers are collected, and a regional load feature database is constructed; predicting the load demand of each computing power center in a future time window by using a long short-term memory network model to generate a load prediction value; calculating a resource gap coefficient and a resource margin coefficient of each region according to the load prediction value and the current resource capacity, identifying a resource insufficient region and a resource surplus region, and generating a resource collaborative matching matrix between the regions; and optimizing and solving the cross-regional task allocation scheme by adopting an improved genetic algorithm to generate an optimal resource allocation strategy, and allocating the to-be-processed task to a corresponding computing power center according to the optimal resource allocation strategy. According to the method, the cooperative utilization rate and the distribution efficiency of the computing power resources in the cross-regional scene are effectively improved.

The invention provides a server GPU computing power allocation method and system and a server, belongs to the field of server resource management, and solves the problems of low utilization rate and inflexible decision of traditional allocation resources. The method comprises the steps that a central agent constructs a directed weighted game diagram of tasks and a GPU cluster, and a global allocation strategy is solved based on Nash equilibrium; a local agent distributes tasks to a specific GPU through reinforcement learning, intelligent migration is achieved in combination with a dynamic threshold value and anomaly detection, and a strategy is iteratively optimized through federal learning. The system comprises a central agent, a local agent, a GPU cluster and an experience pool module, and a server carries the system execution method. According to the scheme, through double-layer agent cooperation and multi-algorithm fusion, the task cluster adaptation relation is quantified, the allocation strategy is dynamically adjusted, the resource utilization rate and the task processing efficiency are remarkably improved, and the adaptivity and reliability in a complex scene are enhanced.

The invention provides an internet of things equipment collaborative task scheduling optimization method and system oriented to edge computing, and relates to the technical field of the internet of things, and the method comprises the steps: collecting a task request from an internet of things terminal, and extracting task features; constructing a dynamic resource state table; generating a candidate node set; selecting a target computing node from the candidate node set, and deploying a task execution unit on the target computing node; generating a resource utilization rate index based on the real-time operation index of the node; and generating a task performance index based on task execution feedback, dynamically adjusting a task scheduling rule, or isolating an abnormal computing node from an abnormal task. Through the task scheduling method and device, the technical problems of low resource utilization rate and high response delay caused by lack of global collaboration of task scheduling in the prior art are solved, and efficient task allocation between edge devices is realized by extracting task features and establishing an optimization model, so that the overall scheduling efficiency and the system performance are improved.

The invention discloses an open architecture system for an underwater acoustic signal processor, and belongs to the technical field of underwater acoustic signal processors. The system comprises a hardware resource pool, a standardized algorithm container group, a visual processing link construction module, a containerized deployment and scheduling platform and a data processing and communication plane. According to the system, special hardware resources are pooled, an underwater sound processing algorithm is packaged into a standardized container, a user is allowed to arrange a processing flow in a graphical mode, an intelligent scheduling platform dynamically deploys the algorithm container to matched hardware for execution according to real-time resource conditions, and meanwhile, an efficient and reliable data communication channel is provided. According to the method, the problems that a traditional system is closed, poor in expansibility and low in resource utilization rate are solved, elastic sharing of hardware resources, agile deployment of an algorithm and flexible construction of a flow are achieved, and openness, flexibility and overall efficiency of an underwater acoustic signal processing system are remarkably improved.

The invention discloses a computing power resource dynamic scheduling method and system based on artificial intelligence, relates to the technical field of artificial intelligence, and solves the technical problems that the resource utilization rate of resource data is relatively low and the accuracy and efficiency of a resource dynamic scheduling method are relatively low due to the fact that the change condition of resource use is not considered and predicted in the prior art. Predicted resource data is generated based on task data; according to the predicted resource data and the task queue, generating a capacity expansion and shrinkage resource quantity; generating task priorities based on the task parameters; dynamically adjusting a task queue based on the task waiting duration; and generating a task scheduling scheme according to the capacity expansion and contraction resource quantity and the adjusted task queue, predicting resource data required by tasks in a future period of time in advance, and dynamically adjusting the capacity expansion and contraction opportunity and the expansion and contraction capacity of each piece of resource data in combination with real-time data in the current task queue, so that the utilization rate of the resource data can be maximized, and the task scheduling efficiency is improved. And the accuracy and efficiency of the resource dynamic scheduling method are improved.

The invention discloses a heterogeneous computing power resource dynamic scheduling system and method based on multi-objective optimization, and the system is deployed in a digital ecological platform complex based on multivariate consensus and embedded intelligent management. Comprising a platform access module, a data acquisition and perception module, a multi-target modeling and optimization module, a dynamic scheduling and execution module and a fault-tolerant mechanism module. The system registers as a computing power scheduling service node through an intelligent contract interface, obtains and verifies the compliance of a computing task, collects heterogeneous computing power resource node state data, and generates a scheduling decision by dynamically adjusting the weight of a target function; evaluating task suitability by using an AMCU suitability scoring model, and executing a scheduling AMCU strategy; according to the method, the technical problems of low resource scheduling efficiency and poor fault-tolerant capability in a heterogeneous computing power environment are solved, and the system throughput and the resource utilization rate are improved.

The invention discloses a job migration method and device, a storage medium and a program product, and relates to the technical field of computers, and the method comprises the steps: obtaining multi-dimensional feature data of each arithmetic unit, preprocessing the multi-dimensional feature data into normalized resource state data, determining a to-be-migrated job in combination with a resource demand parameter of a running job, and carrying out the migration of the to-be-migrated job; the limitation that a migration decision is triggered based on a single resource index in the prior art is solved, the resource condition of each operation unit can be comprehensively considered, and new problems caused by insufficient other resources of the migrated target migration calculation unit are avoided. Meanwhile, neighborhood operation unit sets corresponding to different neighborhood radiuses are generated with the source operation unit as the center, the target migration operation units are screened through the preset target function, the traditional mode of indiscriminate traversal or random selection of the target migration operation units is changed, the search range is accurately narrowed, the search time consumption is reduced, the migration delay is reduced, and the search efficiency is improved. The problem of unbalanced resource allocation is effectively relieved, and the overall cluster utilization rate is improved.

The invention belongs to the field of computer resource management, particularly relates to a computer resource dynamic allocation management method based on cloud computing, and aims to solve the problems of locality of a decision view angle and hysteresis of a time dimension in related technologies. The method comprises the following steps: acquiring the resource utilization rate of each virtual machine in a cloud environment, wherein the resource utilization rate comprises a historical resource utilization rate and a real-time resource utilization rate; according to the resource utilization rate, predicting a resource demand of each virtual machine in a future preset time period; determining a resource competition conflict according to the predicted resource demand, and constructing a resource competition graph model; mapping the resource competition graph model into a non-cooperative game model; and determining a scheduling scheme in the cloud environment based on the non-cooperative game model, and executing a resource allocation operation according to the scheduling scheme. According to the method, potential conflicts can be identified and scheduled before resource competition actually occurs, and performance reduction caused by insufficient resources is reduced.

The invention discloses a cloud-side-end collaborative meat intelligent detection system and a cloud-side-end collaborative meat intelligent detection method. The system realizes high precision, low delay and low resource consumption of meat detection through three-level cooperation of the mobile terminal, the edge device and the cloud server and a dynamic task scheduling strategy based on confidence, and is suitable for meat safety supervision in a large-scale and complex environment.

Disclosed in the present application are resource configuration methods, a resource configuration system, an electronic device and a storage medium. A resource configuration method comprises: acquiring configuration requirement information corresponding to a computing power service; on the basis of the configuration requirement information, performing resource configuration on a general-purpose server and a graphics card resource pool to obtain a configuration result, wherein the general-purpose server is obtained on the basis of processor resource configuration, and the graphics card resource pool comprises a plurality of graphics processing units that support a hot-swapping function; and on the basis of the configuration result, providing a first instance for the computing power service, wherein the first instance comprises a processor resource allocated by the general-purpose server and a processor resource allocated by the graphics card resource pool. The present application solves the technical problems in the related art of low resource utilization rate and poor flexibility of the entire unit caused by heterogeneous models being constrained by the fixed configuration of processor resources of the entire unit.

The invention relates to the technical field of meteorological and ocean numerical forecasting, and aims to solve the problems of strong manual dependence, poor timeliness, high error risk and low resource utilization rate in coupling operation of a WRF atmosphere model and an ROMS ocean model. The method comprises the following steps of: storing an initial file into a date catalogue which is dynamically generated according to a YYYMMDD format after the initial file is subjected to integrity verification; the timed task checks nc.lock / nc.tone files every X minutes, a lock starting model is created when no lock exists and an initial file exists, and an ROMS needs to confirm nc.tone of a WRF result directory first; automatic exception processing is carried out when the model is overtime, and after completion of the automatic exception processing, a Done is created, a lock is deleted, resources are released, and computational nodes are dynamically allocated. According to the method, automatic scheduling is achieved through double-file marking, manual operation is shortened to be within one minute, model connection delay is shortened to be within five minutes, the failure rate is reduced, the resource utilization rate is improved, and a reliable scheme is provided for model business operation.

The invention discloses a resource scheduling method and device, a storage medium and electronic equipment, and relates to the technical field of computers, and the method comprises the steps: determining a predicted load state matched with each node according to a load index parameter set matched with each node in a current node set, the load index parameter set comprises a plurality of parameters used for indicating the resource use condition of the node; determining a resource allocation strategy according to a load index parameter set matched with each node under the condition that the predicted load state of the node in the current node set meets a load peak condition; determining a target node set indicated by the resource allocation strategy and a target weight value matched with each node in the target node set; and performing resource allocation on each node according to the target weight value matched with each node. The technical problem of low resource scheduling efficiency in the prior art is solved.

The present application provides a storage space allocation method and a server. The method is applied to a server. Upon reception of a configuration request, the server registers the external storage of a server cluster into the storage space of the server cluster on the basis of the configuration request to obtain a target storage space, thereby realizing capacity expansion of the storage space of the server cluster by using the external storage of the server cluster. Upon reception of a storage space allocation request, the server uses the target storage space to provide a corresponding storage space, so as to store a temporary file generated in the process of realizing a service function. In the method provided by the present application, the external storage of a server cluster is used to realize expansion of storage resources of the server cluster, and the internal storage and external storage of the server cluster can be invoked to provide a storage space for a computing node, thereby effectively supporting the realization of a service function, improving the success rate of task execution, and avoiding task execution failure due to insufficient storage resources.

The invention discloses an idle computing power resource dynamic scheduling method based on node allocation, and relates to the technical field of node allocation. Sequentially executing to-be-scheduled calculation tasks to perform lightweight simulation execution to generate task feature vectors, collecting idle node hardware architecture feature data to generate node hardware feature tags, calculating a matching degree based on a bidirectional adaptability evaluation matrix, allocating main execution nodes, and performing task scheduling; the redundancy is determined by combining the task key level and the overall reputation score of the node pool, the standby redundant nodes are allocated for parallel execution, and the closed-loop scheduling process of the consensus verification window period and the arbitration mechanism is synchronously set, so that compared with the prior art, the accuracy of idle-time computing power scheduling and the credibility of an execution result can be improved, and the scheduling efficiency is improved. Therefore, the problems of low resource utilization rate and high error rate of calculation results caused by disjunction of task and node adaptation and lack of effective verification in the execution process in the existing idle computing power scheduling can be solved.

The invention relates to the technical field of data processing, in particular to a resource adjustment method and device, equipment and a storage medium. The resource adjustment method comprises the following steps: in response to a received resource adjustment request, obtaining available resource amount in a resource pool and a task running state index of a target task, determining a task type of the target task, and in response to the fact that the task type is a first task type, that is, the target task is in a reasoning stage, sending a resource adjustment request; determining a resource adjustment decision based on the task running state index of the target task and a first preset threshold value; in response to the fact that the task type is a second task type, that is, the target task is in a training stage, determining a resource adjustment decision based on a task running state index of the target task, a second preset threshold value, the available resource amount and the reserved resource amount; and the corresponding capacity expansion or capacity reduction operation is executed. According to the method, the technical problems of static resource allocation and low resource utilization rate are solved, and the technical effects of dynamically allocating the resources and improving the resource utilization rate are achieved.

The present disclosure proposes a task processing method, edge computing device, computer device and computer-readable medium, the method comprising: receiving a task sent by a network device; determining the allocation type of the task, and dividing the task into task units according to the allocation type, and placing the task units into a task pool; determining the dispatch priority of each task unit in the task pool according to the type of each task unit in the task pool, the type of the task unit being the allocation type of the task to which the task unit belongs; determining the task unit to be processed according to the dispatch priority; if the current processing capacity of the edge computing device does not meet the preset first requirement, determining the first network element device, and sending the task unit to be processed to the first network element device. When the computing power resources of the edge computing device are insufficient, the first network element device is selected from the surrounding network element group to perform task transfer calculation, which improves the computing power of the edge computing device and reduces the task calculation delay.

The invention relates to the technical field of resource scheduling, in particular to a server resource optimization scheduling method and system. The method comprises the steps that node index data, a historical load data set and environment parameters of a current server within a period of time are collected, and the node index data comprise the node-level CPU utilization rate, the memory occupancy rate, the disk I / O throughput and the network bandwidth utilization rate; based on the node index data, the historical load data set and the environmental parameters, performing parallel hybrid prediction by using a time sequence prediction model and a trend decomposition model to obtain predicted load data; inputting the predicted load data into a preset optimization model for optimization processing, and determining adjustment strategy parameters; and scheduling the scheduling resources of the current server by using the adjustment strategy parameters to determine an optimal scheduling scheme of the server resources. The invention aims to solve the problems that the dynamic response delay of the existing scheduling method is high, the service degradation is easily caused, and the resource allocation efficiency is low.

The invention discloses a multi-dimensional parallel artificial intelligence processor for large model reasoning, relates to the technical field of artificial intelligence processors, and solves the problem that fixed and rigid hardware parallel configuration cannot adapt to a dynamically changing reasoning load. In order to solve the problems of low resource utilization rate, aggravated communication conflicts and unstable overall performance caused by the fact that a current task is not required, the number and the type of required virtual modules are accurately calculated according to the batch size and the sequence length which are changed in real time, and physical calculation units are dynamically combined according to the number and the type of the required virtual modules, so that processor resources can elastically match the actual requirements of the current task; the contradiction that small tasks occupy large resources or large task resources are insufficient under fixed configuration is fundamentally overcome, so that the average utilization rate of the computing unit is improved, and invalid power consumption is reduced.

A management device according to the present invention comprises a control unit. The control unit executes processing that acquires resource information that stipulates remaining amounts of resources needed to process a workpiece for respective processing systems, processing that, when input of a production plan that stipulates at least a type of workpiece to be produced and a quantity for the workpiece to be produced has been received, allocates processing of the relevant quantity of the workpiece to a plurality of processing systems and transmits a processing order that designates the type of workpiece to be processed and the quantity of the workpiece to the processing systems that are to do the processing, and processing that determines whether the resources at the processing systems that are to do the processing are insufficient on the basis of the processing order and the remaining amounts of the resources at the processing systems that are to do the processing and, when it has been determined that the resources are insufficient, outputs a warning that indicates that the resources are insufficient.

The invention discloses a register renaming device, a processor and a method for maintaining a register mapping table, and belongs to the technical field of computers. The register renaming device comprises a mapping table caching module, a control module, a recovery module and a reordering caching ROB module, and the recovery module is configured to rename a renamed register when a speculative path of a branch instruction is wrong and a copy distributed for the branch instruction is covered. And if the number of the program instructions between the program instruction pointed by the retirement pointer of the ROB module and the branch instruction in the ROB module is greater than the number of the program instructions on the speculative path, restoring the modification of the first register mapping table based on the branch instruction according to the copy and the ROB module. According to the register renaming device provided by the invention, unexpected modifications in the first register mapping table can be recovered by combining the copies and the ROB module under the condition that the number of the copies is smaller than the number of the branch instructions, the recovery efficiency is high, and the resource overhead is low.

The invention discloses a resource allocation method, a computer program product and electronic equipment. The method comprises the following steps: receiving an application data operation request, wherein the application data operation request carries a namespace identifier matched with a first application for triggering operation; determining a node list of a node database occupied by the application data of the first application in the first storage database cluster based on the namespace identifier; obtaining a node demand resource load of a node database in the node list in a second processing period; and determining a first node database from the second storage database cluster and migrating the application data of the first application to the first node database under the condition that the first application is determined to meet the resource redistribution condition based on the node demand resource load. The technical problem of low resource utilization rate caused by neglecting the resource use demand change condition of the application is solved.