55 results about "Gang scheduling" patented technology

The invention discloses a multi-tenant management method and system based on containers and Kubernetes, and relates to the technical field of computer resource management.Tasks of tenants are packaged into a plurality of container instances, the container instances are submitted to a Kubernetes cluster for calculation, each container instance is subjected to Gang scheduling in the Kubernetes cluster, GPU resources of an AI platform are subjected to critical analysis, and when a critical edge is reached, the Kubernetes cluster is subjected to Gang scheduling in the Kubernetes cluster. All the tenants are classified according to the payment conditions of the tenants, scheduling priority ranking is carried out on all categories in combination with cluster center values, and dynamic allocation and management strategy generation are carried out on GPU resources of each tenant in each category by using a dynamic scheduling algorithm. A distributed deployment architecture is adopted, computing power resource management is performed through a containerization technology and a Kubernetes cluster, GPU resource allocation is performed by using a dynamic scheduling algorithm, and the management efficiency of GPU resources is effectively improved.

The invention discloses a metro crew scheduling plan compilation optimization method and system based on an algorithm, relates to the technical field of metro scheduling, and is used for solving the problem that a conventional metro crew scheduling system is not timely in peak allocation. In the prior art, traffic roads are directly matched with aircrafts one by one, but the adaptability between the traffic roads and the aircrafts is not fully considered, and optimization is not carried out aiming at special requirements in peak hours. In the peak period, the demand is far greater than that in the low peak period, more aircraft shifts are possibly needed for support, and in the peak period, the important demand is that the route and the aircraft shifts with high matching degree need to be preferentially matched. According to the method, through matching optimization of the route and the locomotive crew and combination of historical passenger flow data, efficient subway team scheduling is achieved, and it is ensured that the shift demand in the peak period is preferentially met.

The invention relates to the field of industrial park scheduling optimization, in particular to an industrial park multi-time scale unit scheduling optimization method and system considering source-load-storage matching, and the method comprises the steps: employing a pre-constructed day-ahead pre-scheduling model to initialize a scheduling plan of an energy supply unit on that day; a pre-built intra-day rolling correction scheduling model is adopted to correct the scheduling plan every first preset time, meanwhile, a pre-built real-time feedback scheduling model is adopted to correct the output of the energy supply unit in operation every second preset time, and the time length of the first preset time is larger than that of the second preset time. According to the method, the day-ahead pre-scheduling model is firstly adopted to roughly plan scheduling, the intra-day rolling correction scheduling model is adopted to perform scheduling optimization in the specific scheduling process, and the real-time feedback scheduling model is adopted to perform real-time correction, so that finer adjustment is realized, and real-time adjustment and control can be performed according to the operation condition of the industrial park.

The invention relates to the field of container and virtual machine mixed scheduling, in particular to a container and virtual machine mixed scheduling method and device based on a domestic AI server, and the method comprises the steps: judging whether the server is used for operating a virtual machine or a container, if the server is used for operating the virtual machine, operating the virtual machine and executing corresponding operation, and if the server is used for operating the container, executing corresponding operation; if yes, the container is operated, corresponding operation is executed, a Volcano scheduling framework is integrated, and a Gang scheduling strategy and a BinPack scheduling strategy are defined; and a resource pooling management and dynamic adjustment mechanism is adopted, so that the resource utilization rate and the task execution efficiency are improved. According to the method, the container and the virtual machine can share and efficiently utilize the Ascend-910B resources, collaborative operation of multiple types of tasks is achieved, deep fusion and efficient collaboration of the container, the virtual machine and the domestic AI resources are achieved, and the scheduling intelligence level and the resource utilization efficiency of the domestic AI infrastructure are improved.

The invention discloses an electric power system optimization scheduling method and system based on sea-land wind power cooperation, terminal equipment and a storage medium, and belongs to the field of electric power system scheduling. A final lag period is accurately determined through alignment calculation of the minimum path accumulation distance, an association rule base is constructed, the sea-land wind power association space-time dynamic characteristics are effectively captured, and the problem that the sea-land wind power space-time dynamic characteristics are ignored in the prior art is solved; generating joint output data according to a rule base and current data, inputting a model to generate a prediction scene, and clustering to obtain a typical scene set, thereby improving prediction accuracy; based on a typical scene set, a system operation cost minimization objective function and constraint condition solving are combined, and a unit scheduling strategy of each scene is determined, so that the strategy is more accurate, and the problem of inaccurate scheduling strategy caused by neglecting of space-time dynamic characteristics associated with sea-land wind power in the prior art is solved.

The invention belongs to the field of electric power automation, and discloses a calculation task weaving method and a related device, which are based on a data association graph, acquire a target node corresponding to a calculation target, find out a calculation task node associated with the target node, and construct a task directed acyclic graph according to the calculation task node, thereby realizing automatic construction of a workflow. The dependency relationship between tasks and resource requirements are fully represented, and local optimum and resource mismatching are avoided. And then obtaining a task key path and associated node floating time, allocating resource priorities to associated nodes based on the key path, quantifying the floating time to guide resource allocation, realizing efficient resource utilization and risk avoidance, and ensuring a key calculation task. And meanwhile, the Pareto optimal leading edge of the associated node is generated during multi-target optimization scheduling, and a group of scheduling schemes are calculated through balancing among multiple targets, so that the flexibility and adaptability of services are greatly improved, different weaving schemes can be output according to different optimization requirements, and efficient scheduling of complex workflows is realized.

An apparatus and method for efficiently scheduling kernels for execution in a computing system. In various implementations, a computing system includes a cache and a processing circuit with multiple compute circuits and a scheduler. The scheduler groups kernels into scheduling groups where each scheduling group includes particular kernels of the multiple kernels that access a same data set different from a data set of another scheduling group. Each of these scheduling groups is referred to as a “cohort.” The scheduler accesses completion time estimates of kernels of the cohorts. Using the completion time estimates, the number of kernels currently executing, and the number of remaining kernels that have not yet begun execution of each currently scheduled cohort, the scheduler determines whether to immediately schedule a next cohort or delay scheduling the next cohort. By doing so, the scheduler balances throughput and cache contention.

The invention provides a unit scheduling method and device considering source load uncertainty, equipment and a medium. The method comprises the following steps: determining a unit scheduling strategy meeting a first constraint condition set and a fixed cost value corresponding to the unit scheduling strategy in a deterministic output scene by taking minimization of scheduling cost as a target; the unit scheduling strategy is adjusted with the goal of minimizing the adjustment cost, it is ensured that the number of scenes meeting the second constraint condition set in all the uncertainty output scenes is maximized, and the adjustment cost value of the unit scheduling strategy is calculated based on the adjustment variable of the unit scheduling strategy; and when the unit scheduling strategy does not meet the preset cost requirement, adding a new first constraint condition to the first constraint condition set according to the adjustment cost value and the unit scheduling strategy so as to continue to determine a new unit adjustment strategy. According to the method, the unit scheduling strategy can be determined through two-stage unit optimization scheduling considering new energy uncertainty, and the safety and robustness of the unit optimization scheduling strategy are ensured.

A GPU cluster-oriented deep learning training task scheduling system comprises an offline performance characterization module and an online task scheduling deployment module. The offline performance characterization module respectively predicts the similarity of different training tasks to determine the hardware index set to be collected for a newly submitted task and the performance degradation degree, i.e., the performance interference degree, generated after mixed deployment of two training tasks, and collects the calculation time of the training task under different batch sizes in the exclusive GPU mode to fit the batch size-calculation time relationship. The online task scheduling deployment module takes out a task to be scheduled from a task queue and generates the GPU affinity and batch size distribution strategy of the task to be scheduled according to the batch size-calculation time relationship and the performance interference degree, and divides the training task submitted by a user into appropriate GPUs according to the GPU number specified by the user, i.e., the GPUs with the minimum performance interference after mixed deployment, to minimize the average job completion time and improve the cluster resource utilization. The problem of maximizing the overall system throughput and improving the cluster resource utilization under the condition that part of the training tasks cannot be fully utilized by the GPU resources alone and need to comply with the group scheduling principle in the GPU cluster is completely solved.

The invention discloses a rail transit vehicle multi-target scheduling method based on a V2X architecture. The method comprises the following steps: S1, obtaining a vehicle-road-cloud cooperative state; s2, scene passenger flow prediction based on multi-source data; s3, global multi-target scheduling optimization facing the prediction demand; s4, grouping and ungrouping scheduling based on prediction and real-time passenger flow; s5, operation feedback and adaptive parameter adjustment are carried out; based on a self-adaptive adjustment mechanism of an operation evaluation index, a prediction model structure or hyper-parameters in S2, a target weight in S3 and a passenger flow threshold value, a congestion degree threshold value and a safe vehicle distance coefficient in S4 are adjusted according to different deviation modes, so that all links of prediction-scheduling-marshalling are cooperatively converged in the long-term operation process, and the prediction-scheduling-marshalling efficiency is improved. And systematic performance degradation of a traditional fixed parameter model after the passenger flow structure changes is avoided. According to the method, the scheduling refinement level is improved, scene modeling of passenger flow can be realized, and the passenger flow prediction accuracy in a complex scene is improved; the method has robustness and adaptability in long-term operation.

The invention relates to the technical field of unit scheduling of a gravity energy storage system, in particular to a unit scheduling optimization method, system and equipment of the gravity energy storage system and a medium, and the method comprises the steps: obtaining historical scheduling demand data of a power grid and real-time capacity parameters of each unit in the gravity energy storage system; based on the historical scheduling demand data and the capacity parameters, solving a unit combination scheme meeting the power grid demand through a combination optimization algorithm; layering the units according to the capacity, selecting the unit combination scheme with the highest proportion of the middle capacity layer as a first target scheduling instruction, and outputting the first target scheduling instruction to the execution end. The method has the advantages that intelligent decision support can be provided for power grid dispatching, the system operation cost is reduced, the dispatching flexibility and reliability are improved, the overall performance of an energy storage system is improved, resource waste is reduced, and the power grid stability is improved.

The invention discloses a cross-border import supply chain intelligent scheduling system and method based on big data analysis, and the method comprises the following steps: forming an original scheduling data set through a multi-source data fusion collection module; performing structured processing on the original scheduling data set, extracting market emotion features by adopting an emotion analysis algorithm based on an ELCTRA model, constructing scheduling feature vectors, and generating a scheduling feature vector set; inputting the scheduling feature vector set into a demand prediction model, and fusing and outputting a demand prediction result; generating a plurality of groups of scheduling candidate solutions by adopting an improved dominant Pareto evolutionary algorithm, and selecting a scheduling execution scheme according to scheduling target preference; in the scheduling execution process, execution data is collected through the Internet of Things equipment, execution state monitoring and three-dimensional visualization of a scheduling result are carried out, and risk early warning information is pushed to a high-risk scheduling event. According to the invention, intelligent closed-loop optimization of cross-border import supply chain scheduling is realized, and the accuracy and efficiency of demand prediction and risk response are improved.

The present application provides a method and system for short-term dispatch of units under the advance water storage at the end of the flood season in a reservoir. The method includes the following steps: establishing a scheduling plan for advance water storage in a reservoir; based on the set scheduling plan for advance water storage in a reservoir, constructing a short-term dispatch model for the unit with the objective function of maximizing the total power generation of the reservoir during the planning period and satisfying the model operation constraints; using the reservoir water storage at the end of the period as the decision variable of the short-term dispatch model for the unit, and using a genetic algorithm to seek a short-term dispatch strategy for the unit under the optimal power generation criterion. The method for short-term dispatch of units under the advance water storage at the end of the flood season provided in the present application is based on the framework of "advance water storage-unit dispatching-bionic optimization", constructs and efficiently solves a general model for short-term dispatch of units under the water-based electricity mode of a reservoir, and proposes a method for short-term dispatch of units under advance water storage at the end of the flood season in a reservoir, providing a theoretical basis and technical support for the scientific formulation of short-term dispatch plans for units under advance water storage at the end of the flood season in a river basin.

This invention discloses a method, system, terminal, and medium for optimizing the production scheduling of an aluminum alloy aerospace component creep forming production line. The method includes: constructing a multi-dimensional objective function for optimizing the scheduling of the aluminum alloy aerospace component creep forming production line; generating a set of reference points; and initializing a parent population P of size N. t Using genetic operators on P t The operation yields a child population Q of size N. t P t and Q t The mixture yields a population U of size 2N. t From U t N individuals are selected. The selection process involves three stages: first, clustering-based selection; second, non-dominated sorting-based selection; and third, vector angle-based selection. This process is repeated until the iteration termination condition is met. A set of scheduling scheme solutions is obtained, and a fuzzy decision method is applied to select a high-quality production scheduling scheme. This three-stage selection method improves the ability to distinguish non-dominated solutions, strengthens the balance between population convergence and diversity, and yields high-quality production scheduling schemes.

The invention provides a scheduling information transmission method and device, a storage medium and a program product, relates to the technical field of communication, and is used for improving the data or signal transmission performance of a communication node. The method comprises the following steps: acquiring K groups of scheduling information; wherein at least one group of scheduling information in the K groups of scheduling information comprises one or more modulation and coding schemes (MCS), and K is an integer greater than 1; generating a first signaling according to the K groups of scheduling information; and sending the first signaling.

The invention discloses a hardware event group scheduling method for a computer processor performance monitoring unit. The method comprises the following steps: S1, acquiring an event group set and the number of counters; s2, performing descending sorting on the event groups according to the number of events by using a non-incremental sorting strategy; s3, initializing an empty round; s4, iteratively distributing the event group to an existing round or creating a new round; and S5, outputting the round set of the existing round or the new round obtained in the step S4 as a scheduling scheme of the event group on a performance monitoring counter, and the method has the advantages that a hardware event group scheduling problem is converted into a one-dimensional boxing problem, and a non-incremental sorting strategy and an iterative greedy distribution strategy are utilized, so that the time complexity of a hardware event scheduling process is remarkably reduced, and the scheduling efficiency is improved. A low-overhead and high-precision performance data observation scheme is constructed, on the premise that accuracy is guaranteed, efficient execution of large-scale performance analysis tasks is achieved, and performance monitoring efficiency and system performance observation precision are improved.

The present invention provides a gas turbine unit scheduling method based on multi-stage dual dynamic programming, belonging to the technical field of power scheduling. The specific scheduling method is as follows: Predict the load situation and new energy power generation in the corresponding power supply area, and construct a multi-stage scheduling model for the power supply area on the same day with the goal of minimizing cost; Solve the multi-stage scheduling model based on dual dynamic programming to obtain the scheduling strategy of the gas turbine unit; Set feedback nodes. When reaching the feedback nodes, obtain the operation information of the gas turbine unit and the fluctuation information of the prediction results within the corresponding time period, and correspondingly adjust the model parameters and cycle length of the multi-stage scheduling model; Resolve the adjusted multi-stage adjustment model again to update the scheduling strategy until the gas turbine unit scheduling on the same day is completed. While using the introduced pre-expanded multi-stage dual dynamic programming algorithm for scheduling planning, the present invention sets a feedback adjustment mechanism, taking into account both the solution speed and accuracy of the gas turbine unit scheduling planning.

This invention discloses a parallel pump group scheduling method and system based on dynamic programming, belonging to the technical field of intelligent pump group scheduling. The method involves data acquisition and preprocessing, discretizing and equivalently treating the variable frequency pumps as n equivalent fixed frequency pumps. Then, single-pump performance curves and pipeline network performance curves are fitted. Based on the single-pump performance curves, the high-efficiency operating range of each pump is determined. Then, using water supply flow demand, water supply head demand, and the high-efficiency operating range of each pump as hard constraints for pump group operation, candidate pump group combinations are selected. The start-up and shutdown costs and usage time of each pump are incorporated as soft constraints into the total cost, constructing a minimum objective function for the total cost. Finally, a dynamic programming algorithm is used to obtain the global optimal solution, generating the optimal pump group scheduling sequence within the scheduling cycle. This invention balances low energy consumption and equipment health, and is applicable to mixed pump group scenarios including fixed-frequency and variable frequency pumps, significantly reducing energy consumption in traditional scheduling and mitigating equipment wear.

The present invention relates to the technical field of intelligent manufacturing workshops, in particular to an optimization method for solving the group scheduling of distributed heterogeneous flow shops, including determining the problem-solving goal of minimizing the product assembly completion time and initializing parameters; initializing the solution; generating probability random numbers, and judging whether to select to perform product and group operations or product and workpiece operations; updating the best solution, and judging whether the termination time is reached. If satisfied, the iteration ends and the current best solution is output. The application of the present invention can effectively solve the group scheduling problem of distributed heterogeneous flow shops with a product assembly stage, and improve the production efficiency of manufacturing enterprises through optimized scheduling.

The invention discloses a blocking flow shop group scheduling rapid evaluation method, relates to the technical field of shop scheduling, and constructs an inter-group and intra-group rapid evaluation method based on hierarchy according to the problem characteristics of blocking scheduling by taking a distributed blocking flow shop with starting time as an example. Comprising the steps that 1, the starting time # imgabs3 #, the completion time # imgabs4 # and the leaving time # imgabs5 # of a workpiece # imgabs1 # in a group # imgabs0 # on a machine # imgabs2 # are calculated; 2, the layer # imgabs8 # of the workpiece # imgabs7 # in the group # imgabs6 # is calculated; 3, the layer # imgabs11 # of the workpiece # imgabs10 # in the group # imgabs9 # is updated, and the offset # imgabs12 # is calculated; 4, the maximum completion time is obtained; wherein the number of the imgabs is 13 #, the number of the imgabs is 14 #, and the number of the imgabs is 15 #. Aiming at the problem of group scheduling of the blocked flow shop, the invention provides a method for quickly evaluating the maximum completion time target value, so that the calculation time is effectively shortened, repeated evaluation is avoided, the calculation cost is reduced, and meanwhile, the scheduling efficiency is remarkably improved.

The invention relates to the technical field of power system operation optimization, and discloses a federated physical drive unit scheduling method considering information barriers and regulation authority, comprising the steps of constructing a plurality of scheduling agent proxy models, each scheduling agent proxy model corresponding to a scheduling subject, the scheduling agent proxy model is a decision proxy model based on a space-time diagram convolutional network; based on a personalized federated learning framework, cooperatively training a plurality of scheduling agent proxy models; and inputting a power grid operation state obtained in real time into the trained scheduling agent proxy model, and directly outputting a cooperative scheduling scheme meeting physical constraints and economical efficiency targets through forward reasoning. The method has the advantages that the scheduling method can effectively cope with multi-province information barriers and can efficiently process complex control permissions, meanwhile, the weak supervised learning strategy fused with physical rules remarkably reduces dependence on large-scale label data, and the generalization ability and decision-making safety of the model under complex working conditions are improved.

The invention provides a thermoelectric unit scheduling method and system considering a coal quality measurement result, and belongs to the technical field of unit scheduling, and the method specifically comprises the steps: taking a time period in which a calorific value variation coefficient does not meet a requirement as a calorific value variation time period, and employing the distribution data of the calorific value variation time period to determine the calorific value variation coefficient; when it is determined that the measurement uniformity coefficient of the calorific value of the pulverized coal of the thermoelectric unit does not meet the requirement, determining matching time periods in different calorific value intervals according to the measurement results of the calorific values in different time periods, and determining matching conditions with different preset scheduling strategies in combination with the measurement results of the calorific values in different matching time periods, the method comprises the steps of obtaining a matching condition of a thermoelectric unit, determining a matching scheduling strategy by utilizing the matching condition, obtaining load requirements of the thermoelectric unit in different time periods, and determining the load scheduling strategy of the thermoelectric unit according to the matching scheduling strategy when load scheduling processing is carried out by utilizing the load requirements in different time periods, so that the operation safety of the unit is improved.

The present application relates to the technical field of smart grid, and discloses a generator set optimization scheduling system and method based on real-time electricity price prediction, the generator set optimization scheduling system based on real-time electricity price prediction comprises: calculating electricity price direction incentive; calculating lower section conductor temperature; calculating induced target output; calculating price-heat misalignment characteristic quantity; obtaining optimized active output and generator set start-stop state through unit optimization model solution; calculating generator set pre-order output; outputting final active instruction and continuous climbing trajectory. The present application calculates line dynamic thermal characteristic parameters based on conductor heat balance equation, generates price-heat misalignment characteristic quantity in combination with coupling power flow squeezing effect, thermal inertia hysteresis effect and climbing asymmetric effect, realizes explicit quantification of unit scheduling risk, embeds price-heat misalignment constraint into rolling optimization model for risk weighted power distribution, can take into account the economy of scheduling process and the safety of power grid operation, and thus helps to improve the stability of power grid operation.

The invention discloses a virtual power plant scheduling method based on a Genethink shark optimization algorithm, and relates to the technical field of virtual power plant scheduling, and the method comprises the steps: obtaining each piece of cost data generated in the scheduling process of a virtual power plant, and carrying out the limitation of an enforceable numerical range of each piece of cost data; substituting the cost data into a scheme optimization algorithm, initializing to generate a group of scheduling schemes, and calculating the minimum operation cost of the virtual power plant under the current group of scheduling schemes; continuously simulating a scheduling scheme corresponding to the minimum operation cost of the current virtual power plant through a scheme optimization algorithm, and exploring other scheduling schemes similar to the current scheduling scheme to obtain a current optimal scheduling scheme; a self-protection mechanism is added in the scheme optimization algorithm to prevent the scheme optimization algorithm from falling into a local optimum condition; the method has the beneficial effects that the optimization algorithm is effectively prevented from falling into local optimum, and meanwhile, complex nonlinear, non-convex and multi-target optimization problems can be solved.

The invention discloses a flow shop group scheduling multi-objective optimization method and system based on a dual-evolution mechanism, and the multi-objective evolutionary algorithm based on the dual-evolution mechanism is used for solving the processing optimization problem of a specific group sequence and a workpiece sequence. A unique coding scheme effectively solves the problems of processing division and speed distribution of a group sequence and a workpiece sequence, a feasible solution can be obtained and a target value can be calculated according to related information in decoding, and through three heuristic methods and a random division strategy, specific knowledge of the problem is fully utilized, population diversity is enhanced, and a high-quality initial population is generated. The solution is divided into two populations according to the dominating relation of the solution, a proper evolutionary strategy is selected by using a penalty-based boundary crossing method, the convergence and diversity of the solution are guaranteed, and the algorithm can effectively solve related problems and has excellent performance in the aspects of optimizing a target value, improving the quality and diversity of the solution and the like.

The invention discloses a multi-target energy-saving distributed flow shop group scheduling method and system, and relates to the technical field of intelligent production scheduling and optimizing.The method comprises the steps that operation parameters of a distributed flow shop group scheduling scene are obtained, constructing a multi-objective mixed integer linear programming model taking minimization of the maximum completion time and the total energy consumption as optimization objectives; generating an initial population containing multiple feasible solutions by adopting multiple initialization strategies, and dividing the initial population into two sub-populations by adopting a Pareto frontier-based classification mechanism; constructing a dual Q learning mechanism, establishing two independent Q tables to respectively optimize two targets, generating a candidate solution in each sub-population by adopting an adaptive iteration greedy algorithm, a Q learning algorithm and a hybrid algorithm of the adaptive iteration greedy algorithm and the Q learning algorithm, screening an optimal solution by adopting non-dominated sorting and crowding distance sorting to update the sub-population, and continuously iterating and updating until a termination condition is met; and combining the solutions of the two sub-populations to form a Pareto frontier solution set, and obtaining an optimal scheduling scheme with a better effect.

The invention provides a shipyard air compressor unit intelligent optimization scheduling method and system, and relates to the technical field of industrial intelligent optimization. The method comprises the following steps: firstly, constructing an optimal scheduling model for the shipyard air compressor unit, and then solving the optimal scheduling model by adopting a dynamic planning method. The method breaks through the limitation of a traditional scheduling mode, constructs an air compressor unit optimization scheduling model and solves the model under the constraint conditions of flow load supply and demand balance and the like, supports reasonable formulation of an air compressor unit scheduling scheme, ensures efficient and stable operation of a shipyard pneumatic system, saves energy, and improves the scheduling efficiency of the shipyard pneumatic system. And the total energy consumption of the air compressor unit is optimal under the condition that the flow load is met.

The invention discloses a GPU computing power grouping scheduling method and system based on large model pre-training, and relates to the technical field of GPU resource management, and the method comprises the steps: obtaining multi-source heterogeneous data pre-trained by a large model, preprocessing the multi-source heterogeneous data, and extracting multi-source heterogeneous feature data; performing exception analysis on the multi-source heterogeneous feature data by using an exception analysis algorithm, and identifying exceptional computing power nodes influencing the GPU cluster; and establishing a packet scheduling model based on the abnormal computing power node, dynamically adjusting the member composition and task mapping of the GPU computing power packet by using the packet scheduling model, and optimizing the global resource scheduling of the GPU cluster. According to the method, the overall resource utilization rate of the GPU cluster is improved by accurately identifying the abnormal computing power nodes and establishing the packet scheduling model, and the task execution delay is effectively reduced and the system throughput is improved based on a dynamically adjusted packet member composition and task mapping mechanism.

The invention provides an off-line task scheduling system and method, electronic equipment and a storage medium, and relates to the artificial intelligence field of distributed storage, automatic analysis, off-line data processing and the like. The system can comprise a scheduling layer and an execution layer, the scheduling layer comprises a group of scheduling nodes, and the execution layer comprises a group of execution nodes; each scheduling node has a peer-to-peer and complete scheduling capability and cooperatively works through a standardized application program interface; the scheduling node is used for determining the to-be-executed task as a target task in response to determining that any to-be-executed task meets the execution condition, determining a target node for executing the target task from the execution nodes, and sending a task execution instruction to the target node; and the execution node is used for completing task processing corresponding to the target task according to the obtained task execution instruction. By applying the scheme of the invention, the continuity, timeliness and the like of task processing can be improved.

The application discloses a processor, a display card, a device, a resource allocation method and device, and belongs to the technical field of computer resource management. The processor comprises a thread group assembling unit, a thread group scheduling execution unit and at least two processing units. The thread group assembling unit is used for assembling tasks into one or more thread groups based on attribute information of the tasks in a task assembling stage, and allocating computing resources to the one or more thread groups in a thread group granularity, and each thread group is allocated with one processing unit. The thread group scheduling execution unit is used for scheduling the one or more thread groups to the corresponding processing units of each thread group according to a unified scheduling rule, and the unified scheduling rule is a scheduling rule shared by different types of tasks. The at least two processing units are used for executing the scheduled thread groups. The allocation of the computing resources in the thread group granularity helps to improve the execution efficiency of the processor.