84results about How to "Reduce response delay" patented technology

The invention belongs to the technical field of e-commerce logistics management, and particularly relates to an e-commerce product storage full-process monitoring management method, which comprises the steps of synchronously acquiring multi-source heterogeneous monitoring data in real time through a sensing equipment network; performing space-time alignment preprocessing on the original data, and converting heterogeneous data into a uniform data format; semantic fusion is carried out on the standardized data, and a job behavior knowledge graph with a complete spatio-temporal context is constructed; inputting the atlas into a pre-training hybrid model, and detecting an inventory abnormal event; and generating a strategy instruction in combination with the real-time order and resource data, carrying out joint analysis on an operator, and outputting a compliance score and a specific violation operation type. Through adoption of the technical scheme, complete reconstruction of the context of the operation behavior can be realized, the anomaly detection accuracy is remarkably improved, the abnormal event response delay is shortened, and the inventory loss rate caused by human errors is reduced.

The invention provides an energy storage optimization scheduling method and system under microgrid priority fusion constraint, and the method comprises the steps: S1, carrying out the reading and preprocessing of multi-source data, carrying out the unified time axis of historical photovoltaic output, user load, future weather, electricity price, demand unit price, and completing the missing value interpolation and abnormal value elimination, and forming a time series data set; s2, setting decision variables, namely charging power, discharging power, maximum charging power, maximum discharging power, energy storage equipment capacity, commercial power purchase unit price and demand unit price; s3, the overall income is maximized under the specified priority, and the expenditure is reduced; s4, establishing a priority constraint, an energy storage hardware constraint, an electrical load constraint and other constraint conditions, and avoiding target scheduling conflicts according to the corresponding constraint conditions; s5, carrying out two-stage MINLP rapid solution, and searching a feasible solution in a short time by using a rapid exploration method; and deep optimization is carried out through a fine mode.

The invention discloses a dual-mode communication method and system based on intelligent decision making. The method comprises the following steps: acquiring historical operation data and actual operation data of HPLC (High Performance Liquid Chromatography) and HRF (High Performance Framework); creating a reinforcement learning model which comprises a state space, an action space and a reward function; the state space is a combined vector of a dual-mode channel quality index, the action space comprises HPLC priority, HRF priority and dual-mode self-adaption, and the reward function is a weighted calculation formula of throughput, power consumption and packet loss rate; and performing iterative training on the reinforcement learning model based on the historical operation data to obtain an intelligent optimization model, inputting the actual operation data into the intelligent optimization model to obtain an optimal action, and adjusting the dual-mode communication parameters according to a corresponding preset parameter adjustment rule. According to the invention, communication mode selection and parameter adjustment are more suitable for actual environment requirements, the communication efficiency is improved, the energy consumption and data loss risk are effectively reduced, the system architecture is simplified, and the deployment cost and response delay are reduced.

The invention discloses a building vector graph multi-scene intelligent simplification method and system, and belongs to the technical field of geographic information systems. Comprising the following steps: establishing a four-level data system comprising an original data layer, an engineering cache layer, a display cache layer and a terminal view layer; monitoring data change and user interaction events through an event bus; the event is responded, the current application scene is intelligently recognized, and the corresponding simplified parameter configuration is dynamically loaded; based on the configuration, executing simplification logic on the target building vector graph; querying a topological relation to carry out cooperative processing on the contours of the adjacent buildings; and synchronously updating a processing result to the four-level data system. Through four-stage data caching, dual-drive triggering and scene self-adaptive judgment, efficient, high-quality and self-adaptive processing of building simplification in multiple scenes such as cartographic integration, real-time display and spatial analysis is realized, and the system flexibility, the resource utilization rate and the user experience are improved.

The application discloses an emoticon package selection and text generation method based on prompt learning and a multi-modal model, and comprises the following steps: S1, intercepting a big model agent and a user's recent round of dialogue, and compressing the dialogue into a dialogue prompt mark; S2, arranging an emoticon package set into a candidate emoticon package input set graph and pairing the candidate emoticon package input set graph into an emoticon package reasoning sample; S3, accessing an improved RT-DETR model to extract and fuse multi-scale visual features; S4, an object query group organization structure is truncated, independently detected, and inquired into an entry, and is grouped into a query group; S5, a Transformer decoder with a query group cross-layer inheritance structure forms a candidate emoticon package to undertake a query; S6, a prediction head reads out a path constraint candidate reply text generation; and S7, after dialogue closure verification, emoticon package images and reply texts are output. The application relates to the technical field of artificial intelligence multi-modal content processing, and realizes emoticon package selection and reply text generation collaborative processing.

The invention discloses a cloud local application non-inductive switching method, system and device and a storage medium, and relates to the technical field of cloud virtualization and intelligent vehicle machine interaction, and the method comprises the steps: loading and uniformly displaying cloud and local applications when a vehicle machine is started; when the network is available, preloading the high-frequency cloud application based on the historical use data, and obtaining the unique serial number of the process of the high-frequency cloud application; when a user calls any cloud application, response data (such as application video streams) are directly obtained from the cloud according to whether the serial number is cached or not. According to the scheme, complete cloud mirror image pre-loading is abandoned, the single application serves as the minimum operation unit, the response time is reduced to be within the driving safety threshold value, and safety is improved; the delay is obviously reduced through intelligent preloading and instance multiplexing; seamless fusion of cloud local experience is realized and interaction splitting is eliminated through unified entrance, adaptive vehicle machine UI and cockpit hardware linkage.

The invention relates to a TSN (Time Sensitive Network) timestamp recovery method based on an FPGA (Field Programmable Gate Array), which belongs to the technical field of TSNs and comprises the following steps of: S1, collecting synchronous messages of different TSN nodes in parallel by multiple ports of the FPGA, and analyzing timestamp data by taking a message serial number as an anchor point; s2, constructing an incomplete timestamp matrix and a mask matrix by adopting a time sliding window mechanism, and automatically marking missing data; s3, a random index set is generated through LFSR, and matrix sparse sampling is achieved through a bitmap judgment unit; and S4, based on an FPGA accelerated CUR matrix decomposition algorithm, solving a pseudo-inverse matrix through parallel MGS-QR decomposition, and reconstructing a complete timestamp matrix through block parallel matrix chain multiplication. The FPGA parallel architecture, ping-pong cache and fixed-point number processing are adopted, microsecond-level low-delay recovery is achieved, hardware resource occupation is greatly reduced, robustness in a severe network environment is improved, and the method is suitable for a high-precision clock synchronization scene of an industrial TSN node.

The invention belongs to the technical field of power equipment safety monitoring and intelligent sensing, and particularly relates to an overhead line bird repelling device and method based on multi-mode sensing dynamic fusion and edge calculation. The system comprises a sensing layer which is used for sensing the activity of birds around the overhead line and the ambient brightness in real time; the edge calculation layer is used for processing the data of the sensing layer by adopting a prediction model, outputting bird behavior prediction, and outputting a corresponding bird repelling strategy by adopting a self-adaptive bird repelling decision model according to the bird behavior prediction; the execution layer is used for executing bird repelling operation according to the bird repelling strategy and the ambient brightness; the power supply layer is used for supplying power. According to the bird repelling system, through combination of the microwave radar unit, the camera unit and the photosensitive sensor, multi-mode sensing dynamic fusion is formed to capture bird activities, a prediction model and a self-adaptive decision model are locally operated in combination with an edge calculation layer, bird behaviors are predicted in advance, response delay is compressed, and the bird repelling effect is improved. The delay problem caused by a traditional scheme depending on the cloud is solved.

The invention relates to the technical field of big data management, in particular to a whole-process performance data management method based on big data. The specific implementation process comprises the steps of obtaining heterogeneous behavior traces, deconstructing and packaging the heterogeneous behavior traces into space-time cross-domain data streams, and inputting the space-time cross-domain data streams into a distributed cooperative monitoring model; an asynchronous pause frequency and a cross-domain retrieval track are extracted, a cooperative friction coefficient is calculated, and a cooperative retardation degree is generated; inputting the collaborative friction coefficient and the collaborative retardation degree into a performance evolution engine, generating a targeted dynamic weight, and outputting a targeted collaborative performance report; and outputting a resource intervention instruction when the cooperative retardation degree change slope exceeds a preset warning threshold. According to the method, real-time unified access of multi-source heterogeneous process data, continuous association modeling of cross-system behavior tracks, detection of flow blocking nodes and online dynamic updating of evaluation weights are realized by utilizing a distributed collaborative monitoring model and a performance evolution engine, so that the association processing delay of the cross-system data is reduced; and the timeliness and accuracy of flow anomaly detection are improved.

The invention discloses a working condition self-adaptive injection molding process energy consumption optimization control device and method, and belongs to the technical field of injection molding process control. The method comprises the following steps: acquiring current quasi-static working condition data, and calculating a theoretical energy consumption reference value and a theoretical melt viscosity fingerprint reference value based on the quasi-static working condition data; in the plasticizing stage of the current injection molding period, a torque signal of a screw driving motor is collected, integral operation is conducted on the torque signal, and melt viscosity fingerprints in the current period are extracted; calculating to obtain a plasticizing power deviation value based on the current-period melt viscosity fingerprint and a theoretical melt viscosity fingerprint reference value; and based on the plasticizing power deviation value and the theoretical energy consumption reference value, a corrected energy consumption target for the next injection molding period is generated, and a corresponding equipment control instruction is output. According to the method, cycle-by-cycle feed-forward correction is performed on the energy consumption target by utilizing the process fluctuation information, so that the problem of failure of the preset energy consumption model caused by non-uniform components of the recycled materials is solved, and the energy-saving precision and the product percent of pass are improved.

The invention is applicable to the technical field of information processing, and provides a design consultation service system based on case similarity comparison, comprising: a case database module for storing multi-dimensional design case data including text description, images, 3D models, user feedback data and metadata tags; according to the method, high-precision matching of multi-modal cases is achieved through the similarity comparison module, the accuracy and efficiency of design consultation services are remarkably improved, an embedding extraction unit of the similarity comparison module adopts a multi-modal embedding model, texts, images and 3D models are uniformly converted into vector representation, cross-modal matching is supported, and the accuracy and efficiency of design consultation services are improved. Limitations of traditional keyword retrieval are overcome, a similarity calculation unit applies a weighted cosine similarity algorithm, different modal information is fused, it is ensured that a matching result meets design requirements, and a dynamic weight adjustment sub-module automatically optimizes weight parameters according to historical preferences of a user and provides a personalized matching strategy. And the cross-domain transfer learning subunit realizes transfer of knowledge in the field of building design to the field of product design through fine tuning of a pre-training model.

The invention discloses an equipment control method and device for smart home equipment, electronic equipment and a medium, and belongs to the technical field of equipment control. Comprising the steps of obtaining one or more pieces of environment data in a process that one or more pieces of home equipment operate according to a first scene mode; generating an environment feature vector; inputting the environment feature vector into a preset scene classifier, and outputting a prediction scene corresponding to the environment feature vector and a prediction probability of the prediction scene; when it is judged that the first scene mode is triggered to be switched into a second scene mode corresponding to the prediction scene according to the prediction scene and the prediction probability, an equipment control sub-module corresponding to the second scene mode is loaded in a local database, and the equipment control sub-module and a cloud training meta-model share the same convolutional backbone structure; performing fine tuning on the equipment control sub-model based on the environment feature vector in the second scene mode; and in the second scene mode, controlling one or more pieces of home equipment by adopting the finely-adjusted equipment control sub-model.

The invention relates to the technical field of natural language processing and database query, in particular to a semantic alignment optimization method and system based on meta-information guidance and a storage medium. Obtaining a keyword vector; obtaining an overall matching confidence coefficient according to the original database description text vector and the keyword vector; judging whether the overall matching confidence is smaller than a threshold value or not; if the overall matching confidence coefficient is smaller than a threshold value, three types of label probability distribution are obtained, and an output label type is obtained; judging the type of the output label type; if the output label is a second type label, generating candidate information, selecting one meta-information item by the user, and updating a synonym mapping table according to the selected meta-information item; and if the output tag is the third-class tag, generating a boundary prompt. According to the method, the two-stage detection architecture based on the lightweight model and the vector index is constructed, and semantic extension fed back by the user is combined, so that low-cost and high-efficiency semantic matching and self-adaptive optimization are realized.

The application provides a kind of abnormal flow discernment method, device, electronic equipment and computer readable storage medium, it is related to network information security technical field.Discernment method includes: identifying abnormal flow and its corresponding IP address, wherein, abnormal flow refers to one or more kinds of abnormal network flow;According to IP address and preset device information library, the port bandwidth data of abnormal device is monitored, wherein, abnormal device refers to one or more devices to which abnormal flow belongs;According to port bandwidth data, normal flow and target abnormal flow in abnormal flow are discerned.It at least solves the core pain points of low accuracy, poor timeliness, difficult to troubleshoot and fragmented collaboration in related technologies, leading to high network security risk, high operation and maintenance cost and large business loss.Adapted to abnormal flow detection and diagnosis scene.

The application provides a radio frequency source circuit and a laser frequency shift lock cavity signal generation method, which can be applied to the field of cold atom optical lattice clock technology. The radio frequency source circuit comprises a control sub-circuit, a phase-locked loop sub-circuit and an output power control sub-circuit. The control sub-circuit is used for receiving a digital control instruction from an upper computer and taking a synchronous clock as a working clock, performing amplitude-phase conversion and phase waveform conversion in parallel based on preset modulation data carried by the digital control instruction, and outputting a phase control word and a phase demodulation signal. The phase-locked loop sub-circuit is used for frequency multiplication of a reference clock and outputting a clock signal suitable for the working frequency of a frequency synthesizer as the working reference clock of the frequency synthesizer. The frequency synthesizer is used for taking the working reference clock as a frequency reference, realizing phase synchronization by combining the synchronous clock, and adjusting the phase of an output signal based on the phase control word. The output power control sub-circuit is used for adjusting the power of the output signal of the frequency synthesizer based on preset power modulation data to output a laser frequency shift lock cavity signal.

The embodiment of the invention provides a protection section state-based cross-interlocking concentration area protection turnout locking method, which comprises the following steps that: an ATS sends a route handling command to a CI1 of a first interlocking concentration area to which a route belongs, the CI1 handles the route and feeds back completion information, and then the ATS sends a locking command aiming at a protection section in which the route is positioned to the CI1, the turnout in the protection section is located in a first interlocking concentration area, and the protection turnout is located in another second interlocking concentration area. The CI1 responds to a locking command to expropriate a turnout and a protection turnout, and sends a protection section state set to be locked to the CI2 of the second interlocking concentration area; the CI2 responds to the state to move the protection turnout to a corresponding protection position and sends a turnout driving mark set to be successful to the CI1; and the CI1 responds to the mark, the turnout is moved to the corresponding position of the route, a locking command is executed, and the turnout and the protective turnout are locked. In this way, the efficient, reliable and safe locking process of the protection turnout crossing the interlocking concentration area can be achieved.

The invention discloses an event identification method and system for pig house intelligent monitoring, a camera and a medium. The method comprises the following steps: collecting and sampling pig house video streams; performing target identification on the sampled image to obtain an identification result including categories and position coordinates of pigs, personnel, vehicles and the like; performing label matching on the identification result based on a pre-configured label rule, wherein the matching rule at least comprises a target appearance rule, a region intersection rule, a target continuous appearance rule and a target disappearance rule; matching a pre-configured event rule according to a label matching result, and generating hog house monitoring events such as entry of pigs into a corridor and entry of personnel into an effective area; and intercepting an event associated video clip, monitoring a 4G network signal state, and adaptively selecting to locally store or report the event data and the video clip to a cloud platform. According to the invention, accurate and flexible event identification aiming at a pig house scene is realized, and the reliability of monitoring data in an environment without a wired network or a weak network is guaranteed.

The invention discloses an edge computing resource scheduling and self-healing method and device based on containerized micro-service. The method comprises the following steps: acquiring operation data in a multi-energy scene through a heterogeneous terminal; when a certain micro-service in the edge computing gateway breaks down, the edge computing gateway based on the containerized micro-service architecture dynamically dispatches operation of the micro-service to realize fault self-healing of the micro-service; when a certain edge computing gateway in the edge computing gateway cluster formed through the negotiation of the local area network breaks down, another healthy gateway is selected to bear the responsibility of the broken-down edge computing gateway; when the edge computing gateway detects that the operation data reach a trigger condition, executing control by adopting a scheduling strategy based on event driving, computing a new output distribution scheme of each energy unit through edge-cloud cooperation, and issuing an adjustment instruction to the corresponding energy equipment by the edge computing gateway according to the output distribution scheme; the system reliability is improved, the real-time response capability is enhanced, and the energy utilization rate is optimized.

This invention discloses a BIM-based hoisting construction supervision and optimization management system, comprising: a perception layer: real-time acquisition of multi-source heterogeneous data; a decision layer: the bottom layer uses an incremental RRT# algorithm to generate candidate paths that satisfy the kinematic constraints of the crane, and the upper layer embeds a distributed Q-learning framework, where each crane acts as an independent intelligent agent, collaboratively learning through a shared experience pool, dynamically optimizing path planning strategies based on environmental perception data and construction progress requirements, and simultaneously developing a fuzzy sliding mode control algorithm combined with an LSTM-Transformer model of crane trolley and hook travel speeds and hook lifting speeds, as well as a wire rope disturbance prediction model to calculate crane motion compensation parameters, and generating optimal control commands in advance based on predicted crane movement and lifting data; an execution layer: the commands generated by the decision layer are issued to the construction equipment; and an optimization layer: a BIM model is constructed, and the equipment execution results and structural safety monitoring data are fed back to the decision layer.

According to the double-fed fan gearbox residual life prediction system, synchronous sensing of multiple physical quantities such as vibration, oil and thermodynamics is achieved through the multi-source heterogeneous data acquisition module, and sensing blind areas of traditional single-point vibration monitoring under complex working conditions on the sea are effectively overcome in combination with space-time correlation analysis of the feature fusion and enhancement module; the dynamic prediction model module adopts a dual-channel deep network architecture, and synchronously processes differentiated features of progressive degradation and sudden faults, so that the fault early warning accuracy is improved, and the response delay is reduced; the decision optimization module shortens the operation and maintenance decision time based on a digital twinning three-dimensional visualization and hierarchical maintenance strategy, and reduces the offshore operation and maintenance cost in cooperation with an environment-adaptive resource scheduling algorithm. The whole set of system forms a'perception-analysis-prediction-decision 'closed loop through an industrial Internet of Things platform, the technical breakthrough of full life cycle health management of the gear box is realized, and the system is particularly suitable for the intelligent operation and maintenance requirements of deep and far sea megawatt wind turbine generator sets.

The invention relates to the technical field of data acquisition and processing, and discloses a multi-level time sequence optimization logic method for FPGA high-frequency data acquisition and processing, which can significantly break through the performance bottleneck of a traditional static hardware architecture in processing non-stationary high-frequency signals, and can improve the performance of the FPGA high-frequency data acquisition and processing by introducing a dynamic value evaluation system based on signal entropy. According to the method, accurate capture and priority scheduling of key mutation data are achieved, occupation of invalid data on computing resources is thoroughly eliminated, and by means of time sequence and resource coupled topology planning and an elastic assembly line reconstruction technology, hardware logic has the self-adaptive capacity of real-time deformation along with data features, and the real-time performance of the hardware logic is improved. Not only are time sequence conflicts and metastable state risks effectively avoided under extreme working conditions, but also self-evolution and calibration of system parameters are realized through a closed-loop feedback mechanism. And the utilization efficiency of logic resources and the robustness of multi-task parallel processing are remarkably improved.

The application discloses a memory optimization method and device based on state classification. The method comprises: monitoring the available memory capacity of a computing device in real time, dividing the memory pressure state into N levels according to a plurality of preset pressure thresholds, wherein N >= 3; obtaining the load type identifier of the currently running application or process, and dividing the load type into at least two priority levels; according to the memory pressure state level and the load priority level, querying a preset scheduling strategy mapping table to obtain a corresponding resource allocation coefficient; and generating a resource scheduling instruction according to the resource allocation coefficient, wherein the instruction comprises: adjusting the resource allocation quota of the current process, adjusting the priority of the preload queue, and releasing the specified memory area. The application realizes dynamic optimization and scheduling of memory resources by combining and classifying the memory state and the load type, can be widely applied to various computing devices and application scenarios, and improves the system running fluency and memory utilization efficiency.

An MULTI-LoRA-oriented large model memory optimization configuration system comprises a memory block manager based on a tree structure and a revenue-based memory block scheduler, a GPU memory and a main memory are uniformly divided into memory blocks with the same size by the memory block manager to fill the whole memory space, and LoRA weight and KV cache are stored according to the granularity of the memory blocks; and the memory block scheduler performs dynamic memory migration on the LoRA weight and the KV cache according to the revenue by continuously evaluating the revenue of reserving the specific memory block in the GPU memory or the main memory. Aiming at a multi-LoRA reasoning scene, the LoRA weight and the KV cache in the GPU memory and the main memory are efficiently managed and scheduled, so that the overall resource utilization efficiency is improved.

This invention discloses a method for implementing AI search in the gaming field, comprising: training a language model, supervising fine-tuning, and optimizing rewards to obtain an optimized language model; an online service module, based on the optimized language model, performing risk control and intent recognition on user-input questions; and, based on the intent recognition results, employing relevant strategic intervention methods to input the question into the optimized language model, obtaining an answer, and replying to the user. It has the following advantages: 1. Low cost: Using a smaller model with high-quality data for training achieves the same effect as a larger model. 2. Fewer illusions: This invention uses a high-quality document when generating answers and simulates adding interfering documents during training, which strengthens the language model's resistance to interference and reduces the illusion of answers. 3. Low response latency.

The invention discloses an HITL method and system for unloading a large model reasoning task in a heterogeneous GPU computing power cloud. The method comprises the following steps: constructing a system model of the heterogeneous GPU computing power cloud; an HITL system architecture based on multiple agents is designed; human expert feedback is introduced through an HITL method, and recognition of a large model on a user intention is corrected; human expert feedback is introduced through an HITL method, and recognition of a large model on a user intention is corrected; the multi-agent cooperation module maps the intention of a user to a utility function through interaction among a plurality of agents; an optimized cue word method is used, the thinking process of a large model is decomposed, and the intention recognition accuracy is improved in a chain reasoning mode. According to the method, dynamic scheduling of computing power resource management and large model reasoning tasks can be optimized, and the service quality and the user experience in the heterogeneous GPU computing power cloud environment are improved.

This invention relates to the field of data retrieval technology, specifically disclosing a database-based method for rapid retrieval of skills learning data, comprising the following steps: S1: Preset a retrieval period, statistically analyze the types, occurrences, and frequencies of retrieval fields within the period, and stop if the threshold is not met; S2: Construct binary retrieval field combinations, calculate the frequency and correlation coefficient of each combination field; S3: Single-field retrieval matches initial courses, calculates popularity based on the number of enrollees and collections, and sorts them; S4: Divide primary and secondary fields to correct popularity, sort and filter popular courses, and establish a connection between combinations and courses; S5: Prioritize displaying related popular courses, continuously update the period, and repeat the process after decomposing multi-field combinations. This method can improve retrieval accuracy and efficiency, ensure course timeliness and recommendation quality, adapt to the needs of skills training platforms, optimize resource allocation, and enhance the platform's core competitiveness.

The invention relates to the technical field of vehicles, in particular to a vehicle safety certification and control system, a vehicle and a control method.The vehicle safety certification and control system comprises a multi-mode biological characteristic quantization processing module which is used for obtaining at least one biological characteristic of a user and conducting coding fusion on the biological characteristics in a hardware operation environment to generate a quantum-state identity label; the dynamic key management and vehicle control module is used for completing identity authentication according to the quantum state identity credential, generating an encryption key of a vehicle control instruction according to the traffic situation information if the identity authentication is successful, and outputting the vehicle control instruction after safely packaging the vehicle control instruction by using the encryption key in response to user operation or a preset event, and the vehicle is controlled to execute corresponding operation. Therefore, the problems that in the related technology, biological characteristic transmission of a vehicle-mounted safety system is prone to being stolen, secret key management is static and rigid and cannot respond to road conditions in real time, and quantum hardware is difficult to stably work in the vehicle-mounted vibration and high-temperature environment are solved.

The application belongs to the technical field of databases, and specifically discloses a multi-mode time series data processing method and system. The method comprises the following steps: establishing a row storage table and a column storage table according to a table creation request, and establishing an associated mapping relationship between the row storage table and the column storage table; receiving time series data to be inserted and extracting label data and index data, determining a target data node based on label data hash calculation, writing the label data into the row storage table, and writing the index data into a target time series data block of the column storage table; performing partition pruning and scanning on the stored time series data based on a screening condition, and generating a query result by parallel execution of split query tasks and merging of results based on cross-mode query requirements; determining the affected time series data block based on an operation request, and updating the label data and the index data or deleting invalid data according to the type of the operation request. The application can realize low-redundancy storage, high-concurrency read-write and fast query analysis of time series data.

The invention relates to a thread work distribution method and device based on dynamic assembly, and the method comprises the steps: obtaining a first number of to-be-processed tasks belonging to a target priority in to-be-processed batch tasks, and judging whether a first maximum task number which can be processed by a first residual resource of a current thread pool is greater than the first number or not according to the first number, if the first maximum task number is greater than the first number, dynamically allocating the first residual resources to the first number of to-be-processed tasks; and if the first maximum task quantity is not greater than the first quantity, dynamically selecting partial or all to-be-processed tasks in the first quantity of to-be-processed tasks for allocation according to the first residual resource and the task attribute of each to-be-processed task. Through the method provided by the invention, the throughput rate in unit time is remarkably improved, and the overall response delay is shortened.