128 results about "Concurrency" patented technology

The application provides a sideband compensation acceleration method, an acceleration device, a computing device and a storage medium. The sideband compensation acceleration method comprises the following steps: obtaining sideband data; adaptively setting a candidate concurrency parameter and determining an optimal concurrency parameter P; dividing the sideband data into a plurality of vectors according to the determined optimal concurrency parameter P; performing parallelization calculation on each divided vector and judging whether each vector matches; and updating the sideband data according to the matching state of the plurality of vectors. Through the sideband compensation acceleration method of the application, sideband compensation data can be calculated in parallel based on an adaptive compensation algorithm, and the coding efficiency is improved.

This application provides a strategy scheduling method, system, electronic device, and storage medium that combines method-level strategy encapsulation with pre-compiled execution, relating to the fields of software architecture and runtime scheduling optimization. The method encapsulates multiple strategy methods with the same predefined commonalities within each strategy category using a small number of methods, changing "class-level" encapsulation to "method-level" encapsulation to reduce the number of strategy classes at the source. It optimizes calls by replacing traditional reflection calls with pre-compiled execution handles and constructs a hierarchical concurrent caching structure, designing a concurrent index layer and a hotspot caching layer. High-frequency strategies are directly retrieved from the hotspot caching layer, and all strategies are distributed across the two caching layers to reduce hash collisions. A hotspot update mechanism maintains the caching of high-frequency strategies in the hotspot caching layer, improving retrieval and scheduling efficiency under high concurrency. Compared to traditional class-level strategy implementation schemes, this method achieves significant performance and resource optimization in scenarios with a large number of strategies and high concurrency scheduling.

This invention relates to an enterprise-level full-stack low-code development platform based on a model-driven and plug-in architecture, belonging to the field of computer technology. It addresses the problems of poor scalability, performance bottlenecks, high security risks, and inflexible deployment inherent in existing platforms. The invention includes: a visual development tool providing a form designer and workflow engine, supporting drag-and-drop development and real-time preview; a plug-in extension module, implemented based on the PF4J framework, supporting dynamic loading, unloading, and isolation of plug-ins; a data integration and management module providing an API gateway, multi-source data integration, and data model management; and an application deployment and maintenance module for one-click automated deployment and log auditing. This invention achieves dynamic functional expansion through a plug-in architecture, optimizes performance in high-concurrency scenarios through a workflow engine, enhances security, and provides more flexible deployment, improving the development and maintenance efficiency of enterprise-level applications. It is suitable for application scenarios such as government approvals.

The present application relates to the technical field of log field processing, and provides a log field processing method and system based on a description model and a pluggable strategy to solve the problems of inflexible configuration, insufficient performance and difficulty in dynamic updating commonly faced by log field processing technologies.The log field processing method based on the description model and the pluggable strategy comprises receiving current log data and structurally analyzing the same to obtain a log data structure description model; reading pre-generated structured field processing configuration data and analyzing the same to determine a field path to be processed in the current log data, locate a target field to be processed in the log data structure description model, call a corresponding independently encapsulated pluggable field processing strategy unit, and then execute a corresponding field processing operation, write the processed field value back to a corresponding field position in the log data structure to update the log data content, and meet the log field processing requirements in modern high-concurrency and complex business scenarios.

The application discloses a Java-based distributed transaction management system and relates to the technical field of distributed systems. In view of the contradiction between transaction consistency and high concurrency and high availability in a distributed system, the scheme comprises the following: a transaction coordination module, which is used for constructing a cluster based on a Raft protocol to guarantee continuous availability of a service, realizing transaction full-life cycle management by relying on an extended state machine engine, and dynamically selecting an optimal communication path of a transaction according to scheduling decisions and a real-time network topology state; a local transaction management module, which is embedded in each microservice and is used for realizing local execution and state reporting of a transaction; a multi-level log module, which is used for realizing traceability of a transaction state and recovery of a fault through multi-level storage and event tracing; an intelligent scheduling module, which is used for realizing dynamic optimization of transaction processing based on an LSTM model and QoS indicators; and a monitoring and alarming module, which is used for collecting transaction full-link logs and performance indicators in real time and realizing full-link observability. The application is suitable for high-reliability and high-concurrency scenarios.

This invention relates to the field of computer file system technology, and in particular to a file system kernel service and high-concurrency scheduling method based on a unified protocol. The method includes the following steps: Step 1: Starting a receiving thread and a file operation thread on the server side, the file operation thread including a lightweight thread and a heavyweight thread; Step 2: Listening to the network port through the receiving thread to receive JSON-formatted requests sent by clients, the requests containing timestamps and operation types; operation types include copy, move, create directory, delete, list, and restore; Step 3: Storing the received requests in a receiving queue; Step 4: Retrieving requests from the receiving queue and distributing them according to the operation type; Step 5: Executing the operation; This invention exposes file operation capabilities through independent processes and a unified JSON protocol. The six types of operation semantics, as well as the semantics of abort and recoverable deletion, are all uniformly defined at the protocol layer. The dual-type threads and request queue improve throughput and stability under multiple clients and large-scale tasks.

This invention discloses a real-time data anomaly monitoring and repair method based on machine learning, comprising the following steps: S1, constructing data fragments and generating a consistent constraint set; S2, performing self-consistent representation construction; S3, performing anomaly detection and generating anomaly label set; S4, performing masking processing to generate masked data fragments, inputting them into an improved UniTS model, introducing a structural perturbation suppression mechanism, and generating a repair candidate set; S5, performing structural consistency evaluation and screening target repair data that meets the minimum structural perturbation condition; S6, performing data repair and generating repaired data fragments; S7, feeding back the updated consistent constraint set to the improved UniTS model. This invention achieves highly sensitive real-time identification of anomalous data items, effectively reducing the anomaly false judgment rate and repair distortion rate, and improving the data security, stability, continuity, and reliability in high-concurrency data stream environments.

The application discloses a GC mechanism processing method and device and medium, and is suitable for the technical field of solid state disks. When the RGW service is started, a GC main thread, a queue and a work thread are created, wherein the GC main thread starts the GC mechanism; garbage data obtained by the GC main thread traversing a storage pool is stored in the queue; and the queue sends the garbage data to the work thread in a de-queue mode so that the work thread deletes the garbage data. The method uses the thread division mode, the GC main thread starts the GC mechanism, when the GC mechanism is started, garbage collection is processed, the main thread is responsible for traversing the garbage data, and the work thread deletes the garbage data, thereby effectively improving the processing speed of the GC mechanism, relieving the pressure of only executing garbage collection through a single thread at present, and guaranteeing the stability of the system. Meanwhile, the multiple threads execute the deletion of the garbage data, high concurrency is realized, and available disk space is released.

This application provides a method, apparatus, device, medium, and product for multi-threaded concurrent log reading. It relates to the field of data processing technology. The method includes: controlling one thread to acquire a mutex lock through a preset mutex lock management mechanism; determining whether concurrent readability conditions and overall readability conditions are met; in response to both conditions being met, determining whether logs can be read from a preset buffer based on a preset log header threshold and the current preset buffer length; if it is determined that logs can be read from the preset buffer, verifying whether the current log file status is normal based on the determined current total log entry length, a preset file layout threshold, and the current preset buffer length; in response to successful verification, updating the global two-pointer flag value and controlling the thread to release the mutex lock through the preset mutex lock management mechanism. The method of this application achieves safe log reading for multiple concurrent threads, significantly improving concurrency performance in a multi-threaded environment while effectively avoiding data conflicts.

A black screen control system for nitric acid production, comprising network hardware nodes deployed in a local area network and control units running on the network hardware nodes, the control units are independently deployed at each operation station and adopt a multi-thread isolation architecture. The invention allocates time-consuming bottom layer data collection, processing and front-end UI interaction to independent threads for asynchronous execution, ensuring the ultimate smoothness and stability of the system under high concurrency; the state control and abnormal response are separated, the system enters the energy-saving black screen state under normal conditions to reduce power consumption, and when an abnormal signal occurs, the system is instantly awakened across threads and a safety alarm is triggered, realizing true safe unattended operation; the control units are independently deployed and run at each operation station, and are uniformly scheduled by a master control module, avoiding the whole line paralysis caused by single point failure or network fluctuation, and ensuring the absolute independence and reliability of a single device.

This invention discloses a data model-driven digital human emotion guidance method, belonging to the fields of artificial intelligence and human-computer interaction. It extracts emotional features from speech and text separately using a lightweight gated convolutional network and a pre-trained model incorporating multi-head self-attention, and aligns emotional semantics using a cross-modal attention mechanism to obtain fused features. Based on these features, user emotions are identified, and emotional guidance instructions are generated using a knowledge base to drive the digital human to output emotional voice feedback. Efficient message synchronization in high-concurrency scenarios is achieved through a NATS message middleware based on a publish / subscribe model, and a dual-mode approach is introduced to ensure interaction continuity. This invention significantly improves the accuracy of emotion recognition and the naturalness of feedback, while achieving extremely low CPU utilization and high network throughput with limited hardware resources, stably supporting tens of thousands of concurrent users.

Methods, systems, apparatuses, devices, and computer program products are described. A group-based communication system may use machine learning techniques to classify teams of the system, determine discount messaging for teams of the system, or both. The group-based communication system may receive concurrency data for a team of users and may input the concurrency data (e.g., with one or more other features associated with the team) into a machine learning model to generate a classifier for the team. The classifier may indicate whether the team is a work team, an educational team, or a social team. Based on the classifier for the team, the system may send a message to at least one user of the team (e.g., an administrative user). In some examples, the system may use another machine learning model to generate a discount message for sending to the at least one user based on the team classifier.

Techniques for detecting potential concurrency issues and / or causality bugs are disclosed. To this end, the system maintains at least one object graph of event objects. The event objects represent events, and the event objects in an object graph are interconnected by references that represent causal and / or temporal relationships between the corresponding events. Before a thread attempts a task that could feasibly conflict with another task, the system may compel the thread to consult the object graph. For instance, if a task entails accessing a shared object, the system may compel a thread to determine if the last event to mutate the shared object is represented in an object graph of event objects before an event corresponding to the thread's current attempt to complete this task. If the last mutation event is absent from the object graph, the system may prevent or delay the thread from completing the task.

This invention discloses a log collection method, apparatus, device, and medium for a multi-core system. The method is executed by the master core and multiple slave cores in a multi-core system, including: writing log strings into a buffer in the shared memory of the slave cores and generating log descriptors; writing the log descriptors into a lock-free circular queue; monitoring the state count value of the lock-free circular queue in real time by the master core, and obtaining a target log descriptor when the lock-free circular queue is determined to be non-empty; determining a target buffer in the shared memory of the slave cores and transferring the log strings in the target buffer to the master core memory; and determining that the log upload conditions are met based on a timer or master core memory usage, determining at least one target master core memory block in the master core memory and uploading the log strings in the target master core memory block. This technical solution can improve log writing efficiency and log upload real-time performance in high-concurrency scenarios of multi-core systems.

The application relates to the technical field of digital information processing, and discloses a digital information processing system for agent task distribution, which comprises the following modules: a state quantization module, which is used for realizing binary processing on operating parameters according to a benchmark threshold array and generating an ability bitmap vector; a topology construction module, which is used for sequentially stacking the vector to construct a global state bitmap matrix; a constraint conversion module, which is used for compiling task constraints into a demand bitmap vector; an activity screening module, which is used for realizing logical filtering on the matrix by using an activity bit mask to generate a candidate resource bitmap; and a matching distribution module, which is used for calculating a bitwise AND result and issuing an instruction. According to the application, continuous state quantity is mapped to a Boolean logic space, the matching cost is changed from floating-point operation into bit operation which is supported by a processor originally, the computing power deadlock under high-frequency concurrency is eliminated, and it is ensured that the distribution action is completed within a determined clock period.

The application relates to the technical field of model configuration, in particular to a software architecture framework configuration and generation method based on UML meta-model extension technology. The application proposes the following scheme: a UML activity diagram is constructed to depict function execution flow, and UML interaction constraint conditions containing time sequence, concurrency, branching and interface constraint are generated from the activity diagram; under the constraint of the interaction constraint, a UML component meta-class is extended, meta-attributes such as framework semantic type, running environment and generation configuration are introduced, and an initial software architecture framework model is constructed; in combination with a component knowledge graph library, an outwardly-biased distribution strategy is adopted to distribute activity nodes to reusable components or newly-built components, and a software architecture framework meeting interaction consistency and generation constraint is generated. The application can improve the traceability, reuse stability and engineering generability of architecture configuration.

The application discloses a kind of dynamic rights allocation method, device and equipment based on user contribution value and medium.The method is captured by front-end system user interaction behavior event, is forwarded to B system event hub after being verified and enhanced by A system, then asynchronously is delivered to message queue and is handled.Consumption end calculates contribution value variation according to dynamically configured rule base table, supports positive and negative adjustment of civilization contribution value and platform contribution value.System uses MySQL and Redis double storage architecture, and guarantees data consistency under high concurrency by atomic operation.Based on formula β (i)=U (i) / (G (t)-G (0)) Real-time calculation user contribution proportion, provide proportion type rights release, ladder type ranking empowerment, threshold type screening incentive Three allocation strategies.Through event-driven architecture and cache consistency algorithm, millisecond-level contribution value update and second-level proportion accounting are realized.The application solves the problems of single contribution measurement mechanism, rights allocation rigidity and calculation performance bottleneck in traditional integral system.

The application discloses a resource processing method, device and equipment of a computing task and a readable storage medium. The method comprises the following steps: obtaining running index information of a computing task in a first time period; the computing task comprises N subtasks running concurrently, and the running index information of the computing task comprises running index information of each subtask; based on the running index information of each subtask, the demand amount of computing resources of each subtask in a second time period is predicted and analyzed to obtain the predicted demand resource amount corresponding to each subtask; based on the predicted demand resource amount corresponding to the N subtasks respectively, M subtasks satisfying a resource adjustment condition are obtained from the N subtasks; and the M subtasks are adjusted in terms of resources in the second time period. According to the application, the computing resources of the computing task can be effectively adjusted without changing the concurrency of the computing task, and the task backlog can be reduced.

This application provides a method, apparatus, and device for preventing deadlocks in batch data writing based on primary key ordering. The method involves extracting the primary key and normalizing its type from the set of records to be written to generate a normalized primary key string; performing primary key ordering sorting based on the normalized primary key string to generate a logically ordered record sequence; performing intelligent batch sharding on the logically ordered record sequence to generate sharded record batches; and performing batch data writing based on the sharded record batches according to their ordered arrangement, while also implementing adaptive concurrency adjustment and backoff retries. This scheme, by forcing all concurrent transactions to acquire locks in a completely consistent order, mathematically breaks the circular wait condition that leads to deadlocks, reducing the deadlock rate to an extremely low level and improving data processing throughput and system stability in high-concurrency scenarios.

The application discloses a weight-based dynamic flow control method and system. The gateway method comprises the following steps: capturing multi-dimensional request data carried by an access request sent by each target service as a real-time metering sequence; performing abnormality detection on the sequence and a pre-trained abnormality detection model to generate an abnormality detection result of each target service; obtaining basic parameters of each service; calculating a normalized dynamic weight of each target service at a current time according to the basic parameters and the abnormality detection result; updating a current flow control parameter of each target service according to the normalized dynamic weight to obtain a latest flow control parameter of each target service; and performing flow control on the access request based on the latest flow control parameter of each target service. Therefore, the application realizes adaptive and differentiated control of the distributed system flow, and improves the stability, load balancing efficiency and overall reliability of the system in a high-concurrency and flow fluctuation scenario.

Techniques for supporting a token based service are described. In some examples, a token based service such as a foundation model service supports a higher performance tier with guaranteed throughput. In some examples, the throughput is expressed in tokens per minute (input and / or output). The service performs concurrency-based throttling for the shared resource according to the guaranteed throughput.

This application discloses a stress testing method, apparatus, device, and computer-readable storage medium, belonging to the field of system performance testing technology. The method includes: determining the concurrency level based on the number of centralized test data sets of the system under test and the number of concurrent processing data sets of a single server in a server cluster; generating a stress testing application corresponding to the server cluster; sending the stress testing application and the concurrency level to the server cluster; and obtaining the stress testing data fed back by the server cluster. The above technical solution can easily obtain the stress testing data fed back by the server cluster and does not require the use of a dedicated stress testing machine, thus simplifying the stress testing method.

The application discloses an omap operation performance test method and device and a medium, and relates to the field of distributed storage. Test parameters are initialized, the test parameters at least including a running time and a concurrency number; an operation queue is generated, the length of the operation queue being equal to the concurrency number; and an omap request is generated, the omap request including an omap write request or an omap read request, the number of the omap request being equal to the concurrency number, each omap request corresponding to a slot of the operation queue; the omap request is processed asynchronously through the operation queue, it is judged whether the processing process meets a preset condition, if yes, test statistical data is output. It can be seen that the above scheme can test the write operation and the read operation of the omap operation in the distributed storage system. Since the concurrency number is set and the asynchronous processing mode is adopted, the test pressure intensity can be controlled through multi-concurrency, so that the limit value of the omap operation performance is obtained, and the method is very reliable.

The application provides a logistics information prediction method, an electronic device, a storage medium and a program product, and relates to the technical field of computers. The method can realize prediction in an offline and online mixed manner through a light gradient boosting machine classification model and a light gradient boosting machine regression sub-model, can achieve a balance between predicted response load and disaster tolerance, and can meet the real-time query requirements of high concurrency and low delay of an e-commerce platform.

This application provides a CDN scheduling concurrency adjustment method, apparatus, computer equipment, computer-readable storage medium, and computer program product, belonging to the field of content delivery network technology. The method includes: acquiring historical actual bandwidth data, a preset bandwidth cost line, and a basic scheduling concurrency within a preset statistical window; determining a bandwidth gap factor based on the historical actual bandwidth data and the preset bandwidth cost line, the bandwidth gap factor being used to characterize the degree of bandwidth deviation; correcting the basic scheduling concurrency based on the bandwidth gap factor to obtain a target scheduling concurrency; and executing user scheduling based on the target scheduling concurrency. The technical solution of this application embodiment can reduce bandwidth usage costs.

The application discloses a large model intelligent question and answer acceleration method for a high concurrency scene, comprising the following steps: receiving a user input question and answer request, extracting core query keywords and semantic features, and generating a query vector; querying a high-frequency intention cache; if the high-frequency intention cache is not hit, querying a high-frequency question and answer pair cache; if the high-frequency question and answer pair cache is not hit, querying a historical question and answer result cache; if none of the three layers of caches is hit, entering a mixed semantic retrieval process; starting vector retrieval and keyword matching double-path parallel retrieval, and generating a final reference knowledge set after deduplication and integration; large model dynamic reasoning pruning: for a new request that is not hit in the cache, dynamic reasoning pruning optimization is performed, and a large model is called to generate a question and answer result. The application also discloses a large model intelligent question and answer acceleration system. The application solves the problems of high response delay, large resource consumption, low retrieval efficiency and insufficient service stability of an existing large model intelligent question and answer system in a high concurrency scene.

This invention relates to the field of 5G NR technology, specifically disclosing a neural network signal processing method and system for 5G NR SDR systems. Addressing the problems of USRP FIFO overflow, difficulty in real-time neural network deployment, and low efficiency of multi-GPU collaboration in existing SDR systems, this invention constructs a multi-threaded PHY processing architecture. The PHY layer is decomposed into independent threads bound to CPU cores, with each thread independently processing all physical channels in a time slot. A GPU task scheduler and GPU pool are designed to achieve asynchronous task submission and intelligent load balancing. Weight preloading, task encapsulation, and a double-buffered weight update mechanism are employed, and asynchronous online training is performed through an asynchronous execution framework. This invention overcomes the single-core CPU bottleneck through time-slot-level parallel PHY processing, reduces multi-threaded concurrency overhead through GPU collaborative scheduling, and achieves uninterrupted real-time inference and continuous adaptation of the neural network through double buffering and asynchronous training, significantly improving channel estimation accuracy and system throughput.

The present application relates to a kind of smart grid monitoring data storage method and system, specifically relates to the field of smart grid, by lightweight feature extraction in edge side, generate the metadata rich in semantic information, significantly reduce data transmission overhead and central storage pressure;Center system is based on reinforcement learning dynamic selection encoding strategy, and according to the intelligent scheduling of algorithm characteristics to heterogeneous computing node execution, realizes the optimal balance of storage efficiency and processing speed;At the same time, through online learning mechanism continuous optimization decision model, form a self-adapting closed loop, comprehensively improve the real-time performance, reliability and resource utilization efficiency of data storage under the high concurrency monitoring scenario of smart grid.

The application discloses a GPU resource efficient scheduling method based on multi-scene intelligent interaction, comprising the following steps: step one, generating an interaction request scheduling descriptor based on an intelligent interaction request; step two, constructing a candidate resource action set, and screening an executable candidate resource action set according to a display memory capacity constraint and a concurrency capacity constraint; step three, generating a feature vector; step four, inputting the feature vector into a pre-trained improved NGBoost model to generate a complete time delay condition distribution for the executable candidate resource action set; step five, calculating a default probability; step six, screening the candidate resource action according to an allowed default upper limit, and determining a to-be-executed resource action according to the default probability; and step seven, executing GPU resource configuration and updating tail distribution parameters by using an execution result. The application realizes controllable GPU scheduling risk and improved resource utilization efficiency.