1250 results about "Hash table" patented technology

The invention discloses an android container rendering optimization method based on cross-domain hard real-time Fen synchronization, which comprises the following steps: by taking a swan-mong system as a host and an android system as a container, creating a hash table, a synchronous thread and a shared memory corresponding to a GPU core when the host is started, acquiring VSync cycle registration callback, transmitting shared memory FD to the container, and finishing shared memory mapping and alignment by the container. Registering a GPU queue to complete callback; when the Android application is started, a container obtains a queue and a physical address of a rendering buffer area, creates a Fen and binds the Fen to the queue, after the queue is submitted, metadata is written into a shared memory to inform a host, after the host receives the metadata, nodes are created and stored in a hash table, and an overtime timer is registered; after the GPU completes the command queue, the container calls back an update state and a verification value to notify the host, and after the host is verified to be valid, the corresponding hash table is updated, the timer is reset, and asynchronous screen loading is triggered; and the host executes buffer area synthesis and submission of the display equipment to finish on-screen, so that the stability of the rendering frame rate is improved, and the reliability of cross-domain synchronization is ensured.

The embodiment of the invention provides a binary translation method, a binary translator, electronic equipment and a readable storage medium, and the method comprises the steps: calling a preset instruction if a current instruction is recognized as an indirect jump instruction; the instruction code of the preset instruction comprises a first operand and a second operand; the first operand points to the first register, and the second operand points to the second register; the target PC value is stored in the first register; the second register is used for storing a calculation result of the preset instruction; the preset instruction executes preset calculation logic to calculate a target index value corresponding to the target PC value in the hash table, and the target index value is written into a second register; the target index value is read from the second register, whether the target PC value hits the hash table or not is inquired through the target index value, and the next instruction is executed based on the inquiry result. According to the embodiment of the invention, the instruction translation number of the high-frequency instruction can be reduced, and the execution efficiency of the binary translator is improved.

The invention discloses a mixed collision detection method in a virtual operation simulation system. The method comprises the following steps: 1) constructing a surgical instrument and soft tissue model; 2) dividing a virtual space into grids with the same size, indexing an object model, accelerating the construction and updating of a spatial information hash table by utilizing GPU multithreading, mapping the grids occupied by all primitive bounding boxes into the hash table, and determining objects in the same space according to the spatial hash table; 3) before object intersection solving, space reduction is carried out, and hash table data is optimized; 4) constructing a hierarchical bounding box for an object which may collide, and performing intersection detection; if it is detected that a certain node or child nodes of the node intersect, the next step is executed, and otherwise, detection is pushed out; and 5) accurately judging whether the basic triangular units contained in the nodes in the step 4) are intersected or not, and completing collision detection. According to the method, the problem of missed detection and wrong detection in the existing discrete collision detection algorithm is effectively solved, and the detection algorithm has excellent accuracy and calculation efficiency.

The invention provides a timestamp-based data change intelligent deduplication and message simplification method and system, and the method comprises the steps: capturing a change event of a database in real time, and generating a change record containing a timestamp; based on a preset timestamp comparison algorithm, performing timestamp analysis and comparison on the multiple change records of the same service entity; the change records are grouped according to the service keys, the latest change records are stored and dynamically updated through a hash table, the latest timestamp data under the same service key are reserved, and intelligent duplicate removal of redundant data is achieved; when the data volume exceeds a preset threshold value, a data fragmentation strategy is triggered, batch data is uniformly segmented into a plurality of subtasks, multi-thread parallel processing of fragmented data is achieved by dynamically allocating thread resources, and processing results are summarized; and carrying out merging processing on the deduplicated changed data, constructing a standardized message body, selecting immediate sending or delayed queue distribution according to a message priority, and controlling a network transmission load through a message batching mechanism.

The invention discloses an intelligent charging and discharging management method and system for a lithium battery, and the method comprises the steps: collecting multi-dimensional parameter data of the lithium battery, and generating a data set with a timestamp; and analyzing the data features based on the battery type classification model, and determining a battery type identifier. And utilizing the recurrent neural network to model the relevance between the capacity attenuation and the health state, and predicting the residual capacity and the health score. If the capacity or health score is lower than the threshold value, extracting the environment temperature and the load demand to generate a temperature-load feature vector; and matching the candidate strategy set from the pre-established index database through the hash table. And distributing weights according to health scores and load demands, sorting strategy efficiency and life influences by adopting a linear regression model, and screening an optimal strategy. And if the strategy calculation complexity exceeds the equipment capability, iteratively optimizing parameters by utilizing a genetic algorithm, simplifying a strategy instruction, generating a charging current, a voltage curve and a discharging rate control instruction, and executing the charging current, the voltage curve and the discharging rate control instruction in real time by a battery management system. The method prolongs the service life of the battery and improves the charge-discharge efficiency.

Provided in the present invention are a method and system for real-time update of a medical knowledge base, and a medium and a device. The method comprises: acquiring medical literature data in real time; analyzing the medical literature data, so as to obtain first structured data; performing correlation analysis on the first structured data and second structured data in a medical knowledge base, so as to obtain an analysis result; on the basis of the analysis result, determining whether to update the medical knowledge base; identifying the differences of the medical knowledge base before and after the update, so as to obtain a difference identification result; assigning a unique version number to each difference in the difference identification result, so as to obtain version update records; and on the basis of the version update records, using incremental hash table technology to perform data storage for each difference. The present application has a higher update efficiency without the risk of omission; and by means of difference analysis and by using incremental hash table technology to store difference data, the present application significantly reduces data storage space requirements, and also improves data access speed.

The invention belongs to the technical field of water surface target identification and tracking, and particularly relates to a tracking system and a tracking method for a water surface target under an incomplete information condition, which can be used for solving the problems of incomplete target information and tracking loss caused by wave shielding of an unmanned ship in a marine environment. The multi-modal feature extraction and fusion module is used for improving a convolutional neural network to extract visual features, constructing motion feature vectors in combination with multi-sensor motion information, and performing weighted fusion to obtain multi-modal features; the dynamic target feature library management module stores feature vectors and timestamps through a hash table-linked list structure, and maintains aging by means of a sliding window; when the target is reproduced, the re-identification algorithm module based on space-time joint matching integrates the feature similarity, the position distance and the time interval to determine the ID; the tracking parameter correction module uses Kalman filtering and combines a motion model and an observation value to correct parameters. The method improves the accuracy and stability of target tracking, and has application prospects in the fields of marine rescue, marine monitoring and the like.

The invention discloses a method for generating a three-dimensional building GLTF file based on index compression. The method comprises the following steps: S1, collecting three-dimensional coordinate data of a building; s2, grouping and classifying the data according to floors to form a classification index structure based on a hash table; s3, selecting a reference model and recording Mesh data and vertex coordinates of the reference model; s4, calculating a translation matrix T of the polyhedral structure, and performing coordinate transformation; s5, generating Mesh data of other polyhedrons based on index compression, and reducing redundant surface data by adopting a topology preserving surface optimization algorithm; s6, independently processing the geometric data of each floor, and generating corresponding Mesh data and Node nodes; and S7, combining all floor data to generate a complete GLTF file, and performing compression by using a dynamic Huffman coding algorithm. Through the improved index compression and dynamic Huffman coding technology, the file size is remarkably reduced, the transmission efficiency is improved, and the geometric and topological integrity of the model is ensured.

The invention provides a safe non-interactive comparison method and device for multi-party set intersection, and relates to the technical field of privacy set intersection comparison methods. The method comprises the following steps: carrying out data collection on software services of a plurality of service providers; determining a communication structure and a multi-party information exchange demand according to the advertisement scene delivery demand; performing offline data storage by adopting a cuckoo hash function to obtain a multi-party hash table; constructing a linear homomorphic hash function according to the multi-party hash table; generating a random number set offline through a trusted third party; according to the multi-party hash table, performing calculation based on one-out-of-two casual transmission to obtain a shared value set; and based on a communication protocol, the random number set, the linear homomorphic hash function and the sharing value set, performing non-interactive comparison according to a multi-party hash table to obtain an advertisement putting sharing data intersection. The non-interactive comparison method is based on non-interaction and is high in processing efficiency and high in universality.

Data structures and methods are described to provide a unique extendable group identifier for efficient aggregation query processing using multiple grouping keys. A method comprises retrieving a database query comprising an aggregate function of a selected column from a database table, grouped by a plurality of columns. The method further comprises maintaining a plurality of hash tables that map column values to grouping keys and identify a maximum grouping key. The method further comprises updating a plurality of bitmasks that define per-column grouping key bit positions, capable of storing the maximum grouping key, within a combined index value. The method further comprises allocating a result array sized according to the plurality of bitmasks. The method further comprises using the plurality of bitmasks and the plurality of hash tables to determine the combined index value to apply the aggregate function on the selected column for each row in the database table.

The invention discloses a pose estimation method and device based on point pair feature matching, a medium and a product. The method comprises the following steps: performing coarse registration on a fusion scene point cloud and each template point cloud, and calculating point pair features; according to the corresponding hash key values, retrieving from each hash table to establish registration point pairs; calculating a translation vector and a rotation quaternion matched with each registration point pair according to the reference point coordinate system and the template coordinate system, and performing clustering processing on each registration point pair to obtain each initial corresponding set; screening out a target pose candidate transformation matrix and a matched target template point cloud from the pose candidate transformation matrixes matched with the initial corresponding sets respectively; a target optimization function between the fusion scene point cloud and the target template point cloud is constructed, iterative optimization is performed on parameters in the target pose candidate transformation matrix, and the target pose of the fusion scene point cloud is obtained at the end of iteration, so that the precision, real-time performance and reliability of six-dimensional pose estimation are improved.

The invention discloses a method for positioning memory leak, which relates to the field of computer science and technology, and comprises the following steps of: initializing a memory use monitoring module, setting an initial memory use rate threshold, and defining a threshold adjustment parameter; the CPU utilization rate and the memory use condition are monitored in real time, historical data are recorded, and a memory use trend model is established. When the CPU utilization rate exceeds the set threshold value, reducing the first preset threshold value, adjusting the second preset threshold value according to the memory use trend, and comprehensively adjusting the memory utilization rate threshold value. And when the memory utilization rate exceeds a first preset threshold value, starting memory leak detection, recording memory allocation and release information by using a hash table, and searching the unreleased memory. And analyzing logs and behavior records of the system before and after memory leak, constructing a causal relationship model, adjusting a memory utilization rate threshold value, and generating a report containing memory leak position and reason analysis thereof. According to the method, the predictability and the accuracy of memory management are better improved, and the self-adaptive capability of the system is enhanced.

The invention discloses a semantic-based text content index automatic identification method, which relates to the technical field of information retrieval, and comprises the following steps: initializing sparse projection and LSH signature, performing iterative optimization by using a Lagrange duality form and gradient update, adjusting hash digits, obtaining a fragment index through a k-d tree, and constructing an inverted index. According to the method, compression is performed through Delta coding, an index map is constructed based on Jaccard similarity, compression is performed through WebGraph, CSNMF is used in combination with Z-Laplacian regularization, a low-rank basis matrix and a low-rank coding matrix are generated, a compressed inverted index is reconstructed after iterative optimization, and reconstructed inverted index entries are generated. According to the method, through multi-resolution hash table initialization, joint feature optimization, local adaptive quantization and low-rank index reconstruction, the semantic expression ability and the compression effect of an index structure are improved, the index precision and efficiency are improved, and intelligent identification of index content is achieved.

The invention discloses a data deduplication method and device, electronic equipment, a storage medium and a product. The method comprises the following steps: under the condition that data to be subjected to duplicate removal is obtained, inserting the data to be subjected to duplicate removal into a first hash table for performing hash duplicate removal until the data in the first hash table reaches a target number of rows, and outputting the data in the first target number of rows in the first hash table; under the condition that the first hash table is not full, continuing to insert the data to be subjected to duplicate removal into the first hash table for performing hash duplicate removal until the first hash table is full; under the condition that the first hash table is full, repeated detection is carried out on the data to be subjected to duplicate removal and existing data in the first hash table, and the data to be subjected to duplicate removal, which are judged to be non-repeated, are subjected to disk brushing; and under the condition that the processing of the data to be subjected to duplicate removal is completed, performing duplicate removal processing on all the data in the disk, outputting the data, and outputting the data except the data in the previous target number of rows in the first hash table. According to the scheme, the previous target number of rows of data obtained through duplicate removal can be returned in advance, and the user experience can be improved.

The invention relates to a laser radar-inertial navigation tight coupling SLAM (Simultaneous Localization and Mapping) method using ground constraint, belongs to the field of mapping of autonomous mobile equipment, and aims to solve the problems of relatively large Z-axis height positioning error, inaccurate point cloud map construction and low rear-end matching speed existing in long-time operation of a laser inertial SLAM method in the prior art. The method comprises the following steps: S100, constructing a laser inertial navigation SLAM system, wherein the construction is divided into IMU pre-integration, laser radar odometer and rear-end sliding window optimization; and S200, adding a functional module for simulating and extracting ground point cloud in the laser radar odometer, and using the functional module to restrain the posture of the robot / unmanned vehicle. And S300, constructing a self-adaptive voxel map by using the pose optimized and output by the rear end. According to the method, pose estimation is optimized by adopting a cloth simulation technology, Z-axis errors are reduced, an adaptive voxel map is introduced, and the matching efficiency and precision of the SLAM rear end are improved in combination with a hash table and a K-D tree.

The invention belongs to the technical field of information security, and particularly relates to a super-threshold data set intersection method and system for security information processing. The method comprises the following steps: S1, carrying out casual key value pair storage OKVS encoding and decoding between each participant and other participants; s2, each participant constructs a hash table according to a decoding result and elements of the participant; and S3, sending the constructed hash table to a specified party, summarizing information in the hash tables of all participants by the specified party, and finally obtaining a super-threshold intersection result by using a secret sharing scheme. The method has the characteristics that the hardware and operation and maintenance cost can be reduced, the data security is enhanced, and the calculation efficiency is improved, so that the industrial control requirement of short-time response is met.

The invention provides a cache cleaning method and device based on large model reasoning. The method comprises the following steps: monitoring the cache capacity of a hash table cache pool; when the current cache capacity of the hash table cache pool is greater than or equal to a preset capacity threshold value, obtaining access information corresponding to one or more cache entries, and screening a first to-be-cleaned cache entry through a preset cache cleaning algorithm according to the access information; and / or obtaining importance information corresponding to one or more cache entries, and screening a second to-be-cleaned cache entry meeting an importance cleaning rule according to the importance information; and according to the first to-be-cleaned cache entry and / or the second to-be-cleaned cache entry, cleaning the cache in the hash table cache pool, and updating the cache capacity of the hash table cache pool. According to the method and the device, the cleaned cache items are identified and determined according to the access information and / or the context importance information of the cache items, so that the cache redundancy and the accumulation of invalid data are reduced, and the cache management efficiency is improved.

The invention belongs to the technical field of laser scanning, and discloses a multi-station foundation laser scanning point cloud registration method and system suitable for a forest environment. Firstly, an original three-dimensional point cloud is converted into a two-dimensional distance image, a trunk is extracted based on continuity constraint, and the position of a trunk root point is searched; secondly, constructing a triangular descriptor, describing the spatial relationship among the trunks in scanning, and storing the spatial relationship in a hash table; and searching the corresponding triangular pairs by using the search operation of the hash table, calculating the similarity between scans, and extracting three-dimensional transformation parameters. And taking a to-be-solved scanning station pose as a node, taking relative transformation as a factor constraint, and obtaining a scanning station pose parameter with optimal overall consistency through factor graph optimization. And finally, by taking the conversion parameter after the factor graph optimization as an initial value, optimizing and adjusting the conversion parameter by using an ICP method, deriving the pose of each scanning station in the unified coordinate system through a depth-first strategy, converting the local point cloud into the unified coordinate system, and realizing global point cloud registration.

The invention provides a multi-dimensional Hash acceleration-based grid ray intersection method, which comprises the following steps of: acquiring original model data, and carrying out standardized preprocessing on the original model data to obtain standardized scene data; self-adaptive space grid division is carried out based on the preprocessed standardized scene data, space coordinates of three-dimensional grid units are mapped to a hash table to construct a one-dimensional index of a space grid, and triangular patches of the scene data are associated with the space grid; performing direction parameter decomposition on rays to be intersected, discretizing direction vectors of the rays, and constructing a ray direction composite hash key; and executing mixed intersection calculation in a mode of combining multi-level cache query construction, direction pruning and invalid intersection filtering and space hash traversal intersection, and determining intersection point information of the ray and the geometry. According to the method, the space grids and the ray direction data are converted into the hash values to be stored, the occupied memory is small, and the ray intersection efficiency is improved.

Techniques and systems described herein relate to monitoring executions of computer instructions on computing devices based on observing and generating a control flow directed graph. The techniques and systems include determining an observation phase for a process or application on a computing device. During the observation phase, CPU telemetry is determined and used to generate a control flow directed graph. After the control flow directed graph is generated, a hash table associated with frequently traversed execution paths is generated. A monitoring phase may be entered where transfers of instruction pointers are monitored based on the control flow directed graph to identify invalid transfers. The frequently traversed execution paths may be identified based on the hash table and be identified as valid if the hash value corresponds to the table.

The invention provides a multi-host network card port aggregation service hardware unloading system under a horizon operating system, which comprises a network card module based on an FPGA (Field Programmable Gate Array), which is configured with an FPGA module, a DMA (Direct Memory Access) module and a PCIE (Peripheral Component Interconnect Express) interface; and a plurality of hosts applying a Tianvessel operating system, each host is connected with the network card module through a PCIE interface, and exchanges data messages with the network card module through the DMA module in a DMA mode to realize network communication. A bond service management table item is configured in an FPGA module of a network card module, and the bond service management table item is composed of a bond member table, a bond binding mode table, a bond hash table and a member port state table. A Japanese vein operating system of each host is configured with a bond service module which is used for port aggregation service control under the Japanese vein operating system, and the creation of bond ports and the management of member ports are realized by creating and deleting a bond member table, a bond binding mode table and a bond hash table in the FPGA module. And updating the bond member table and updating the state of the network card equipment under the horizon operating system of the host through polling detection.

The invention discloses a path selection method and device for a source host end in a data center network, and the method comprises the steps: employing an ECMP Hash linearity principle in the data center network, and achieving the precise path selection method and device for the source host end. According to the method, the correlation between the relative ECMP hash change and the final hash value is deeply mined, and an effective multi-bit adjustment scheme in a data stream header is identified in combination with a known switch single-bit hash table, a network topology structure, routing information and ECMP configuration. In this way, the source host can flexibly and accurately select the multi-hop forwarding path to the target host by adjusting the data stream header. According to the accurate path selection method and device provided by the invention, the source host obtains the autonomy of path selection, so that path distribution is optimized, path overlapping and overuse of switch ports are remarkably reduced, and the overall network performance is further improved.

Examples described herein relate to key-value lookups. A first index is calculated based on a first hash function and a received key. A first hash table including entries associated with a first hash calculation on a key is accessed. A collision hint table is accessed to identify any other hash table to search. If the collision hint table indicates any other table to search, the first hash table and the any other hash table are searched to identify an entry associated with an index that matches the first index.

The invention relates to the technical field of data recognition, and discloses an abnormal data recognition method, device and equipment and a storage medium, and the method comprises the steps: extracting data distribution characteristics in each time window in a high-dimensional time series data stream, and carrying out the recognition of abnormal data based on the data distribution characteristics and an adjusted random projection matrix of the high-dimensional time series data stream; the method comprises the steps of generating dimension-reduced low-dimensional data, configuring local sensitive hash resolution parameters according to local volatility of the low-dimensional data, constructing a multi-resolution hash table, and identifying abnormal data points according to in-bucket data density of the multi-resolution hash table. According to the method, the dimension reduction process is adjusted by extracting the data distribution characteristics, the calculation complexity is reduced while key information is reserved, the multi-resolution hash table is constructed based on the local volatility of the dimension reduction data, adaptive analysis of different data areas is achieved, abnormal data points are identified according to the density of the data in the barrel, and the accuracy of the data in the barrel is improved. The dimension problem and the distribution change problem can be effectively solved, and the processing efficiency is improved while the detection precision is ensured.

An information handling system executing a predictive hash table and carbon dioxide (CO2) minimizing system may comprise a network interface device to receive from a remote data center deduplication process workload metrics, fingerprint hash table size metrics, and a current reserved memory size allocated for storage of a fingerprint hash table, a processor executing code instructions to predict a future workload of deduplication processes and a future fingerprint hash table size to determine a predicted amount of memory to reserve for storage of a future fingerprint hash table based on the predicted future workload of deduplication processes and fingerprint hash table size, and to determine an over-allocated portion of the current reserved memory size where the network interface device transmits an instruction to the data center to limit power provided to the over-allocated portion of the current reserved memory size.

The invention discloses an astronomical cross identification method based on spherical trapezoidal division, and the method comprises the steps: S1, taking a unit spherical surface as a celestial sphere, dividing the celestial sphere into a plurality of strips parallel to an equatorial line, and determining the right ascension span of each strip for the transverse segmentation of the strip; s2, dividing each strip into a same spherical isosceles trapezoid with two waists perpendicular to the equator, and taking the same spherical isosceles trapezoid as a sky area; s3, determining an index of each sky area; s4, dividing a star catalogue A and a star catalogue B, and creating a zipper type hash table based on the row number of the star catalogue B; s5, mapping each celestial body in the star catalogue B into a linked list in a hash table in a multi-thread parallel mode, and inserting celestial body data into the linked list by means of a two-stage insertion strategy; s6, mapping each celestial body in the star catalogue A into a linked list in a group of hash tables in a multi-thread parallel mode, and traversing to obtain celestial body IDs of which the spherical distance is smaller than the error radius; the method not only can completely avoid the problem of result omission, but also can greatly improve the astronomical cross identification calculation performance.

The invention is suitable for the technical field of smart power grids, and relates to a Wi-SUN communication emergency deduplication fixed weighted QoS management method, a medium and a terminal, and the method comprises the steps: S10, carrying out the classification of the service data type and data packet grade of an intelligent electric meter of an AMI system; s20, establishing a QoS mathematical model of the Wi-SUN communication network of the intelligent electric meter, and obtaining an optimization objective function and a communication bandwidth constraint of the Wi-SUN communication network; s30, according to the Wi-SUN communication network optimization objective function and the communication bandwidth constraint, fairly allocating bandwidths to different flow queues by adopting a weighted fair queue flow scheduling algorithm, and considering the weight of each queue at the same time; and S40, de-duplication of the intelligent electric meter service data Wi-SUN network optimization emergency notification is carried out through the hash table. According to the method, dynamic adjustment and planning of the Wi-SUN network communication service can be effectively carried out according to the size of the service data packet, the data characteristics and the data priority degree of the intelligent electric meter, and optimization of the Wi-SUN communication network is achieved, so that the communication acquisition success rate is ensured.

The invention relates to a message rule searching method. The method comprises the following steps: searching a first candidate quintuple rule in a preliminary screening hash table according to address field data of a target message; the storage position of each quintuple rule in the preliminary screening hash table is determined according to the hash value of the address field data in each quintuple rule; searching a second candidate quintuple rule in the collaborative screening hash table according to a quintuple combination template matched with the address field data in the collaborative screening hash table and the quintuple field data of the target message; the storage position of each quintuple rule in the collaborative screening hash table is determined according to the hash value of each quintuple rule; and determining the first candidate quintuple rule and the second candidate quintuple rule as the quintuple rule of the target message. By adopting the method, the rule searching process of the message can be simplified, and the processing performance of the message can be improved.

The invention provides a method, a device and equipment for realizing a high-concurrency extensible hash table (hash table) in a multi-thread environment and a medium. The implementation content of the method comprises two parts: 1) realizing high concurrency by a hash table fine-grained lock: realizing by adding the fine-grained lock to the hash table, locking each bucket of a bucket array of the hash table, and improving the concurrency of accessing the hash table in a multi-thread environment; 2) expandability of the hash table double-bucket array structure: two bucket arrays are used, the original bucket array is used for storing hash table elements in a non-expansion state, the expansion bucket array is used for storing newly added hash table elements and elements migrated from the original bucket array in an expansion state, the expansion bucket array is used as the original bucket array in an expansion state after migration is completed, and the expansion bucket array is used as the original bucket array; according to the method, the problem of realizing the high-concurrency extensible hash table in the multi-thread environment is solved, the high-concurrency access of the hash table in the multi-thread environment is ensured, the expandability of the hash table in the multi-thread environment is also ensured, the high concurrency and the expandability are carried out in parallel without conflict, and the hash table can adapt to the change of the data scale.

The invention relates to a method for realizing NAT on a core network UPF network element, and the method comprises the steps: when a user terminal establishes a 5G session, the UPF pre-allocates a public network IPv4 address and a continuous source port set for the session, the number of ports is dynamically adjusted according to a user package, and resources are bound to a session data structure; receiving a downlink packet of which a destination address is a public network IP based on a private network-to-public network mapping relation, positioning a session level hash table through a flow label direct index mechanism, adding a GTP-U packet header after reverse NAT conversion, and forwarding the packet to a mobile phone through service processing; on the basis of a stream label direct index mechanism, each session exclusively occupies an IP port pool and a hash table through a session-level resource isolation architecture, and is used for avoiding global lock competition in combination with multi-core CPU parallel processing; based on a DPI function switching strategy, an NAT mapping table is synchronized through a UPF cluster session mirroring mechanism for node fault switching zero packet loss.