38results about How to "Reduce latency overhead" patented technology

The invention discloses a MapReduce optimizing method suitable for iterative computations. The MapReduce optimizing method is applied to a Hadoop trunking system. The trunking system comprises a major node and a plurality of secondary nodes. The MapReduce optimizing method comprises the following steps that a plurality of Hadoop jobs submitted by a user are received by the major node; the jobs are placed in a job queue by a job service process of the major node and wait for being scheduled by a job scheduler of the major node; the major node waits for a task request transmitted from the secondary nodes; after the major node receives the task request, localized tasks are scheduled preferentially by the job scheduler of the major node; and if the secondary nodes which transmit the task request do not have localized tasks, prediction scheduling is performed according to task types of the Hadoop jobs. The MapReduce optimizing method can support the traditional data-intensive application, and can also support iterative computations transparently and efficiently; dynamic data and static data can be respectively researched; and data transmission quantity can be reduced.

This invention provides a system and method for processing discrete image data within an overall set of acquired image data based upon a focus of attention within that image. The result of such processing is to operate upon a more limited subset of the overall image data to generate output values required by the vision system process. Such output value can be a decoded ID or other alphanumeric data. The system and method is performed in a vision system having two processor groups, along with a data memory that is smaller in capacity than the amount of image data to be read out from the sensor array. The first processor group is a plurality of SIMD processors and at least one general purpose processor, co-located on the same die with the data memory. A data reduction function operates within the same clock cycle as data-readout from the sensor to generate a reduced data set that is stored in the on-die data memory. At least a portion of the overall, unreduced image data is concurrently (in the same clock cycle) transferred to the second processor while the first processor transmits at least one region indicator with respect to the reduced data set to the second processor. The region indicator represents at least one focus of attention for the second processor to operate upon.

The invention discloses a method for realizing network element level parallel service functions in a network function virtualization environment. The method comprises the steps of traversing a servicefunction chain, operating a service function decomposition analysis algorithm and decomposing service functions into basic message processing units, called as network elements, and determining and storing operation and operation domain information of the service functions for messages; operating a parallel judgment algorithm through utilization of the obtained operation and operation domain information of the service functions, and determining a service function combination on which parallel optimization can be carried out in the service function chain; performing a parallel optimization algorithm, and combining the parallel service functions; and through combination of non-parallel service functions and the parallelly optimized service functions, organizing and combining the service functions according to an original service function sequence, and establishing a new service function chain. According to the new service function chain, the length of the original service function chainis effectively reduced, the parallelism among the service functions is improved, and the delay cost when the messages pass through the service function chain is clearly reduced.

The invention relates to an access authentication method of electric automobiles. The method comprises the following steps that: access authentication systems of the electric automobiles are established; access authentication between the electric automobiles and a charging station is carried out; and the charging station carries out batch signature authentication on the electric automobiles. According to the invention, identity aggregate signatures are introduced into identity validity inspection of the electric automobiles, and bidirectional authentication and session key negotiation are designed between the electric automobiles and the access authentication systems, so that the reasonable usage of the information resources in the station is ensured, and the safety of the whole network system is fundamentally ensured. The batch signature authentication mechanism enables the charging station to verify a plurality of electric automobile terminal signatures in the aggregate manner, the calculation load is lightened, and the performance is improved. The typical identity aggregate signature system is composed of five algorithms: system initialization, private key generation, signing, aggregation and aggregate verification, user identity information can be used to replace a public key thereof, the problem of high certificate cost is solved, and by compressing the plurality signatures into one, the high-efficiency verification is carried out.

The invention discloses a data flow control method for an SSD backend. The SSD backend includes a command scheduling module, a data transmission control module, an interference and CRC and ECC module,and a Nand time sequence control module. In the method, a first command channel and a first information feedback channel are established beforehand between a data transmission control module and a command scheduling module, and a second command channel and a second information feedback channel are established between the Nand time sequence control module and the command scheduling module. By separating operation process areas of read/write commands from other operation commands and adjusting different command receiving channels and completing the information feedback channel, the cost of delaying waiting data at a Nand interface is reduced, and the efficiency of command response at the Nand interface is improved.

The invention provides a method used in an embedded CPU (Central Processing Unit) and used for reducing program skip overhead and circuit implementation thereof. According to the method, historical score values are added into the skip address lookup table entries, score comparison is carried out on the table entries, and therefore the most frequently-occurring skip destination address is reservedin the lookup table entries, and the skip delay overhead is effectively reduced. According to the method, in terms of hardware implementation, a CPU classic cache structure is adopted, and the score value is used as a table item replacement basis.

A file write-in method in a distributed system and an apparatus thereof are disclosed. The file write-in method comprises the following steps that after receiving a write request for a first file, a client sends a data write-in request of the first file to each storage node corresponding to the first file respectively, wherein the data write-in request at least includes a write request identification corresponding to the first file data write-in request and a storage node address; the storage node address includes an address of a main storage node and/or an address of a secondary storage node in each storage node corresponding to the first file; the storage node address is used for interaction between the main storage node and the secondary storage node; and the client receives an execution result of the data write-in request from each storage node corresponding to the first file respectively. In the invention, under the condition that multiple copy byte orders of the data can be guaranteed to be consistent, data sending hop counts are reduced so that write-in time delay is decreased.

The invention discloses an FPGA (Field Programmable Gate Array) virtualization hardware system stack design for cloud deep learning reasoning, and relates to the technical field of artificial intelligence, the design comprises a distributed FPGA hardware auxiliary virtualization hardware architecture, a CPU (Central Processing Unit) server node for running a virtual machine container, a static compiler and a deep neural network DNN, wherein the deep neural network DNN is used for acquiring a user instruction, and the user instruction is compiled into an instruction packet through the static compiler; the FPGA server computing node is used for operating virtualization system service and an FPGA acceleration card, and the FPGA acceleration card comprises a virtualization multi-core hardware resource pool and four double-rate synchronous dynamic random access memories (DDRs); and the master control node is used for managing each node in the CPU server node and the FPGA server computing node through a control layer. According to the scheme, the technical problem that in the prior art, an FPGA virtualization scheme oriented to deep learning reasoning application cannot be expanded to a distributed multi-node computing cluster is solved.

This invention discloses a method for realizing positioning based on a mobile station including the following steps: a mobile station starts a position request based on WAP, a positioning service customer processor receives the request and asks for position information to the terminal the mobile station starts the beginning of the positioning process request information, the mobile station and the mobile positioning center interact to finish authorization authentication and certification, the mobile station, positioning entity and mobile positioning center in interact to finish computation and acquirement of position information, the mobile station sends its position information to the position service customer processor, which provide service based on the position information to the mobile station.

The invention discloses a data consistency system based on a satellite network environment. The system comprises n gateway stations, wherein each gateway station comprises a meta-data server and is used for realizing meta-data synchronization by means of a chain-based high-consistency strategy through a transaction chain when the average speed of the chain is larger than or equal with a threshold,and otherwise, realizing meta-data synchronization by means of a chain-based low-consistency strategy through the transaction chain; a hot-standby meta-data server which is used for replacing the meta-data server for service when shutdown of the meta-data server occurs; and a data server which is used for supplying the chain-based high-consistency strategy for the writing operation of the data block in the data server when the average speed of the chain is larger than or equal with the threshold, and otherwise, supplying the chain-based low-consistency strategy. The data consistency system has advantages of ensuring high granularity consistency, realizing quick read-and-write response and reducing synchronous expenditure of repeated data.

The invention discloses a random access method and device, and the method comprises the steps that: in a random access process, a base station to which a satellite cell belongs sends a random access related message through a ground cell accessed by UE, and indicates a cell type corresponding to the random access related message, and the UE receives the random access related message sent by the ground cell; and a cell type corresponding to the random access related message is determined, wherein the cell type is a ground cell or a satellite cell. The invention provides the method for assisting the satellite cell to perform random access by the ground cell, the random access time delay of the satellite communication system can be shortened through the method, and the time delay overhead is reduced.

The invention relates to a program execution method and device, equipment and a storage medium, and the method comprises the steps: in a main program execution process, when a subprogram parameter is obtained, determining a to-be-executed subprogram and execution data corresponding to the to-be-executed subprogram based on the subprogram parameter; and controlling the to-be-executed subprogram to execute related logic based on the execution data. According to the method, the to-be-executed subprogram and the execution data corresponding to the to-be-executed subprogram are determined according to the subprogram parameters, the number of times of querying the subprogram is reduced, the to-be-executed subprogram is controlled to execute related logic according to the execution data, execution of multiple programs is changed into execution of one program, the execution efficiency is improved, and the execution time is shortened. And the number of times of network message interaction between the client and the server is reduced, so that the network delay overhead is reduced, and the processor overhead of the database is also reduced.