Abstract: In a log management method performed by a server, the server receives a transaction and generates a command log of the transaction. When detecting the transaction is a multi-partition transaction or a non-deterministic transaction, the server generates a data log of the transaction. When the server is faulty, the server recovers data according to the command log or the data log.

Abstract: A method for data processing in a neural network system and a neural network system are provided. The method includes: inputting training data into a neural network system to obtain first output data, and adjusting, based on a deviation between the first output data and target output data, a weight value stored in at least one in-memory computing unit in some neural network arrays in a plurality of neural network arrays in the neural network system using parallel acceleration. The some neural network arrays are configured to implement computing of some neural network layers in the neural network system. The method may improve performance and recognition accuracy of the neural network system.

Abstract: A data table partitioning management method and apparatus are disclosed. The method includes: determining a type and a join key of each data table in a table group, where the type of the data table includes a one-dimensional table, a multidimensional table, or a fact table; and performing one-dimensional partitioning on row replica space of each data table in the table group, and performing one-dimensional or multidimensional partitioning on column replica space of the data table according to the type of the data table and based on the join key of the data table. Different partitioning management methods are applied to data tables of different types and different dimensions, so that data processing mechanisms of OLTP and OLAP are efficiently implemented in a system, and resource consumption is reduced.

Abstract: A neural network system includes P computing units configured to perform an operation of a first neural network layer and Q computing units configured to perform an operation of a second neural network layer. The P computing units are configured to perform computing on first input data based on N configured first weights to obtain first output data after receiving the first input data. The Q computing units are configured to perform computing on second input data based on M configured second weights to obtain second output data after receiving the second input data. The second input data includes the first output data. A ratio of N to M corresponds to a ratio of a data volume of the first output data to a data volume of the second output data.

Abstract: A method for partitioning an association table in a distributed database, where a manager determines a first data table in data tables requiring partition and generates a colocation partition (CP) table set of the first data table, the CP table set of the first data table includes the first data table and at least one CP table of the first data table, and a CP table of the first data table includes a data table whose partition key includes a subset of a partition key of the first data table. The manager partitions the first data table according to the partition key and partitions each CP table in the CP table set, a partition range of a partition key of each CP table is the same as a partition range of a corresponding partition key in the first data table.

Abstract: In a log management method performed by a server, the server receives a transaction and generates a command log of the transaction. When detecting the transaction is a multi-partition transaction or a non-deterministic transaction, the server generates a data log of the transaction. When the server is faulty, the server recovers data according to the command log or the data log.

Abstract: An index table update method, and a device are presented. The method includes adjusting an index density of a second index table in a memory when determining that storage space occupied by the second index table reaches an upper limit or a lower limit of a preset range, where the index density of the second index table is a ratio of a quantity of index records included in the second index table to a quantity of index records included in a first index table in an external storage; and selecting an index record from the first index table according to the adjusted index density, and updating the second index table using the selected index record. In this way, the second index table is dynamically adjusted according to a storage space occupation status of the memory, and system performance and retrieval efficiency are further ensured.

Abstract: A data table partitioning management method and apparatus are disclosed. The method includes: determining a type and a join key of each data table in a table group, where the type of the data table includes a one-dimensional table, a multidimensional table, or a fact table; and performing one-dimensional partitioning on row replica space of each data table in the table group, and performing one-dimensional or multidimensional partitioning on column replica space of the data table according to the type of the data table and based on the join key of the data table. Different partitioning management methods are applied to data tables of different types and different dimensions, so that data processing mechanisms of OLTP and OLAP are efficiently implemented in a system, and resource consumption is reduced.

Abstract: A method for partitioning an association table in a distributed database, where a manager determines a first data table in data tables requiring partition and generates a colocation partition (CP) table set of the first data table, the CP table set of the first data table includes the first data table and at least one CP table of the first data table, and a CP table of the first data table includes a data table whose partition key includes a subset of a partition key of the first data table. The manager partitions the first data table according to the partition key and partitions each CP table in the CP table set, a partition range of a partition key of each CP table is the same as a partition range of a corresponding partition key in the first data table.

Abstract: An index table update method, and a device are presented. The method includes adjusting an index density of a second index table in a memory when determining that storage space occupied by the second index table reaches an upper limit or a lower limit of a preset range, where the index density of the second index table is a ratio of a quantity of index records included in the second index table to a quantity of index records included in a first index table in an external storage; and selecting an index record from the first index table according to the adjusted index density, and updating the second index table using the selected index record. In this way, the second index table is dynamically adjusted according to a storage space occupation status of the memory, and system performance and retrieval efficiency are further ensured.

Abstract: A P2P based method for playing media is provided. The method includes logging in a network; selecting a live broadcasting service or a video-on-demand service, and receiving a judgment result on whether to buffer a data flow corresponding to the live broadcasting service or the video-on-demand service; obtaining the corresponding data flow according to a selection result, and buffering the data flow when the judgment result is yes; and providing the buffered corresponding data flow for a video-on-demand node or a live broadcasting node as a data source of the live broadcasting service or the video-on-demand service.

Abstract: An automatic expansion method, a management device, and a management system are disclosed in the embodiments of the present invention. In the embodiments of the present invention, information about added management nodes is received from a node management server when a new node is added, where the information about the added management nodes comprises at least address information of the added management nodes; the address information of the added management nodes is used to obtain an index corresponding to a continuous identifier (ID) segment; and an adding message of the new node is sent to all added management nodes, where the adding message includes at least a continuous ID segment managed by the new node and address information of the new node.

Abstract: A distributed network construction method includes: determining a domain that covers a requesting node according to an address of the requesting node which requests to join a distributed network, where the domain has a domain identifier (ID); generating a peer ID of the requesting node according to the number of peers of the domain; concatenating the domain ID with the peer ID to form the node ID of the requesting node; and sending the node ID to the requesting node, where the requesting node obtains information about a neighboring node according to the obtained node ID to join the distributed network.

Abstract: In the field of communications, a method, a system, and a device for searching for information, and a method for registering a vertical search engine are provided. The method for searching for information includes the following steps. A first search request sent by a user is received. A logic node is determined according to attribute information in the first search request. A vertical search engine selected by the logic node for the user is received. A second search request is sent to the vertical search engine. The method for registering the vertical search engine includes the following steps. A mobile search server receives and saves first registration information sent by the vertical search engine. Location information of a logic node corresponding to the vertical search engine is determined according to the first registration information. Second registration information is sent to the logic node according to the location information.

Abstract: The present invention provides a data uploading method, a data downloading method, and a data system. The uploading method includes: receiving a data uploading request of a user and obtaining a content ID of data to be uploaded; determining, according to the content ID, whether the data to be uploaded is already stored; and if the data to be uploaded is not stored, uploading the data to be uploaded to a local data center and storing the data to be uploaded. According to the embodiments of the present invention, a data traffic load between different networks is reduced and response efficiency is increased; uniform management and quick query of content copies in different networks are realized, and the number of distribution of copies of the same content in the network in the system is reduced.

Abstract: An automatic expansion method, a management device, and a management system are disclosed in the embodiments of the present invention. In the embodiments of the present invention, information about added management nodes is received from a node management server when a new node is added, where the information about the added management nodes comprises at least address information of the added management nodes; the address information of the added management nodes is used to obtain an index corresponding to a continuous identifier (ID) segment; and an adding message of the new node is sent to all added management nodes, where the adding message includes at least a continuous ID segment managed by the new node and address information of the new node.

Abstract: A method for transmitting shared contents includes: pre-configuring rules for constructing coefficient vectors corresponding to the shared content on each content terminal; by the content source peer, dividing the original content into blocks, constructing coefficient vectors corresponding to the original content according to the rules for constructing coefficient vectors, and combining each original content block with the corresponding coefficient vector linearly to obtain code blocks, with the ID of each code block same as the ID of the corresponding coefficient vector; and by the content receiving peer, downloading the code blocks from the content source peer according to the IDs of code blocks. A system for transmitting shared contents includes: a content source peer and a content receiving peer. The present invention also discloses a content terminal. The present invention may save the transmission bandwidth and improve the content sharing efficiency.

Abstract: In the field of communications, a method, a system, and a device for searching for information, and a method for registering a vertical search engine are provided. The method for searching for information includes the following steps. A first search request sent by a user is received. A logic node is determined according to attribute information in the first search request. A vertical search engine selected by the logic node for the user is received. A second search request is sent to the vertical search engine. The method for registering the vertical search engine includes the following steps. A mobile search server receives and saves first registration information sent by the vertical search engine. Location information of a logic node corresponding to the vertical search engine is determined according to the first registration information. Second registration information is sent to the logic node according to the location information.

Abstract: A distributed network construction method includes: determining a domain that covers a requesting node according to an address of the requesting node which requests to join a distributed network, where the domain has a domain identifier (ID); generating a peer ID of the requesting node according to the number of peers of the domain; concatenating the domain ID with the peer ID to form the node ID of the requesting node; and sending the node ID to the requesting node, where the requesting node obtains information about a neighboring node according to the obtained node ID to join the distributed network.

Abstract: A method for implementing parallel download applicable to a system supporting the download of chunked download content includes: acquiring download source information corresponding to download content, wherein the download source information comprises multiple content sources providing the download content, and the multiple content sources corresponds to multiple download servers respectively; dividing the download content into download content chunks according to the download source information; sending a download content chunk request to the multiple download servers corresponding to the multiple content sources respectively, and downloading the download content chunks from the multiple download servers in parallel. A download agent for implementing parallel download is provided. The method and download agent of the embodiments of the present invention may improve the download speed and robustness effectively. Therefore, object files with a large amount of data may be downloaded reliably and rapidly.