78results about How to "Eliminate performance bottlenecks" patented technology

The invention discloses a method and a system for managing data of a cluster storage system and belongs to the technical field of cluster storage. The method includes that a disk file system is used for storing and managing meta data, and the data are distributed by using a catalogue as a basic unit; a client senses data distribution information and sends a data access instruction, a storage server judges whether a finding path is a root catalogue according to the data access instruction of the client, a specific active node is selected to be a target node for data access if the finding path is the root catalogue, otherwise, a Hash value is calculated according to path input parameters to obtain the data distribution information, and finding the target node for data access. According to the method and the system for managing the data of the cluster storage system, a series of relevant problems of performance bottlenecks, single point faults, data inconsistency and the like of meta data management are eliminated, system expansibility is obviously improved, and linear expansion increase of system concurrency and performances is achieved.

The invention discloses a method and a system for controlling data storage in a cluster file system and a method for creating a file. The cluster file system stores files by grouping storage nodes. The method for data storage comprises the following steps of: collecting and measuring binary load information containing I / O access load space utilization rate of each storage node in the cluster file system during interval period; measuring whether the load of the cluster file system is balanced or not according to the binary load information of each storage node; and migrating data of in cluster file system if the load of the cluster file system is not balanced.

The invention relates to a system and a method for mobility management under a separate mapping mechanism. A proxy mobile Internet protocol version 6 (PMIPv6) and the separate mapping mechanism are combined. In an access network of a separate mapping network, the PMIPv6 is greatly improved by the scheme, and three functions comprising an access gateway, a control gateway and a data gateway are designed, thereby realizing the separation of data forwarding and control signaling and the separation of a host identifier and a host position. The separation of the access network and a core network is realized in combination with the separate mapping mechanism.

The invention discloses a container-based server resource supply method. The container-based server resource supply method comprises the following steps: diagnosing and evaluating performances of servers through performance anomaly detection, finding out an idle service node and a performance anomaly node, and then, realizing self-adaptive adjustment through detection, feedback and control. According to the invention, relative monitoring data of the servers and a container is collected and subjected to idle judgement based on a threshold value and performance analysis for abnormal performance diagnosis based on service times; therefore, the states of each server and each container cluster node can be obtained; then, various clusters are adjusted; supply of server resources can be ensured according to requirements; the use amount of the servers can be reduced to the most extent; the resource consumption is reduced; and the operating cost of a data centre is reduced.

The invention discloses a data distribution and mutual conversion method of a storage system. A continuous storage space is formed by a plurality of disks, the disks also form a disk array by strips to partition the disk array into continuous storage subspaces, namely sub array, and then redundancy level can be respectively configured as zero, one, or five according to requirements; during conversion, corresponding modes are selected according to differences of transversion types. The invention has the major concept of adopting suitable data distribution mode for I / O requirements with different inquiry characteristics according to partial characteristic of I / O requirements, thereby enabling the storage system to realize optimal performance. The storage system data distribution conversion method disclosed by the invention facilitates the system being dynamically suitable for changes of I / O requirement characteristic, reduces cost caused by data distribution conversion among different redundancy levels and eliminates performance bottleneck generated by data transfer, thereby greatly improving performance of storage subsystem.

A system of a distributed database concurrence storage virtual request mechanism comprises a client for sending storage requests, a web application server, an access control server and a distributed database cluster. Based on the system, a distributed database storage management mechanism scheme is put forward, the mechanism particularly considers the flow request required by the number of concurrence users to reasonably store the user data flow in a distributed manner, that is, under the framework of the cloud database, the user data flow is distributed to the most reasonable database cluster for storage according to the current network load and the current database concurrence connected load. The mechanism realizes management of the distributed database cluster by utilizing the access control server, and an ACK (acknowledgement character) information packet fed back by the database cluster calculates the optimized storage strategy and completes final data storage through a corresponding flexible algorithm.

A multiplier circuit comprises a fused Booth encoder multiplexer which produces partial product bits, a tree which uses the partial product bits to generate partial products, and an adder which uses the partial products to generate intermediate sum and carry results for a multiplication operation. The fused Booth encoder multiplexer utilizes encoder-selector cells having a logic tree which carries out a Boolean function according to a Booth encoding and selection algorithm to produce one of the partial product bits at a dynamic node, and a latch connected to the dynamic node which maintains the value at an output node. The encoder-selector cells operate in parallel to produce the partial product bits generally simultaneously. A given one of the encoder-selector cells has a unique set of both multiplier operand inputs and multiplicand operand inputs, and produces a single partial product bit. The fused Booth encoder multiplexer unit, tree unit and adder unit function in a pipeline manner with the units operating on sequential data sets during a given processing cycle. The fused Booth encoder multiplexer unit may be advantageously laid out in a design of an integrated circuit chip with no gap present in the layout, which allows uniform wire length and avoids the necessity of large transistors to drive long interconnection wires.

The invention discloses an implementation method for hardware-assisted dynamic priority scheduling of slave core tasks. The method comprises the following steps: architectural support is provided for scheduling by an operating system based on a main core plus slave core system so as to assist the operating system scheduler to schedule the slave core tasks, a hardware scheduler maintains hardware queues of three ready tasks, arrived tasks are inserted into the corresponding ready task queues according to a task priority, and the circumstance that the slave core task is starved is avoided by periodic priority promotion. The method effectively reduces the difficult in supporting an embedded heterogeneous multi-core system by a software system through the architectural support, lightens the burden of the main core scheduling and the slave core tasks management, and obviously increases the throughput of the slave core tasks. The method has more obvious effects under the conditions of numerous types and a great number of slave cores and complex scheduling algorithms.

The invention discloses a file transmission method for realizing a point-to-point file transmission mechanism. The method is characterized by comprising the following steps that: a client makes a request for downloading a specified file to a fixed server end; and a fixed server orders the client to download the specified file from the fixed server end or a temporary server, wherein the temporary server refers to another client which was downloaded as the client, stores the specified file and can provide a file downloading service for the client; and the fixed server stores a temporary server list in which the Internet protocol addresses of other clients forming the temporary server are recorded.

The invention provides a shared data updating device. The device comprises a storage unit, a discrepant data computing unit, a data persistence unit and a data verification unit, wherein the storage unit is used for saving a basic data base table and a cache data base table; the basic data base table is used for saving summarized data of shared data; the cache data base table is used for saving discrepant data of the shared data; the discrepant data calculating unit is used for computing discrepant data of the shared data caused by updating operation; the data persistence unit adopts isolated transaction to insert the discrepant data into the cache data base table; and the data verification unit is used for inquiring the discrepant data of the basic data base table and the cache data base table and determining whether to submit isolated transaction according to a verification result. The invention further provides a shared data updating method. According to the technical scheme, locking process can be eliminated during data updating and verification processes on the basis that the data updating correctness is guaranteed, so that the concurrency supported by a system is improved.

The invention discloses a DRM controller, which comprises a client which communicates with a server and is installed on a mobile terminal, and a security device capable of communicating with the client. The DRM controller is characterized in that the client is connected to the security device via interfaces, and the security device is a key type device provided with a function of an IC (integrated circuit) card. By the method, resource requirement of hardware is lowered, performance bottleneck is eliminated, and convenience is brought to software update and maintenance. Meanwhile, by integrally combining a mobile payment card with a CMMB (China mobile multimedia broadcasting) card, existing infrastructure is fully utilized, seamless integration with mobile payment is realized, and accordingly, the DRM system based on the distributed key is convenient in implementation and use. Moreover, the security device can support online DRM system by the aid of low-cost communicator, so that uniformity of mobile network and Internet is realized, implementation and management cost is lowered greatly, and evident quality and price advantages are achieved.

The invention discloses a cloud testing environment scheduling method based on load forecast. The cloud testing environment scheduling method based on the load forecast comprises the steps that a load changing database of testing host machines is established, and selection strategies of physical nodes are configured; the average load of all the physical nodes is obtained; according to the average load and a low threshold value, the testing host machines needing to be migrated on the physical nodes are judged; according to the load changing database, loads of the physical nodes are forecasted, and forecasting results are fed back; according to the forecasting results and the node selection strategies, the testing host machines needing to be scheduled are migrated to the selected physical nodes. According to the cloud testing environment scheduling method based on the load forecast, through the forecasting of the loads of all the physical nodes in the cloud testing environment, the most suitable target node of the cloud testing host machines needing to be scheduled is found, the dual targets of load balance and energy conservation of the cloud testing environment are achieved, the performance bottleneck of the cloud testing environment is eliminated, in this way, the stability of the whole cloud testing environment is improved, and the energy consumption of the system is lowered. The invention further discloses a cloud testing environment scheduling system based on the load forecast.

The invention discloses a data service network system as well as an access method of data services, and the system comprises: an access gateway, a service gateway, a first border router which is connected with the access gateway for converging and classifying a user data flow and transmitting the user data to a second border router via virtual private network; and the second border router which is connected with the service gateway and connected with the first border router via the virtual private network for transmitting the user data to the service gateway. In the technical proposal, the service flow does not pass through a GRE tunnel for encapsulation and decapsulation due to the introduction of VPN in IP bearer network, thereby breaking through limitations of GRE encapsulation efficiency and bandwidth and eliminating the performance bottleneck problem of the whole network brought by NAT conversion.

The invention discloses an index caching method and system based on an off-heap memory. The method comprises the steps that when Lucene is in a starting state, an off-heap cache index is preheated after a memory with a designated size is distributed for index data in the off-head memory and placed into a memory pool; when Lucene is in an indexing state, the index capacity of the off-heap memory is judged, and if the index capacity reaches a demand value, an output stream is opened in an off-heap memory index to write the index data; and when Lucene is in a searching state, whether the index data needing to be read currently exists in the off-heap memory index is judged, and if yes, an input stream is opened in the off-heap memory index to read the index data. The system corresponds to the method. Through the index caching method and system, a performance bottleneck brought by disk IO is greatly lowered, and data can be persisted; meanwhile, the situation that data real-time writing delay is caused when merging speed is lowered due to the disk IO bottleneck is avoided.

The invention discloses a data storage method, a data searching method, a related device and a system, relates to the technical field of communication and aims to reduce the problems of single-point failure and performance bottleneck. The data storage method comprises the steps of reading data by a query server; determining a mark of a data block belonging to the data by the query server according to a data partitioning strategy, wherein the data partitioning strategy is used for determining the mark of the data block belonging to the data; determining a query server mark set corresponding to the mark of the data block by the query server according to a data block storage allocation strategy, wherein the query server which is represented by a query server mark in the query server mark set is used for storing the data block which is represented by the mark of the data block, and the data block storage allocation strategy comprises a corresponding relation between the mark of the data block and the query server mark set; storing the data by the query server if the query server mark set comprises the mark of the query server; enabling the data not to be stored by the query server if the query server mark set does not comprise the mark of the query server.

The invention discloses a data storage method, device and system, computer equipment and a storage medium, and the method comprises the steps: processing a write-in request of to-be-stored data in a memory, and caching the to-be-stored data in the memory; when the data cached in the memory reaches a first preset threshold value, converting the to-be-stored data in the memory into a multi-layer matrix array with an index, and storing the multi-layer matrix array in a storage layer of a matrix storage structure preset in the nonvolatile storage equipment; and when the data in the nonvolatile storage device reaches a second preset threshold value, performing merging processing on the data stored in the nonvolatile storage device, and writing the data into the next layer of the merging tree structure. According to the inventionmethod, the read-write performance of the system can be ensured, and the space size of each layer is increased, so that the write-amplification performance cost of the system caused by the multilayer log structure tree structure is solved, and the system fluctuation and the pause delay are reduced.

The invention discloses a JPEG (Joint Photographic Experts Group) image decompression method, device and platform. The method includes the following steps that: an FPGA (Field Programmable Gate Array) end receives the encoding information of images to be decompressed sent by a CPU (Central Processing Unit) end; the FPGA end adaptively determines the number of the processing paths of Huffman decoding through using a preset JPEG image decoding algorithm and according to the number of the images to be decompressed; the FPGA end performs Huffman decoding operation in parallel on the plurality of paths of the data of the images to be decompressed according to the encoding information and the number of the processing paths of the Huffman decoding; the FPGA end adaptively determines the number of the processing paths of inverse discrete cosine transformation and RGB transformation according to the number of the processing paths of the Huffman decoding; and the FPGA end performs inverse discrete cosine transformation and RGB transformation in parallel on the plurality of paths of the data of the images to be decompressed which have been subjected to the Huffman decoding according to the determined number of the processing paths, so that decompressed image data are obtained. Thus, with the JPEG image decompression method, device and platform of the invention adopted, the bottleneck of decompression performance can be eliminated, and decompression efficiency can be improved.

The invention provides a virtual volume access method and a virtualization system and relates to the field of virtualization. The virtual volume access method of the invention comprises following steps: a virtual machine sends a first request of accessing a virtual volume to a corresponding storage gateway; the storage gateway utilizes protocols of a virtualization platform to convert the first request into a second request and sends the second request to a resource pool; the resource pool sends feedback content to the storage gateway; the storage gateway utilizes the protocol of the first request to convert the feedback content protocol and then send it to the virtual machine. By means of the method, when a virtual machine accesses virtual volumes in different virtualization platforms, protocol conversion can be executed by the storage gateway of a virtualization platform where the resource pool is located so that cross-platform virtual volume access is realized; the storage gateways are positioned inside virtualization platforms, which reduces the conversion load of storage gateways and eliminates the network performance bottleneck of virtual volume access.

The invention relates to an index-based remote cloud data synchronizing method. Data occupying a larger space is stored in fewer servers in a cloud system, indexes of summarization information of all data are backed up on each server, a user judges whether cloud data summarization information which is interested by the user is stored in the system through searching the indexes, when the user needs to download resources in a cloud storage system, a server connected with the user firstly sends a synchronous request sent by the server storing the resources in the cloud system, and after synchronization is completed, the user downloads the resources from the server. According to the index-based remote cloud data synchronizing method, the cloud data in the cloud storage system is more conveniently downloaded by the user in each place, resource synchronization is blindly avoided being carried out by the cloud storage system, the quantity of the servers during cloud storage is greatly reduced, system building cost is lowered, one-point failures and property bottleneck caused by the servers in the system are reduced, data transmission links are reduced, system performance property and efficiency are improved, and efficient and stable operation of the whole system is ensured.

The related multistage distributed network system comprises: a full SP / CP business server, a central business platform connected to last device to provide business access point for mobile terminal, control the access for user and server, distribute business content, position provider and generate charge menu; multiple boundary business platforms connected to last platform to receive and provide business content, and a WAP gateway connected to last two platforms to access user and provide business content to mobile terminal.

This invention discloses a method for organizing ring nodes in the broadcast-selection system of peer-to-peer network flow media including: 1, a newly added node P gets an initial member node list from an index server, 2, P utilizes the initial member node to set up its adjacent nodes and distributes them to the corresponding [l d l / alpha] layer of ring based on the distance from the adjacent node to itself so as to constitute an initial logic ring and form far and near hop neighbor nodes, 3, node P transmits position information to a neighbor node on the ring to inform its existence and receives the forward information from the neighbor node to discover new member nodes to ensure there are k nerghbors in the inner ring and 1 neighbor in the outer ring to maintain a stable ring structure and users can quickly find out the neighbor nodes near the target position when dragging, 4, node P stops the current channel and informs all neighbor nodes on the ring about its leaving.

The invention discloses a distributed-data management method and belongs to the field of distributed storage. The technical scheme includes that scattering metadata management functions to multiple nodes, creating mapping relations, selecting an implementation structure that a common multi-fork tree and a hash table are combined in metadata tree structure design, adding full path fields to the data structure of metadata, and writing new data prior to logs in data writing; after file metadata is acquired, acquiring the storage location of file data in a mapping mode; meanwhile, supporting file metadata copies and file data copies, adopting the different mapping relations to ensure that the copies are distributed to the different nodes when the file metadata copies or the file data copies are used, and enabling a copy storage terminal to take over a main storage end in operation rapidly when the file metadata or the file data is written and the main storage end fails. By the arrangement, performance and reliability of distributed data management can be effectively improved.

The invention provides a method for a semantic switch loose coupling system to process information. The method comprises the steps that a service pool is arranged at each service processing node, and each service pool is a semantic switch cluster used for finishing the same semantic operation and composed of n semantic switches with the same function; a runtime database, a configuration sub-module and a semantic processing program are configured for each semantic switch, wherein the semantic processing program is the program compiled by a user according to the client service logic, and the semantic switches can have different service processing functions through compiling of different programs; a non-relation database cluster is adopted in each runtime database and used for storing data records with the JSON format, and each data record has a unique record ID; the record IDs serve as marks of a URI and can be accessed; body descriptions and relation descriptions are defined in the configuration sub-modules. The method has the advantages that a flue model of a traditional information system is eliminated, and data sharing of a multi-platform heterogeneous system is achieved.

To reduce FIFO access cycles across a system bus in a multi-processor system in which two processors communicate across a system bus through a FIFO, two separate FIFO descriptors are provided. The first descriptor is maintained by the processor located on-board with the FIFO, and the second descriptor is maintained by an off-board processor which communicates with the FIFO across the bus. When one processor performs a FIFO operation, the processor updates the other processor's descriptor via a memory access across the bus. Additionally, one module passes credits to the other to indicate that the latter has permission to perform a plurality of FIFO operations consecutively. In one embodiment a special non-valid data value is used to indicate an empty FIFO position.

A multiplier circuit comprises a fused Booth encoder multiplexer which produces partial product bits, a tree which uses the partial product bits to generate partial products, and an adder which uses the partial products to generate intermediate sum and carry results for a multiplication operation. The fused Booth encoder multiplexer utilizes encoder-selector cells having a logic tree which carries out a Boolean function according to a Booth encoding and selection algorithm to produce one of the partial product bits at a dynamic node, and a latch connected to the dynamic node which maintains the value at an output node. The encoder-selector cells operate in parallel to produce the partial product bits generally simultaneously. A given one of the encoder-selector cells has a unique set of both multiplier operand inputs and multiplicand operand inputs, and produces a single partial product bit.

The embodiment of the invention provides a service logic processing method and device, a service server and a system. The method comprises the steps that whether or not a first completion message usedfor indicating that a first service logic is executed exists in a first message queue is monitored; if it is monitored that the first completion message exists in the first message queue, the first completion message is extracted from the first message queue, and a second service logic associated with the first service logic is executed. Through the method, the message indicating that the servicelogic is completed is stored in the message queue, subsequent associated service logics are executed by monitoring the message queue, and then the execution of the whole service process is promoted.The pushing of the service process does not need updating and reading of a database, so that the situation that the database is frequently updated and read due to state changes of the service processis avoided. Therefore, under the condition of high concurrency, the I / O performance bottleneck of the database can be reduced to a great extent, and the system processing capability is improved.

The invention provides a digital video monitoring system which is characterized in that an encoding collection unit collects video and audio data in a monitoring site and serves as a client of a network file system to store the collected video and audio data to a memory cell serving as a server terminal of the network file system; a decoding play unit serves as a client of the network file system, and directly reads and plays video and audio data required to be played from the memory cell; and a center management unit manages and controls the encoding collection unit, the memory cell and the decoding play unit through a hypertext transfer protocol. According to the system, the encoding collection unit can be installed in any networking place, so that the problems of performance bottleneck caused by the center management unit, as well as single point of failure caused by center management unit failure are solved, and network flow brought by video and audio transmission are greatly reduced. Therefore, 'memory' and 'center management' can be separated in terms of region and manage sovereignty.

The invention relates to a URI-based Internet of Things identifying and servicing method. A URI-based Internet of Things identification comprises a proxy server identification and an item object identification. According to the URI-based Internet of Things identifying and servicing method, firstly, a two-dimensional code of the URI-based Internet of Things identification is scanned by a client app, and a web request is initiated to access a proxy server; the proxy server calls an Internet of Things identifying service function module to communicate with a national Internet of Things identification management public service platform so as to initiate an Internet of Things identifying service process; after acquiring information fed back from the national Internet of Things identification management public service platform, the proxy server initiates web redirection and feeds the web redirection back to the client app; and the client app loads a content of a redirected URI page and displays the content to a user. In the method, the Internet of Things item identification is named in a URI form, and the Internet of Things identifying service process is initiated by establishing the proxy server to ensure that a client can enjoy the convenience brought by an Internet of Things identifying service without any changes.

The invention provides a semantic switch. Distributed type computing architecture based on REST non-state transmission is adopted, and the semantic switch works on a semantic layer above an application layer protocol. The semantic switch comprises a receiving end network / transmission layer data processing module, an application layer protocol analyzing module, a receiving end semantic preprocessing module, a semantic processing engine, a sending end semantic preprocessing module, an application layer protocol packaging module, a sending end network / transmission layer data processing module, a runtime database and a log audit module. The semantic processing engine comprises a semantic analyzing sub-module, a semantic firewall sub-module, a semantic processing sub-module, a semantic processing program, a load balancing sub-module and a configuration sub-module, and a body description and a relation description are defined in the configuration sub-module. A non-relation database cluster is adopted in the runtime database, and the semantic processing program is the program compiled by a user according to the client service logic. The semantic switch has the advantages that a flue model of a traditional information system is eliminated, and data sharing of a multi-platform heterogeneous system is achieved.

The invention provides a data center intelligent monitoring system and method based on a distributed architecture and the system and the method are used for solving a performance bottleneck problem ofa centralized architecture intelligent monitoring system and improving the reliability and safety of intelligent monitoring system service. The data center intelligent monitoring system based on thedistributed architecture comprises a data access unit, a real-time calculation unit, a data storage unit, an application function unit and a distributed coordination processing unit. The system is characterized in that the data access unit is used for acquiring monitoring data from a monitored system and converting the monitoring data into data in a preset format, the real-time calculation unit isused for judging whether alarm information is generated or not according to the monitoring data, the data storage unit is used for storing the configuration data, real-time monitoring data and historical monitoring data acquired by the data access unit and alarm data generated by the real-time calculation unit, and the application function unit is used for receiving a data display request submitted by a client, obtaining display data from the data storage unit according to the data display request and sending the display data to the client.