-
公开(公告)号：KR101285078B1
公开(公告)日：2013-07-17
申请号：KR1020090126035
申请日：2009-12-17
Applicant: 한국전자통신연구원
CPC classification number: G06F9/5027 , G06F15/17393 , G06F2209/5017
Abstract: 대용량의 데이터를 다수의 컴퓨팅 노드를 이용하여 MapReduce 방식으로 분산 병렬 처리하는 시스템으로서, 이미 수집되어 있는 대용량 저장 데이터는 물론 분산 병렬 처리 작업이 수행되는 동안에도 연속적으로 수집되는 대량의 스트림 데이터에 대해서 점진적인 MapReduce 기반 분산 병렬 처리 기능을 제공하기 위한 분산 병렬 처리 시스템이 제공된다.
분산 병렬 처리, 스트림 데이터, MapReduce, Incremental-
公开(公告)号：KR1020130010314A
公开(公告)日：2013-01-28
申请号：KR1020110071031
申请日：2011-07-18
Applicant: 한국전자통신연구원
Abstract: PURPOSE: An explosion data stream processing meth and an apparatus thereof are provided to maintain the accuracy of a process result and to process data without the degradation of performance by processing the data stream. CONSTITUTION: A service managing module(312) manages task information allocated to a service and request information including QoS(Quality of Service) information about a service. A QoS monitoring module(313) collects execution state information about the allocated task. If the service does not satisfy threshold throughput of the QoS, the QoS monitoring module distinguishes a task which decreases process performance of the service. A scheduling module(314) divides data stream inputted to the task. [Reference numerals] (311) Service managing unit; (312) Service managing module; (313) QoS monitoring module; (314) Scheduling module; (322) Task managing module; (323) Area monitoring module; (401) Task 1; (402) Dividing task; (4031) Duplicated task 2-A; (4032) Duplicated task 2-B; (4033) Duplicated task 2-C; (406) Merging task; (407) Task 3; (AA) User; (BB) Data stream processing service request; (DD) Task information including a performance problem; (EE) Allocation, duplication, dynamic configuration; (FF) Data stream binding, control; (GG) Task execution unit; (HH) Task control command; (II) Monitoring; (JJ) Task execution state information
Abstract translation: 目的：提供一种爆炸数据流处理技术及其装置，以保持处理结果的准确性，并通过处理数据流来处理数据而不降低性能。 构成：服务管理模块（312）管理分配给服务的任务信息，并且请求包括关于服务的QoS（服务质量）信息的信息。 QoS监控模块（313）收集关于所分配的任务的执行状态信息。 如果服务不满足QoS的阈值吞吐量，则QoS监控模块区分降低服务进程性能的任务。 调度模块（314）将输入的数据流划分为任务。 （附图标记）（311）服务管理单元; （312）服务管理模块; （313）QoS监控模块; （314）调度模块; （322）任务管理模块; （323）区域监控模块; （401）任务1; （402）分工任务; （4031）重复任务2-A; （4032）重复任务2-B; （4033）重复任务2-C; （406）合并任务; （407）任务3; （AA）用户; （BB）数据流处理服务请求; （DD）包括性能问题的任务信息; （EE）分配，复制，动态配置; （FF）数据流绑定，控制; （GG）任务执行单位; （HH）任务控制命令; （二）监测; （JJ）任务执行状态信息
-
公开(公告)号：KR1020120067886A
公开(公告)日：2012-06-26
申请号：KR1020100129527
申请日：2010-12-16
Applicant: 한국전자통신연구원
CPC classification number: H04L43/0852 , H04L41/5006 , H04L47/805 , G06F9/30079 , G06F9/30087 , G06F9/30094
Abstract: PURPOSE: Data processing method and apparatus are provided to satisfy quality of service(QoS) targets which are individualized according to each application service. CONSTITUTION: A data processing apparatus divides a plurality of operators based on each calculated maximum permissible delay time(320). The apparatus calculates QoS satisfaction margin with respect to the operators(330). The apparatus sets the execution sequence of the operators by reflecting the calculated QoS satisfaction margin(340). According to the execution sequence, the apparatus executes the operators(350).
Abstract translation: 目的：提供数据处理方法和设备，以满足根据每个应用服务个性化的服务质量（QoS）目标。 构成：数据处理装置基于每个计算出的最大允许延迟时间来划分多个操作者（320）。 该装置相对于运算符计算QoS满足余量（330）。 该装置通过反映所计算的QoS满足余量来设置运营商的执行顺序（340）。 根据执行顺序，装置执行运算符（350）。
-
公开(公告)号：KR1020100072777A
公开(公告)日：2010-07-01
申请号：KR1020080131285
申请日：2008-12-22
Applicant: 한국전자통신연구원
CPC classification number: G06F17/30961 , G06F17/30979
Abstract: PURPOSE: A system and a method of distribution index based on multi length signature file thereof are provided to configure independent signature file of different length in computing node of terminal in dispersion index tree according to data distribution and generate signature of bits much more in case of the size of the cluster is small, thereby enhancing filtering effect through effective search. CONSTITUTION: A specific vector extractor(130) extracts N-dimension specific vector from a multimedia object(100) and identifier. A high dimensional index unit(140) configures dispersion index based on tree according to an identifier of the multimedia object and the specific vector of N-dimension. A high dimensional index unit determines the length of a signature by comparing between the end node number of the dispersion index tree and the size of the standard cluster. A high dimensional index management unit(150) generates signature by end node which reflected the determined length.
Abstract translation: 目的：提供一种基于多长度签名文件的分布索引的系统和方法，根据数据分布配置分散索引树中终端计算节点的不同长度的独立签名文件，并在 集群的大小很小，从而通过有效的搜索增强了过滤效果。 构成：特定向量提取器（130）从多媒体对象（100）和标识符提取N维特定向量。 高维索引单元（140）根据多媒体对象的标识符和N维的特定矢量，基于树构成色散索引。 高维度索引单元通过比较分散索引树的结束节点号和标准簇的大小来确定签名的长度。 高维度索引管理单元（150）生成反映所确定的长度的末端节点的签名。
-
公开(公告)号：KR1020100071491A
公开(公告)日：2010-06-29
申请号：KR1020080130226
申请日：2008-12-19
Applicant: 한국전자통신연구원
Abstract: PURPOSE: A method for data management based on a cluster system and a system using the same are provided to supply a data duplication method of a high performance by using a distribution file system of a cluster system which provides a physical data duplication function. CONSTITUTION: A duplication partition server group(14-1 to 14-n) provides identical partitions simultaneously, and a master server(12) allocates a duplication partition server group. When a node error happens, the duplication partition server group recovers data to reconfigure the data in a memory. The duplication partition server group is designated as a first partition server which provides a search service and a modification service through a master server and a second partition server which provides the search service.
Abstract translation: 目的：提供一种基于集群系统的数据管理方法和使用该系统的系统的方法，通过使用提供物理数据复制功能的集群系统的分发文件系统来提供高性能的数据复制方法。 构成：复制分区服务器组（14-1至14-n）同时提供相同的分区，主服务器（12）分配复制分区服务器组。 当发生节点错误时，复制分区服务器组恢复数据以重新配置存储器中的数据。 复制分区服务器组被指定为通过主服务器提供搜索服务和修改服务的第一分区服务器，以及提供搜索服务的第二分区服务器。
-
Patent Agency Ranking
公开(公告)号：KR100946987B1
公开(公告)日：2010-03-15
申请号：KR1020070133589
申请日：2007-12-18
Applicant: 한국전자통신연구원
Abstract: 본 발명은 분산 병렬 처리 시스템의 다중 Map 태스크 중간 결과 정렬 및 결합 장치 및 방법에 대하여 개시한다. 본 발명은 입력 데이터를 맵(Map) 함수에 의하여 처리가능한 소정크기의 데이터로 분할하는 단계; 상기 소정크기의 데이터를 각각 하나 이상의 맵(Map) 태스크에 입력하고, 상기 맵 함수를 적용하여 하나 이상의 키/값 쌍을 각각 추출하는 단계; 각각 추출된 상기 하나 이상의 키/값 쌍에 리듀스(Reduce) 태스크를 적용하여 태스크 수준에서 중복 키를 제거한 하나 이상의 제1 중간 결과를 생성하는 단계; 및 상기 소정크기의 데이터에 대한 상기 각 맵 태스크의 상기 맵 함수 처리가 완료되면, 상기 하나 이상의 제1 중간 결과에 대한 중복 키를 제거하여 하나의 노드에서는 중복 키가 포함되지 않는 하나 이상의 제2 중간 결과를 생성하는 단계를 포함하는 점에 그 특징이 있다.
MapReduce, Map 태스크, Reduce 태스크, 분산 병렬 처리, 결합기-
公开(公告)号：KR100931310B1
公开(公告)日：2009-12-11
申请号：KR1020070123489
申请日：2007-11-30
Applicant: 한국전자통신연구원
Abstract: A device and a method for processing an integrated continuous query using a user-defined shared extension trigger are provided to improve a query processing ability quality to reduce a query processing time by registering a query for a user-defined shared trigger result used for each integrated continuous query as the user-defined shared external trigger result and searching whole user-defined shared extension trigger result without searching the user-defined shared trigger result. An integrated continuous query manager(12) informs a trigger manager(13) that an integrated continuous query includes a query for a user-defined shared trigger result. The trigger manager registers a user-defined shared extension trigger for the integrated continuous query according to the information informed from the integrated continuous query manager. A trigger result manager(14) manages the user defined shared extension trigger result set by the trigger manager as a set. An integrated continuous query processor(15) processes the integrated continuous query by using a data stream received from a data stream manager(11) and the trigger result set received from the trigger result manager.
Abstract translation: 提供了一种用于使用用户定义的共享扩展触发器来处理集成连续查询的设备和方法，以通过注册用于每个集成的用户定义的共享触发结果的查询来提高查询处理能力质量以减少查询处理时间 连续查询为用户自定义的共享外部触发结果，并在不查询用户自定义的共享触发结果的情况下搜索整个用户自定义的共享扩展触发结果。 集成连续查询管理器（12）通知触发管理器（13）集成连续查询包括对用户定义的共享触发结果的查询。 触发管理器根据集成连续查询管理器通知的信息为集成连续查询注册用户定义的共享扩展触发器。 触发结果管理器（14）将由触发管理器设置的用户定义的共享扩展触发结果集作为一组进行管理。 集成的连续查询处理器（15）通过使用从数据流管理器（11）接收的数据流和从触发结果管理器接收的触发结果集来处理集成的连续查询。
-
公开(公告)号：KR1020090065134A
公开(公告)日：2009-06-22
申请号：KR1020070132586
申请日：2007-12-17
Applicant: 한국전자통신연구원
IPC: G06F15/16 , H04L12/865
Abstract: A distribution parallel task processing system and a method thereof are provided to efficiently revent intermediate result transmission requests from being concentrated on a specific map task performing device and have log time complexity, thereby improving performance. A plurality of reduce task performing devices(131~133) receives notification of intermediate result path information from a task management device(120). The reduce task performing devices dividedly allocate transmission planned intermediate results to a priority queue per an area according to an identifier order of a map task. The reduce task performing devices select the transmission planned intermediate results of the areas of the priority queue according to priority.
Abstract translation: 提供分布并行任务处理系统及其方法，以有效地将中间结果传输请求集中在特定地图任务执行装置上并具有日志时间复杂度，从而提高性能。 多个减少任务执行装置（131〜133）从任务管理装置（120）接收中间结果路径信息的通知。 执行设备的减少任务根据地图任务的标识符顺序将传输计划的中间结果分配给每个区域的优先级队列。 执行设备的减少任务根据优先级选择优先级队列的区域的传输计划中间结果。
-
公开(公告)号：KR1020090065130A
公开(公告)日：2009-06-22
申请号：KR1020070132582
申请日：2007-12-17
Applicant: 한국전자통신연구원
IPC: G06F17/30
CPC classification number: G06F17/3002
Abstract: A method for indexing and searching high dimensional data by using a signature file and a system thereof are provided to enhance a signature about a query feature vector and use the enhanced signature in searching, thereby increasing accuracy of the searching. Index generation related information including a feature vector of high dimensional data, an object identifier, a step signature file identifier and a two step signature file identifier is inputted(S401). A feature vector file including the feature vector and the object identifier is generated(S402). A two step signature is obtained by using 1 step signature information and feature vector(S405). The 2 step signature is stored in a 2 step signature file(S406).
Abstract translation: 提供一种通过使用签名文件及其系统来索引和搜索高维数据的方法，以增强关于查询特征向量的签名，并在搜索中使用增强签名，从而提高搜索的准确性。 输入包括高维数据的特征向量，对象标识符，步签名文件标识符和两步签名文件标识符的索引生成相关信息（S401）。 生成包括特征向量和对象标识符的特征向量文件（S402）。 通过使用1步签名信息和特征向量来获得两步签名（S405）。 2步签名存储在2步签名文件中（S406）。
-
公开(公告)号：KR1020080052344A
公开(公告)日：2008-06-11
申请号：KR1020070100556
申请日：2007-10-05
Applicant: 한국전자통신연구원
IPC: H04N19/139 , H04N19/51
CPC classification number: H04N19/139 , H04N19/132 , H04N19/176 , H04N19/43 , H04N19/44 , H04N19/577 , H04N19/61
Abstract: An apparatus for controlling the skip of motion compensation by using a motion vector characteristic in a video decoder and a method therefor are provided to minimize a data transmission time and system power consumption by determining whether to activate a motion compensation device according to the motion vector characteristic and skipping unnecessary motion compensation. A partition information input unit(111) receives partition information including a motion vector for respective macroblock partitions which are image decoding objects. A preprocessing unit(112) calculates a reference picture region indicated by a corresponding motion vector by using the partition information per the macroblock partition, and confirms whether the corresponding motion vector indicates an integral pixel. A skip control unit(113) provides pixel values of a motion compensation reference picture region based on a basic reference picture region to a motion compensation device, or skips the motion compensation device and provides the pixel values to a picture reconfiguring device according to whether the corresponding motion vector indicates the integral pixel.
Abstract translation: 提供一种通过使用视频解码器中的运动矢量特征来控制运动补偿的跳过的装置及其方法，用于通过根据运动矢量特征确定是否激活运动补偿装置来最小化数据传输时间和系统功耗 并跳过不必要的运动补偿。 分区信息输入单元（111）接收包括作为图像解码对象的各宏块分区的运动矢量的分区信息。 预处理单元（112）通过使用每个宏块分区的分区信息来计算由对应的运动矢量指示的参考图像区域，并且确认相应的运动矢量是否表示整数像素。 跳过控制单元（113）将基于基准参考图像区域的运动补偿参考图像区域的像素值提供给运动补偿装置，或者跳过运动补偿装置，并且根据图像重构装置是否将像素值提供给图像重新配置装置 相应的运动矢量表示积分像素。
-
-
-
-
-
-
-
-
-
Search Results
149
records
(took 0.034 seconds)
