Abstract: Disclosed is a training method for the optimization of a traffic signal. The method includes segmenting a large-scale road network into intersection groups each including a plurality of intersections, training a reinforcement learning model by allocating a training agent to each intersection group, performing traffic signal control through the inference of an optimal signal based on the trained reinforcement learning model, evaluating whether the results of the execution of the traffic signal control satisfy a preset goal of learning for traffic signal optimization, and repeatedly training the reinforcement learning model based on the results of the evaluation.

Abstract: A merge join method by a merge join system including a database is provided. The merge join system unequally divides data after a current position of a join column participating in merge join in the database into K partitions, K being greater than 2, and compares the first value of each of remaining partitions excluding the first partition from the K partitions with a value to be searched for the merge join, in order to check whether the first value and the value to be searched satisfy a specific condition. Further, the merge join system determines a partition in which the value to be searched is expected to exist based on a comparison result, and applies a k-ary search to the determined partition.

Abstract: A merge join method by a merge join system including a database is provided. The merge join system unequally divides data after a current position of a join column participating in merge join in the database into K partitions, K being greater than 2, and compares the first value of each of remaining partitions excluding the first partition from the K partitions with a value to be searched for the merge join, in order to check whether the first value and the value to be searched satisfy a specific condition. Further, the merge join system determines a partition in which the value to be searched is expected to exist based on a comparison result, and applies a k-ary search to the determined partition.

Abstract: Disclosed herein are a method and an apparatus for processing a query based on a heterogeneous computing device. The method includes generating, by the apparatus for processing a query, an optimal query execution plan for processing the query using all of a plurality of computation resources included in a heterogeneous computation resource and dividing data corresponding to the query depending on a data division ration included in the query execution plan and allocating the divided data to each computation resource. Further, the divided data are each processed based on each computation resource.

Abstract: Disclosed are a system and a method for address translation for a flash memory device, and particularly, disclosed is a technology that is capable of efficiently performing address translation between a logical address provided to the outside of a flash memory and a physical address of an actual flash memory in managing the flash memory device. The system includes: a flash memory system writing a corresponding data page by allocating a physical address space when there is a request for writing a data page from storage clients, and performing address translation between a physical address and a logical address; and a logical address space formed between the flash memory system and the storage client to provide the logical address.

Abstract: Disclosed are a system and a method for address translation for a flash memory device, and particularly, disclosed is a technology that is capable of efficiently performing address translation between a logical address provided to the outside of a flash memory and a physical address of an actual flash memory in managing the flash memory device. The system includes: a flash memory system writing a corresponding data page by allocating a physical address space when there is a request for writing a data page from storage clients, and performing address translation between a physical address and a logical address; and a logical address space formed between the flash memory system and the storage client to provide the logical address.

Abstract: Disclosed is an apparatus for in-memory data management includes a hybrid memory including a plurality of types of memories with different characteristics and a storage engine. The storage engine rearranges data among the plurality of memories by the unit of a page by monitoring workloads for data stored in the memories and reconfigures a page layout by the unit of the page based on a data access characteristic of an application for each of pages constituting each of the plurality of memories.

Abstract: An apparatus of managing data according to the present invention includes a query processor, a page monitor, a page layout manager and a data storage manager. The query processor processes a user query. At the time of processing the user query, the page monitor collects accessed column information and selectivity information of accessed columns from the query processor and collects access page information from a data storage manager to create page monitoring information. The page layout manager creates page column group information by grouping columns adjacent to each other for each page at a predetermined time interval based on the page monitoring information. The data storage manager stores data in a main memory by reconfiguring a page based on the page column group information for a candidate page of which an access frequency is greater than a predetermined access frequency based on the page monitoring information.

Abstract: Disclosed are a system and a method for address translation for a flash memory device, and particularly, disclosed is a technology that is capable of efficiently performing address translation between a logical address provided to the outside of a flash memory and a physical address of an actual flash memory in managing the flash memory device. The system includes: a flash memory system writing a corresponding data page by allocating a physical address space when there is a request for writing a data page from storage clients, and performing address translation between a physical address and a logical address; and a logical address space formed between the flash memory system and the storage client to provide the logical address.

Abstract: Disclosed herein are an apparatus and method for managing a data stream distributed parallel processing service. The apparatus includes a service management unit, a Quality of Service (QoS) monitoring unit, and a scheduling unit. The service management unit registers a plurality of tasks constituting the data stream distributed parallel processing service. The QoS monitoring unit gathers information about the load of the plurality of tasks and information about the load of a plurality of nodes constituting a cluster which provides the data stream distributed parallel processing service. The scheduling unit arranges the plurality of tasks by distributing the plurality of tasks among the plurality of nodes based on the information about the load of the plurality of tasks and the information about the load of the plurality of nodes.

Abstract: The present invention relates to a resource allocation apparatus and method. The resource allocation apparatus includes a job information management unit for managing job characteristic information required to execute jobs input by a user. A resource form selection unit selects an initial resource allocation form required to execute each job, based on the job characteristic information. A resource allocation unit allocates resources required to execute the job based on the initial resource allocation form.

Abstract: Disclosed herein are an apparatus and method for managing a data stream distributed parallel processing service. The apparatus includes a service management unit, a Quality of Service (QoS) monitoring unit, and a scheduling unit. The service management unit registers a plurality of tasks constituting the data stream distributed parallel processing service. The QoS monitoring unit gathers information about the load of the plurality of tasks and information about the load of a plurality of nodes constituting a cluster which provides the data stream distributed parallel processing service. The scheduling unit arranges the plurality of tasks by distributing the plurality of tasks among the plurality of nodes based on the information about the load of the plurality of tasks and the information about the load of the plurality of nodes.

Abstract: Disclosed are service providing method and device, including: collecting execution state information about a plurality of tasks that constitute at least one service, and are dynamically distributed and arranged over a plurality of nodes; and performing scheduling based on the collected execution state information about the plurality of tasks, wherein each of the plurality of tasks has at least one input source and output source, and a unit of data to be processed for each input source and a data processing operation are defined by a user, and the scheduling is to delete at least a portion of data input into at least one task or to process the at least a portion of input data in at least one duplicate task by referring to the defined unit of data. In particular, the present invention may effectively provide a service of analyzing and processing large stream data in semi-real time.

Abstract: Disclosed is a data processing method and device according to the present invention, including: calculating a maximum allowable delay time of each of a plurality of operators that constitute a plurality of services and are executed according to a data processing request of a user; classifying the plurality of operators based on the calculated maximum allowable delay time; calculating a quality of service (QoS) satisfaction allowable level with respect to each of the plurality of operators, wherein the allowable level for QoS satisfaction indicates the number of times that there is no need to satisfy one goal of QoS with the premise that there is no problem in satisfying QoS satisfaction requested by the user; setting execution orders of the plurality of classified operators by reflecting the calculated allowable level for QoS satisfaction; and executing the plurality of operators based on the set execution orders.

Abstract: Provided are a content-based searching method and system for multimedia objects using a high-dimensional feature vector data based on a 2-level signature. The method for searching the high-dimensional data using a signature file includes calculating a first-level query signature and a second-level query signature by using the query feature vector, performing a first filtering operation to obtain a primary candidate cell group by searching a second-level signature file, and performing a secondary filtering operation to obtain a secondary candidate cell group having a high similarity in a primary candidate cell group. Accordingly, the high-dimensional data searching method and system can process a query quickly and accurately and can increase the searching accuracy by using an enhanced signature of the query feature vector.

Abstract: Disclosed are a column-based data managing method and apparatus, and a column-based data searching method. The column-based data managing method includes determining whether the size of the column-group data file exceeds a partitioning threshold, dividing the column-group data if the size exceeds the partitioning threshold, and generating divided column-group data files.

Abstract: Provided is a method for data recovery using parallel processing in a cluster data management system. The method includes arranging a redo log written by a failed partition server, dividing the arranged redo log by columns of the partition, and recovering data parallelly on the basis of the divided redo log and multiple processing unit.

Abstract: Provided are a cluster data management system and a method for data restoration using a shared redo log in the cluster data management system. The data restoration method includes collecting service information of a partition served by a failed partition server, dividing redo log files written by the partition server by columns of a table including the partition, restoring data of the partition on the basis of the collected service information and log records of the divided redo log files, and selecting a new partition server that will serve the data-restored partition, and allocating the partition to the selected partition server.

Abstract: Provided are a content-based searching method and system for multimedia objects using a high-dimensional feature vector data based on a 2-level signature. The method for searching the high-dimensional data using a signature file includes calculating a first-level query signature and a second-level query signature by using the query feature vector, performing a first filtering operation to obtain a primary candidate cell group by searching a second-level signature file, and performing a secondary filtering operation to obtain a secondary candidate cell group having a high similarity in a primary candidate cell group. Accordingly, the high-dimensional data searching method and system can process a query quickly and accurately and can increase the searching accuracy by using an enhanced signature of the query feature vector.

Abstract: Provided are a system and a method for indexing high-dimensional data in parallel in a cluster environment. The system for indexing high-dimensional data in parallel in a cluster environment includes a Spill-tree creation means for creating a Spill-tree using an sampled N-dimensional feature vector, a feature vector division storage means for distributedly storing the N-dimensional feature vector in a terminal node of the Spill-tree, and a local signature creation means for creating and managing a local signature for the N-dimensional feature vector dispersed into each node of the Spill-tree.