公开(公告)号：KR100422801B1

公开(公告)日：2004-03-12

申请号：KR1020010031213

申请日：2001-06-04

Applicant: 한국전자통신연구원

Inventor： 김경배 ,  김창수 ,  신범주

IPC: G11B20/10

Abstract: PURPOSE: A large-capacity file storage system and a method of adding and deleting data blocks of dynamic multi-level inode are provided to effectively manage large-capacity files. CONSTITUTION: A large-capacity file storage system includes a pointer having a level different from the level of an inode information area(101A) storing information about root inode. The pointer includes a double indirect inner pointer in which a data block exists through double indirect pointer nodes, a single indirect inner pointer in which a data block exists through a single indirect pointer node, and a direct point that directly points a data block. The pointer of the root inode is allocated according to the direct pointer, single indirect pointer node and double indirect pointer node by increasing and decreasing the level in accordance with the size of the data block.

Abstract translation: 目的：提供一个大容量文件存储系统和一种增加和删除动态多级别inode数据块的方法，以有效管理大容量文件。 组成：大容量文件存储系统包括一个指针，该指针的级别与存储有关根节点信息的信息节点信息区域（101A）的级别不同。 指针包括双重间接内部指针，其中数据块通过双重间接指针节点存在，单个间接内部指针通过单个间接指针节点存在数据块，以及直接指向数据块的直接点。 根据索引节点的指针根据直接指针，单个间接指针节点和双重间接指针节点通过根据数据块的大小增加和减小电平来分配。

公开(公告)号：KR1020030073981A

公开(公告)日：2003-09-19

申请号：KR1020020013800

申请日：2002-03-14

Applicant: 한국전자통신연구원

Inventor： 이용주 ,  김경배 ,  신범주

IPC: G06F17/30

CPC classification number: G06F17/30091 ,  Y10S707/99953 ,  Y10S707/99956

Abstract: PURPOSE: A method for managing a directory of a high capacity file system is provided to manage a directory by forming an indirect block as a semi flat structure by applying a changed extendable hashing technique based on a fixed length extent block and a virtual block mapping technique, and memorizing the next insertion position of a directory using a scan indicator for preventing a sequential search in an extent block. CONSTITUTION: A root block(50) for storing a directory entry is created. The directory entry is stored continuously and sequentially. At this time, the root block(50) has a fixed global depth. If a directory entry is filled entirely in the root block(50), an index value for displaying position information for deciding the number of bits to be referred with respect to the root block(50) from the global depth through a hash function is searched. A block number of leaf blocks(60,70) comprises extent blocks(61-64,71-74) based on four fixed lengths which are logical and continuous objects. The leaf blocks(60,70) are searched using the index value. The block number is stored in the root block(50) sequentially, and a directory entry of the root block(50) is stored in the leaf blocks(60,70).

Abstract translation: 目的：提供一种用于管理大容量文件系统的目录的方法，通过使用基于固定长度范围块和虚拟块映射技术的改变的可扩展散列技术来形成间接块作为半平面结构来管理目录 并且使用扫描指示符来存储目录的下一个插入位置，以防止扩展块中的顺序搜索。 构成：创建用于存储目录条目的根块（50）。 目录条目是连续和顺序存储的。 此时，根块（50）具有固定的全局深度。 如果目录条目完全填充在根块（50）中，则搜索用于通过散列函数从全局深度显示用于决定关于根块（50）要参考的位数的位置信息的索引值 。 基于作为逻辑和连续对象的四个固定长度的块数目（60,70）包括扩展块（61-64,71-74）。 使用索引值搜索叶块（60,70）。 块编号依次存储在根块（50）中，根块（50）的目录项被存储在叶块（60,70）中。

公开(公告)号：KR100392382B1

公开(公告)日：2003-07-23

申请号：KR1020010045621

申请日：2001-07-27

Applicant: 한국전자통신연구원

Inventor： 김창수 ,  김경배 ,  신범주

IPC: G06F12/02

CPC classification number: G06F3/0601 ,  G06F2003/0697

Abstract: A method for managing a logical volume for minimizing a size of metadata and supporting dynamic online resizing, computer-readable recoding medium storing a program or data structure for embodying the method are disclosed. The method for managing a logical volume in order to support dynamic online resizing and minimizing a size of metadata, said method comprising steps of: a) creating a logical volume by rounding up disk partitions in response to a request of constructing the logical volume on a physical storage space; b) generating metadata including information of the logical volume and the disk partitions participating to the logical volume and storing it to the disk partitions participating to the logical volume; c) dynamically resizing the logical volume in response to a request of resizing, and modifying the metadata on the disk partitions participating to the logical volume; and d) calculating and returning a physical address corresponding to a logical address of the logical volume by using mapping information of the metadata containing information of the physical address corresponding to the logical address.

Abstract translation: 公开了一种用于管理用于最小化元数据的大小并支持动态在线调整大小的逻辑卷的方法，存储用于实现该方法的程序或数据结构的计算机可读记录介质。 用于管理逻辑卷以便支持动态在线调整大小和最小化元数据大小的方法，所述方法包括以下步骤：a）响应于构建逻辑卷上的逻辑卷的请求，通过将磁盘分区四舍五入来创建逻辑卷 物理存储空间; b）生成包括逻辑卷和参与逻辑卷的磁盘分区的信息的元数据，并将其存储到参与逻辑卷的磁盘分区; c）响应于调整大小的请求动态调整逻辑卷的大小，并修改参与逻辑卷的磁盘分区上的元数据; 以及d）通过使用包含对应于逻辑地址的物理地址的信息的元数据的映射信息来计算并返回对应于逻辑卷的逻辑地址的物理地址。

公开(公告)号：KR1020030034577A

公开(公告)日：2003-05-09

申请号：KR1020010066218

申请日：2001-10-26

Applicant: 한국전자통신연구원

Inventor： 박유현 ,  김창수 ,  김경배 ,  신범주

IPC: G06F3/06

CPC classification number: G06F3/0632 ,  G06F3/0607 ,  G06F3/0689

Abstract: PURPOSE: A striping system and its mapping and processing method is provided to store striping data at a RAID(Redundancy Array of Inexpensive Disks) for enhancing a disk IO performance, and specially, to make it unnecessary rearranging data on the disk by using a stripping zone information table and a mapping algorithm. CONSTITUTION: The method comprises several steps. A striping system determines whether there exists an additional disk at a storage media(S100). If a new disk is added, the striping system adds the striping zone information of the new disk to the existing striping zone information table(S102). The system converts the logical address of write data in the striping zone information table into the physical one, searches for the physical position, and determines a corresponding striping zone(S104). In a case that the system receives a disk write instruction, the system accesses a physical block within a striping zone of a corresponding disk based on the converted table data, and writes the data on the striping zone(S106, S108). In a case that the system does not receives a disk write instruction, the system accesses a physical block within a striping zone of a corresponding disk based on the converted table data, and reads the data from the striping zone(S106, S110).

Abstract translation: 目的：提供条带化系统及其映射和处理方法，以便在RAID（廉价磁盘冗余阵列）中存储条带化数据，以增强磁盘IO性能，特别是通过使用剥离不必要地重新排列磁盘上的数据 区域信息表和映射算法。 构成：该方法包括几个步骤。 条带化系统确定在存储介质上是否存在附加的盘（S100）。 如果添加新磁盘，则条带化系统将新磁盘的条带区信息添加到现有的条带区信息表中（S102）。 系统将条带区信息表中的写数据的逻辑地址转换成物理位置，搜索物理位置，并确定相应的条带区（S104）。 在系统接收到磁盘写指令的情况下，系统基于转换后的表数据访问相应磁盘的条带区内的物理块，并将数据写入条带区（S106，S108）。 在系统没有接收到磁盘写指令的情况下，系统基于转换的表数据访问对应磁盘的条带区内的物理块，并从条带区读取数据（S106，S110）。

公开(公告)号：KR1020030013619A

公开(公告)日：2003-02-15

申请号：KR1020010047715

申请日：2001-08-08

Applicant: 한국전자통신연구원

Inventor： 박춘서 ,  김창수 ,  김경배 ,  신범주 ,  유재수 ,  송석일 ,  신재룡

IPC: G06F17/30

Abstract: PURPOSE: A method for inserting and searching data of a parallel higher index structure is provided to search a higher index effectively by transforming a higher index structure which uses a parallel property of a SAN(Storage Area Network) thereby increasing a fan out, reducing a height of a tree, and maximizing a parallel property of an input/output at searching a range in searching a similarity. CONSTITUTION: For performing a partial K-most access query in a main server and all sub servers simultaneously, if a K-access query is entered, all servers access to a root node(1300). It is judged whether the accessed root node is a non-terminal node or not(1301). If the accessed root node is a non-terminal node, all servers calculates a similarity with a query and an entry, and each entry is sorted in a list in order of similarity(1302). All servers access to a child node of the first entry stored in the list in parallel and allocate the current node as a root node(1303), and the current stage is returned to the above stage (1301) for judging a non-terminal node of not of the root node.

Abstract translation: 目的：提供一种用于插入和搜索并行较高索引结构的数据的方法，通过转换使用SAN（存储区域网络）的并行属性的较高索引结构，有效地搜索较高索引，从而增加扇出，减少 在搜索相似度的范围内最大化输入/输出的并行属性。 规定：为了在主服务器和所有子服务器中同时执行部分K最多访问查询，如果输入K访问查询，则所有服务器都访问根节点（1300）。 判断所访问的根节点是否是非终端节点（1301）。 如果所访问的根节点是非终端节点，则所有服务器计算与查询和条目的相似性，并且每个条目按照相似度的顺序排列在列表中（1302）。 所有服务器并行地存储在列表中的第一条目的子节点并分配当前节点作为根节点（1303），并且当前阶段返回到上述阶段（1301），以判断非终端节点 不是根节点。

公开(公告)号：KR100529278B1

公开(公告)日：2005-11-17

申请号：KR1020030092526

申请日：2003-12-17

Applicant: 한국전자통신연구원

Inventor： 강동재 ,  김경배 ,  정성인 ,  김명준

IPC: G06F12/16

Abstract: 본 발명은 데이터를 중복 저장하는 읽기 연산 집약 시스템에서 디스크에 저장되는 데이터 블록의 배치를 수정하고 이에 따른 새로운 맵핑 방식을 제공함으로써 기존의 데이터 중복 저장 방식(RAID1)의 장점인 데이터 신뢰성을 유지하면서 동시에 데이터 분산 저장 방식(RAID0)의 빠른 읽기 성능을 갖는 대용량 데이터에 대한 데이터 중복 저장 시스템에 관한 것이다.
본 발명의 데이터 중복 저장 시스템은 원본 데이터를 저장하는 원본 디스크와 상기 원본 데이터에 대한 사본들을 저장하는 다수의 중복 디스크들로 구성되며, 디스크 개수만큼의 연속적인 데이터 블록들을 그룹핑하여 상기 원본 디스크와 상기 중복 디스크들에 대한 SMU들을 생성한 후 각 디스크내에서 SMU 순서에 따라 SMU에 SMUno값을 부여하고 각 SMU내에서의 데이터 블록의 순서에 따라 데이터 블록에 SMUidx값을 부여하며, 동일 SMUno값의 각 SMU내에서의 동일 데이터 블록들은 각 디스크마다 서로 다른 배치 순서를 갖는 것을 특징으로 한다.

公开(公告)号：KR1020040012388A

公开(公告)日：2004-02-11

申请号：KR1020020045951

申请日：2002-08-03

Applicant: 한국전자통신연구원

Inventor： 박선영 ,  손덕주 ,  김경배

IPC: G06F15/16

Abstract: PURPOSE: A method for managing a global buffer capable of executing a load distribution in a cluster system is provided to calculate loads of each host and distribute a request of a global buffer manager for preventing a performance decline caused by a concentration to a specific host in a system of a system load necessary for an operation of a global buffer manager when a buffer cache is shared among many hosts through a global buffer manager in a cluster system. CONSTITUTION: The upper module of a specific host of a cluster system requests a transmission of a buffer block to a global buffer manager(S131). The global buffer manager requests a permission as to a corresponding disk block to a locking manager(S132). The locking manager transmits a response message including an accept of the permission to the global buffer manager(S133). If the global buffer manager acquires a permission, it is checked whether a requested buffer block exists in a buffer cache of a host having a permission(S136). If a requested buffer block does not exist in a buffer cache of a host having a permission, it is checked whether a requested buffer block exists in a buffer cache of other host(S137). If a requested buffer block exists in a buffer cache of other host, a transmission of a buffer block is requested to other host through a communication module(S138,S140). If a transmission of a buffer block requested to other host is failed, the buffer block requested through a disk I/O is copied to the buffer cache(S139). The copied buffer block or a buffer block received from a buffer cache of other host is transmitted to the upper module(S142). If the upper block requests a release of a buffer block use, the global buffer manager returns a permission as to the buffer block(S145).

Abstract translation: 目的：提供一种用于管理能够在集群系统中执行负载分布的全局缓冲器的方法，以计算每个主机的负载并分发全局缓冲管理器的请求，以防止由特定主机的集中引起的性能下降 当通过集群系统中的全局缓冲管理器在许多主机之间共享缓冲区高速缓存时，全局缓冲器管理器的操作所需的系统负载的系统。 构成：集群系统的特定主机的上层模块请求向全局缓冲区管理器发送缓冲区块（S131）。 全局缓冲区管理器向锁定管理器请求对相应磁盘块的许可（S132）。 锁定管理器发送包括对全局缓冲器管理器的许可的接受的响应消息（S133）。 如果全局缓冲器管理器获取许可，则检查在具有许可的主机的缓冲器高速缓存中是否存在所请求的缓冲器块（S136）。 如果在具有许可的主机的缓冲器高速缓存中不存在所请求的缓冲器块，则检查所请求的缓冲器块是否存在于其他主机的缓冲器高速缓存中（S137）。 如果所请求的缓冲区块存在于其他主机的缓冲区高速缓存中，则通过通信模块向其他主机请求缓冲区块的传输（S138，S140）。 如果请求到其他主机的缓冲区的传输失败，通过磁盘I / O请求的缓冲区被复制到缓冲区高速缓存（S139）。 将复制的缓冲器块或从其他主机的缓冲器高速缓存接收的缓冲器块发送到上位模块（S142）。 如果上限块请求释放缓冲块使用，则全局缓冲区管理器返回对缓冲区块的许可（S145）。

公开(公告)号：KR1020030071245A

公开(公告)日：2003-09-03

申请号：KR1020020010855

申请日：2002-02-28

Applicant: 한국전자통신연구원

Inventor： 김경배 ,  이용주 ,  박춘서 ,  박선영 ,  신범주

IPC: G06F17/40

Abstract: PURPOSE: A mass file storage system and a dynamic bitmap management method are provided to dynamically allocate a bit map without being restricted by a file size so that it can effectively manage large sized files. CONSTITUTION: The method comprises several steps. A bit map is read from a storage device(S521). A pattern value is read from the header of the bit map in order to be compared with an object pattern requested by a file system(S522). If the pattern of the bit map is identical to that of the object pattern, the number of used bits is compared with that of the total bits(S523). If the number of the used bits is less than that of the total bits, unused bits are set with "1"(S524). Then, a bit map area is allocated to a corresponding object(S525). If the pattern of the bit map is not identical to that of the object or the number of the used bits is more than that of the total bits, it is checked whether there is a bit map in use allocated at a current bit map area(S526). If there is a bit map in use, a new bit map is read, otherwise an unused bit map is selected from a bit map area and is set with a requested pattern(S527). Then, the first bit of the new allocated bit map is set with "1" for allocating a bit map area to the object(S528).

Abstract translation: 目的：提供大容量文件存储系统和动态位图管理方法来动态分配位图，而不受文件大小的限制，从而可以有效地管理大型文件。 构成：该方法包括几个步骤。 从存储装置读取位图（S521）。 从比特图的头部读取图案值，以便与由文件系统请求的对象图案进行比较（S522）。 如果位图的图案与对象图案的图案相同，则将所使用的位数与总位的数目进行比较（S523）。 如果所使用的位的数量小于总位数，则将未使用的位设置为“1”（S524）。 然后，将位图区域分配给相应的对象（S525）。 如果位图的图案与对象的图案不一致，或者使用的位的数量大于总位数，则检查在当前位图区域是否存在使用中分配的位图（ S526）。 如果存在使用的位图，则读取新的位图，否则从位图区域中选择未使用的位图，并且以请求的图案设置（S527）。 然后，将新分配的位图的第一位设置为“1”，以将位图区域分配给对象（S528）。

公开(公告)号：KR100378598B1

公开(公告)日：2003-04-03

申请号：KR1020010021337

申请日：2001-04-20

Applicant: 한국전자통신연구원

Inventor： 이용주 ,  김경배 ,  김창수 ,  신범주

IPC: G06F15/16

Abstract: PURPOSE: A buffer management system and method is provided to perform an integrated management of individual buffer at a host computer networked with other host computers, and to support a transmission of contents at the buffer among the host systems so that the host systems can conceptually share the buffers with one another. CONSTITUTION: The system comprises a plurality of individual buffer managers, a global buffer manager, a plurality of individual locking managers, and a global locking manager. The individual buffer manager, installed at each host system, manages an individual list of data stored at a buffer directly connected to each host. The global buffer manager, installed at a global buffer server for performing an integrated management of the host systems, integrates data stored at the buffers connected to the host systems and manages a global list of the data. The individual locking manager, installed at each host, manages an individual list of current locking data of each host system. The global locking manager, installed at the global buffer server, manages a global locking list of data to be managed by the global buffer server.

Abstract translation: 目的：提供了一种缓冲器管理系统和方法，用于在与其他主计算机联网的主计算机上执行单个缓冲器的集成管理，并且支持在主系统之间在缓冲器上传输内容，使得主系统可以在概念上共享 彼此之间的缓冲区。 组成：系统包括多个单独的缓冲区管理器，全局缓冲区管理器，多个单独的锁定管理器和全局锁定管理器。 安装在每个主机系统上的单独缓冲区管理器管理存储在直接连接到每个主机的缓冲区中的单独数据列表。 全局缓冲区管理器安装在全局缓冲服务器上，用于执行主机系统的集成管理，集成存储在连接到主机系统的缓冲区中的数据并管理全局数据列表。 单个锁定管理器安装在每个主机上，用于管理每个主机系统当前锁定数据的单独列表。 安装在全局缓冲服务器上的全局锁定管理器管理由全局缓冲服务器管理的全局锁定数据列表。

公开(公告)号：KR1020020081886A

公开(公告)日：2002-10-30

申请号：KR1020010021337

申请日：2001-04-20

Applicant: 한국전자통신연구원

Inventor： 이용주 ,  김경배 ,  김창수 ,  신범주

IPC: G06F15/16

Abstract: PURPOSE: A buffer management system and method is provided to perform an integrated management of individual buffer at a host computer networked with other host computers, and to support a transmission of contents at the buffer among the host systems so that the host systems can conceptually share the buffers with one another. CONSTITUTION: The system comprises a plurality of individual buffer managers, a global buffer manager, a plurality of individual locking managers, and a global locking manager. The individual buffer manager, installed at each host system, manages an individual list of data stored at a buffer directly connected to each host. The global buffer manager, installed at a global buffer server for performing an integrated management of the host systems, integrates data stored at the buffers connected to the host systems and manages a global list of the data. The individual locking manager, installed at each host, manages an individual list of current locking data of each host system. The global locking manager, installed at the global buffer server, manages a global locking list of data to be managed by the global buffer server.

Abstract translation: 目的：提供一种缓冲管理系统和方法，用于在与其他主机计算机联网的主机上执行单个缓冲区的集成管理，并支持在主机系统之间的缓冲区内传输内容，以便主机系统可以概念上共享 缓冲区彼此。 构成：系统包括多个单独的缓冲管理器，全局缓冲管理器，多个单独的锁定管理器和全局锁定管理器。 安装在每个主机系统上的单个缓冲管理器管理存储在直接连接到每个主机的缓冲器上的数据的单独列表。 全局缓冲区管理器安装在全局缓冲服务器上，用于执行主机系统的集成管理，将存储在连接到主机系统的缓冲区上的数据进行集成，并管理数据的全局列表。 安装在每个主机上的单个锁定管理器管理每个主机系统当前锁定数据的单独列表。 安装在全局缓冲区服务器上的全局锁定管理器管理由全局缓冲区服务器管理的全局数据锁​​定列表。