Movatterモバイル変換


[0]ホーム

URL:


CN109521959A - One kind being based on SSD-SMR disk mixing key assignments memory system data method for organizing - Google Patents

One kind being based on SSD-SMR disk mixing key assignments memory system data method for organizing
Download PDF

Info

Publication number
CN109521959A
CN109521959ACN201811296432.6ACN201811296432ACN109521959ACN 109521959 ACN109521959 ACN 109521959ACN 201811296432 ACN201811296432 ACN 201811296432ACN 109521959 ACN109521959 ACN 109521959A
Authority
CN
China
Prior art keywords
key
value
ssd
data
storage system
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN201811296432.6A
Other languages
Chinese (zh)
Inventor
伍卫国
聂世强
孙岚子
李翀
刘钊华
毛海
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xian Jiaotong University
Original Assignee
Xian Jiaotong University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Xian Jiaotong UniversityfiledCriticalXian Jiaotong University
Priority to CN201811296432.6ApriorityCriticalpatent/CN109521959A/en
Publication of CN109521959ApublicationCriticalpatent/CN109521959A/en
Pendinglegal-statusCriticalCurrent

Links

Classifications

Landscapes

Abstract

Translated fromChinese

本发明公开了一种基于SSD‑SMR磁盘混合键值存储系统数据组织方法,在数据存储过程中,将数据存储索引过程中的键值对在由SSD和SMR磁盘组成的混合存储系统中解耦合存储,根据数据特征将数据分布到SSD和SMR磁盘,将键值对中的键存放在SSD,将键值对中的值存放在SMR磁盘,根据接收到上层应用发送请求指令后,在内存和SSD上的LSM树查找目标key,根据查找结果,进行相应指令操作,将键值数据解耦合打散分布,实现混合存储的最大化利用,利用SSD的高性能和SMR磁盘的大容量特点来优化应用对数据访问速度,设计和实现低成本、高性能和大容量的混合键值存储系统。本发明能够优化数据读写速度,实现低成本、高性能和大容量的混合键值存储系统。

The present invention disclosed a data organization method based on the SSD‑SMR disk hybrid key value storage system. During the data storage procedure, the key value in the data storage index process is decoupled in the mixed storage system composed of SSD and SMR disks. Storage, distribute data to SSD and SMR disks according to data features, place key values on SSD, store the value of the key value on the SMR disk, and send the request instruction according to the receiving upper application. The LSM tree on the SSD finds target key, according to the search results, perform corresponding instruction operations, decoupled the key value data to disperse distribution, realize the maximum utilization of mixed storage, and optimize the high performance of SSD and the large capacity characteristics of SMR disks to optimize Applying data access speed, design and realizing hybrid key storage systems with low cost, high performance and large -capacity. The present invention can optimize the data read and write speed and achieve low -cost, high -performance and large -capacity mixed key storage system.

Description

One kind being based on SSD-SMR disk mixing key assignments memory system data method for organizing
Technical field
The present invention relates to computers to mix storage system field, more particularly to a kind of is mixed based on SSD-SMR diskKey assignments memory system data method for organizing.
Background technique
With the rapid development of the emerging technologies such as mobile Internet, cloud computing, Internet of Things, information-intensive society enters networkingBig data era.The quickly global mobile data flow of universal Intelligent mobile terminal (computer, mobile phone, plate etc.) boostingIncrease substantially." the digital universe " of IDC LLC reports display, from 2005 to the year two thousand twenty whole world it is annual create, duplication andThe data volume used will increase by 300 times, rise to 40ZB from 130EB.The current total amount of data of Baidu is more than 1000PB, dailyWeb data to be treated reaches 10PB~100PB;Taobao's aggregate transaction data amount is up to 100PB.It among these include magnanimityThe unstructured datas such as picture, video, text, voice, however traditional relevant database can only be used to storage organizationData, with the exponential growth of data volume, relevant database can not be applicable in big data era.Solve this problemEfficient scheme is exactly to use non-relational database, and the key assignments (Key-Value, KV) in non-relational database is stored in big dataEpoch play vital role, have been widely used in the fields such as search engine, e-commerce, social networks,In more mature product have BigTable, LevelDB of Google company, HBase, Cassandra of Facebook companyAnd RocksDB, the PNUTS of Yahoo, the Atlas etc. of Baidu company.
At present the storage medium of mainstream capacity, cost, performance, in terms of there are great differences, if using singleThe covered range of needs of one storage medium building key assignments storage system institute is limited.Heterogeneous storage medium is integrated in oneIt rises, constitutes mixing storage system, respective advantage can be played, maximized favourable factors and minimized unfavourable ones, demand coverage area can be expanded significantly.It will be highThe novel storage medium of performance and the disk combination of large capacity are the common approach for mixing storage system, due to a watt record(Shingled Magnetic Recording, SMR) disk storage capacity is big and flash memory solid-state disk (the Solid based on NANDState Disk, SSD) the fast characteristic of read or write speed, the integrated of SSD and SMR disk met high capacity, high-performance and low costDemand.
Integrate SSD and SMR disk directly simply to construct the mixing key assignments storage system based on SSD-SMR with level frameworkIt is unable to fully the characteristic of the high-performance and SMR magnetic disk mass using SSD.The reason is that disk manufacturer is in order to be compatible with traditional softPart joined conversion layer to shield internal structure difference, such as flash translation layer (FTL) (the Flash Translation of SSD whereinLayers, FTL) and SMR watt record conversion layer (Shingled Translation Layers, STL), and conversion layer is depositedCausing upper layer application that can not perceive its internal structure, to can not carry out going deep into optimization for its internal structure.
Summary of the invention
The purpose of the present invention is to provide one kind to be based on SSD-SMR disk mixing key assignments memory system data method for organizing,With overcome the deficiencies in the prior art.
In order to achieve the above objectives, the present invention adopts the following technical scheme:
One kind will be counted in data storage procedure based on SSD-SMR disk mixing key assignments memory system data method for organizingAccording to key-value pair decoupling storage in the mixing storage system being made of SSD and SMR disk in storage Index process, according to numberAccording to feature by data distribution to SSD and SMR disk, the key in key-value pair is stored in SSD, the value in key-value pair is stored inSMR disk, after receiving upper layer application transmission request instruction, the LSM tree on memory and SSD searches target key, according toLookup result carries out command adapted thereto operation.
Further, key assignments index LSM tree is stored in SSD by the index based on LSM tree building key assignments storage system.
Further, the node of LSM tree is address tuple, with four-tuple < key, value-address, value-Offset, value-size > expression;Key is globally unique, is the identifier of value;Value-address and value-Offset is respectively directed to address and the deviant of daily record data.
Further, value is stored in a manner of daily record data in SMR disk.
Further, in SMR data in magnetic disk storing process, when the capacity of each value is greater than its own available capacityThreshold value when, and storage system at this time load it is relatively light when, carry out garbage reclamation at this time, the granularity of garbage reclamation is needleTo each key-value pair carry out, and each key-value pair carry out garbage collection operation be it is businesslike, first on SMR diskIt divides data collection according to Zone (region), pointer tuple is then updated on SSD, then it is complete for the garbage reclamation of this key-value pairAt.
Further, after storage system, which receives upper layer application, sends read request, the LSM tree traversed first in memory is looked intoTarget key is looked for, if searching failure, then the LSM tree on SSD searches target key, if searching failure, then returns to userError message shows that this key is not stored in storage system, if finding target key, reads the finger where current keyNeedle nodal information, the value of SMR disk is read according to pointer nodal information, and returns data to user.
Further, after storage system, which receives upper layer application, sends write request, traversal is on memory and SSD firstLSM tree searches target key, if searching failure, shows to be this time write operation, new space is distributed on SMR disk, willValue is written to the first address of free space, and address information and deviant, data size information and key are then combined into fourThis tuple is inserted into LSM tree by tuple;If searching successfully, this operation is to update operation, at this time first on SMR diskThe original of value spatially adds new data, then updates nodal information on LSM tree.
Compared with prior art, the invention has the following beneficial technical effects:
It is of the invention a kind of based on SSD-SMR disk mixing key assignments memory system data method for organizing, in data storage procedureIn, data are stored into the key-value pair in Index process and decoupling in the mixing storage system being made of SSD and SMR disk are depositedStorage, according to data characteristics by data distribution to SSD and SMR disk, is stored in SSD for the key in key-value pair, will be in key-value pairValue is stored in SMR disk, and after receiving upper layer application transmission request instruction, the LSM tree on memory and SSD searches targetKey carries out command adapted thereto operation according to lookup result, and distribution is broken up in the conjunction of key assignments data decoupler, realizes the maximum of mixing storageChange and utilizes.The layout type of data has been largely fixed systematicness in the mixing key assignments storage of SSD and SMR disk compositionCan, the Method of Data Organization of the decoupling key assignments storage of the present invention in most cases can all be hit using the read-write of key and is buffered inThe LSM tree of memory and value are stored by the way of additional write, and maximum dynamics excavates mixing storage system maximum performance, are reducedCarrying cost;Optimized using the high-performance of SSD and the large capacity feature of SMR disk using to data access speed, design andRealize low cost, the mixing key assignments storage system of high-performance and large capacity.The present invention can optimize reading and writing data speed, realize lowThe mixing key assignments storage system of cost, high-performance and large capacity.
Further, pointer tuple is stored in SSD, pointer tuple value as corresponding to key and its is in SMR diskAddress space composition, key assignments storage system can read and write the value on SMR disk according to key access pointer tuple;?SMR disk storage log data block, value is stored in a manner of log, i.e., write-in or more new capital are write in a manner of additional write every timeEnter to improve write performance, Value additional write back in a manner of log brings redundant storage expense, storage system can periodically intoRow cleaning compression, deletes original data information to save storage resource.
Detailed description of the invention
Fig. 1 is that key assignments mixes storage system architecture diagram.
Fig. 2 is key assignments of the present invention separation storage schematic diagram.
Fig. 3 is that key assignments storage system of the present invention reads schematic diagram.
Fig. 4 is that key assignments storage system of the present invention writes schematic diagram.
Specific embodiment
The invention will be described in further detail with reference to the accompanying drawing:
The organizational form of heterogeneous data in storage medium is always research hotspot, and common mode has building Hierarchical storageWith storage extending transversely, i.e. SSD is combined as the cache layer or SSD and HDD of HDD in unified logical address space, the present inventionConsider data organization feature in key assignments storage, selects the storage medium organizational form of storage extending transversely.HM-SSD and SMR diskIt is that difference towards Host Administration, with conventional storage media is its internal physical structure to upper layer application exposure, upper layer is answeredWith can make corresponding optimisation strategy according to its physical structure, therefore both media are that the present invention provides hardware supporteds.The Method of Data Organization that the present invention designs make full use of SSD high-performance and SMR disk large capacity feature come optimize using pairData access speed, design and implementation low cost, the mixing key assignments storage system of high-performance and large capacity.
One kind being based on SSD-SMR disk mixing key assignments memory system data method for organizing, is based on HM-SSD and HM-SMR magneticThe storage system extending transversely of disk composition, the key-value pair Key and Value of key assignments storage occur in pairs, Key Value Data amountIt is small, for indexing;Value data volume is larger, for storing data;Data are stored into the key-value pair in Index process by SSDWith decoupling storage in the mixing storage system of SMR disk composition, according to data characteristics by data distribution to SSD and SMR disk,By the biggish small documents of access frequency, i.e. key in key-value pair is stored in SSD, the less big file of access frequency, i.e. key-value pairIn value be stored in SMR disk, the compromise of performance and cost is realized with this.
As shown in Figure 1, underlying storage medium is made up of HM-SSD and HM-SMR disk physical bus interconnection, host sideConversion layer (TL) is made of multiple software modules, including data management, garbage reclamation, cache management and address of cache function mouldBlock, this partial function module realize that upper layer application externally provides key-value pair storage service in operating system software layer.
It is primarily based on SSD-SMR disk mixing key assignments memory system data treatment process, as shown in Fig. 2, key assignments is indexedLSM tree is stored in SSD, and the node of LSM tree is address tuple, with four-tuple < key, value-address, value-Offset, value-size > expression, key are globally unique, are the identifiers of value, value is in SMR disk with logThe mode of data stores, and value-address and value-offset are respectively directed to address and the deviant of daily record data;
SMR disk itself updates, since the value of SMR disk is that addition is write in a manner of log, it is therefore desirable to periodicallyGarbage collection operation, specific garbage collection procedure: when each value capacity be greater than its own available capacity threshold valueWhen, and storage system at this time load it is relatively light when, carry out garbage reclamation at this time, the granularity of garbage reclamation is for eachKey-value pair carry out, and each key-value pair carry out garbage collection operation be it is businesslike, first on SMR disk according toZone (region) divides data collection, and pointer tuple is then updated on SSD, then completes for the garbage reclamation of this key-value pair.
It is illustrated in figure 3 key assignments storage system and reads schematic diagram, after storage system, which receives upper layer application, sends read request,The LSM tree traversed first in memory searches target key, if searching failure, then the LSM tree on SSD is searched, if searchingFailure, then return to user's error message, show that this key is not stored in storage system, if finding target key, readsPointer nodal information where key at present, the value of SMR disk is read according to pointer nodal information, and returns data to useFamily.So far reading process terminates.
It is illustrated in figure 4 key assignments storage system and writes schematic diagram, after storage system, which receives upper layer application, sends write request,LSM tree of the traversal on memory and SSD first searches target key, if searching failure, shows to be this time write operation, in SMR magneticNew space is distributed on disk, and value is written to the first address of free space, it is then that address information and deviant, data is bigSmall information and key are combined into four-tuple, this tuple is inserted into LSM tree;If searching successfully, this operation is to update operation,It adds new data on original space of value first on SMR disk at this time, then updates nodal information on LSM tree.So far it readsProcess terminates.
In data storage procedure, data are stored into the key-value pair in Index process and are mixed what is be made of SSD and SMR diskDecoupling storage in storage system is closed to deposit the key in key-value pair according to data characteristics by data distribution to SSD and SMR diskBe placed on SSD, the value in key-value pair be stored in SMR disk, according to receiving after upper layer application sends request instruction, in memory andLSM tree on SSD searches target key, according to lookup result, carries out command adapted thereto operation, and the conjunction of key assignments data decoupler is broken up pointCloth realizes maximally utilizing for mixing storage.The layout type for mixing data in key assignments storage of SSD and SMR disk composition is verySystem performance, the Method of Data Organization of the decoupling key assignments storage of the present invention, although the read-write process of data are determined in big degreeMiddle relative increase is once searched, but is buffered in the LSM tree of memory since the read-write of key in most cases can all be hit, andValue is stored by the way of additional write, and the performance of the total system of storage system can't be affected substantially, and the present invention is maximumDynamics excavates mixing storage system maximum performance, reduces carrying cost;It is special using the high-performance of SSD and the large capacity of SMR diskPoint is applied to optimize to data access speed, design and implementation low cost, the mixing key assignments storage system of high-performance and large capacity.The present invention can optimize reading and writing data speed, realize low cost, the mixing key assignments storage system of high-performance and large capacity.

Claims (7)

Translated fromChinese
1.一种基于SSD-SMR磁盘混合键值存储系统数据组织方法,其特征在于,在数据存储过程中,将数据存储索引过程中的键值对在由SSD和SMR磁盘组成的混合存储系统中解耦合存储,根据数据特征将数据分布到SSD和SMR磁盘,将键值对中的键存放在SSD,将键值对中的值存放在SMR磁盘,根据接收到上层应用发送请求指令后,在内存和SSD上的LSM树查找目标key,根据查找结果,进行相应指令操作。1. A data organization method based on SSD-SMR disk hybrid key-value storage system, characterized in that, in the data storage process, the key-value pair in the data storage index process is stored in the hybrid storage system composed of SSD and SMR disks Decoupled storage, distribute data to SSD and SMR disk according to data characteristics, store the key in the key-value pair in SSD, and store the value in the key-value pair in SMR disk, after receiving the request instruction sent by the upper layer application, in The LSM tree on the memory and SSD searches for the target key, and performs corresponding instruction operations according to the search results.2.根据权利要求1所述的一种基于SSD-SMR磁盘混合键值存储系统数据组织方法,其特征在于,基于LSM树构建键值存储系统的索引,将键值索引LSM树存放在SSD中。2. a kind of data organization method based on SSD-SMR disk hybrid key-value storage system according to claim 1, is characterized in that, builds the index of key-value storage system based on LSM tree, stores key-value index LSM tree in SSD3.根据权利要求2所述的一种基于SSD-SMR磁盘混合键值存储系统数据组织方法,其特征在于,LSM树的节点即是地址元组,以四元组<key,value-address,value-offset,value-size>表示;key是全局唯一的,是value的标识符;value-address和value-offset分别指向日志数据的地址和偏移值。3. a kind of data organization method based on SSD-SMR disk hybrid key-value storage system according to claim 2, it is characterized in that, the node of LSM tree is address tuple, with quadruple<key, value-address, value-offset, value-size> means; key is globally unique and is the identifier of value; value-address and value-offset point to the address and offset value of the log data respectively.4.根据权利要求3所述的一种基于SSD-SMR磁盘混合键值存储系统数据组织方法,其特征在于,value在SMR磁盘中以日志数据的方式存储。4. A method for organizing data in a hybrid key-value storage system based on SSD-SMR disks according to claim 3, wherein the value is stored as log data in the SMR disk.5.根据权利要求1所述的一种基于SSD-SMR磁盘混合键值存储系统数据组织方法,其特征在于,在SMR磁盘数据存储过程中,当每个value的容量大于其自身有效容量的阈值时,并且此时存储系统的负载相对较轻时,此时进行垃圾回收,垃圾回收的粒度是针对每个键值对进行的,并且每个键值对进行垃圾回收操作都是事务性的,首先在SMR磁盘上按照Zone(区域)划分回收数据,随后在SSD上更新指针元组,则针对此键值对的垃圾回收完成。5. a kind of data organization method based on SSD-SMR disk hybrid key-value storage system according to claim 1, is characterized in that, in SMR disk data storage process, when the capacity of each value is greater than the threshold value of its own effective capacity , and when the load of the storage system is relatively light at this time, garbage collection is performed at this time. The granularity of garbage collection is for each key-value pair, and the garbage collection operation for each key-value pair is transactional. First, the recycling data is divided according to the Zone (region) on the SMR disk, and then the pointer element group is updated on the SSD, and the garbage recycling for this key value pair is completed.6.根据权利要求2所述的一种基于SSD-SMR磁盘混合键值存储系统数据组织方法,其特征在于,当存储系统接收到上层应用发送读请求后,首先遍历在内存的LSM树查找目标key,如果查找失败,随后在SSD上的LSM树查找目标key,如若查找失败,则返回用户错误信息,表明此key并未存储在存储系统中,如若查找到目标key,读取目前key所在的指针节点信息,根据指针节点信息读取SMR磁盘的value,并将数据返回给用户。6. A kind of data organization method based on SSD-SMR disk hybrid key-value storage system according to claim 2, characterized in that, after the storage system receives the read request sent by the upper layer application, it first traverses the LSM tree in the memory to find the target key, if the search fails, then search for the target key in the LSM tree on the SSD. If the search fails, an error message will be returned to the user, indicating that the key is not stored in the storage system. If the target key is found, read the current key. The pointer node information reads the value of the SMR disk according to the pointer information information, and returns the data to the user.7.根据权利要求2所述的一种基于SSD-SMR磁盘混合键值存储系统数据组织方法,其特征在于,当存储系统接收到上层应用发送写请求后,首先遍历在内存和SSD上的LSM树查找目标key,如果查找失败,表明此次是写操作,在SMR磁盘上分配新的空间,将value写入到空闲空间的首地址,随后将地址信息和偏移值、数据大小信息与key组合成四元组,将此元组插入到LSM树中;若查找成功,则此操作是更新操作,此时首先在SMR磁盘上value的原有空间上追加新数据,随后更新LSM树上节点信息。7. A method for organizing data based on SSD-SMR disk hybrid key-value storage system according to claim 2, characterized in that, after the storage system receives a write request sent by an upper-layer application, it first traverses the LSM on the memory and SSD The tree searches for the target key. If the search fails, it indicates that this is a write operation. Allocate new space on the SMR disk, write the value to the first address of the free space, and then write the address information, offset value, data size information and key Combine into a quadruple and insert this tuple into the LSM tree; if the search is successful, this operation is an update operation. At this time, first add new data to the original space of value on the SMR disk, and then update the nodes on the LSM tree information.
CN201811296432.6A2018-11-012018-11-01One kind being based on SSD-SMR disk mixing key assignments memory system data method for organizingPendingCN109521959A (en)

Priority Applications (1)

Application NumberPriority DateFiling DateTitle
CN201811296432.6ACN109521959A (en)2018-11-012018-11-01One kind being based on SSD-SMR disk mixing key assignments memory system data method for organizing

Applications Claiming Priority (1)

Application NumberPriority DateFiling DateTitle
CN201811296432.6ACN109521959A (en)2018-11-012018-11-01One kind being based on SSD-SMR disk mixing key assignments memory system data method for organizing

Publications (1)

Publication NumberPublication Date
CN109521959Atrue CN109521959A (en)2019-03-26

Family

ID=65774281

Family Applications (1)

Application NumberTitlePriority DateFiling Date
CN201811296432.6APendingCN109521959A (en)2018-11-012018-11-01One kind being based on SSD-SMR disk mixing key assignments memory system data method for organizing

Country Status (1)

CountryLink
CN (1)CN109521959A (en)

Cited By (15)

* Cited by examiner, † Cited by third party
Publication numberPriority datePublication dateAssigneeTitle
CN110321367A (en)*2019-06-282019-10-11北京奇艺世纪科技有限公司A kind of date storage method, data query method and relevant apparatus, electronic equipment
CN110347336A (en)*2019-06-102019-10-18华中科技大学A kind of key assignments storage system based on NVM with SSD mixing storage organization
CN111026329A (en)*2019-11-182020-04-17华中科技大学Key value storage system based on host management tile record disk and data processing method
CN111831423A (en)*2019-04-152020-10-27阿里巴巴集团控股有限公司 A method and system for implementing Redis in-memory database on non-volatile memory
WO2021027340A1 (en)*2019-08-132021-02-18华为技术有限公司Storage method and apparatus for key value (kv) and storage device
CN112395212A (en)*2020-11-052021-02-23华中科技大学Method and system for reducing garbage recovery and write amplification of key value separation storage system
CN113094372A (en)*2021-04-162021-07-09三星(中国)半导体有限公司Data access method, data access control device and data access system
CN113900590A (en)*2021-09-282022-01-07重庆紫光华山智安科技有限公司Shingled disk storage method, apparatus, device and medium
CN113901273A (en)*2021-09-072022-01-07联想(北京)有限公司Data processing method and device
CN114490524A (en)*2022-01-202022-05-13中国科学技术大学High-performance distributed key value storage method based on master-slave copy data decoupling
CN116860722A (en)*2023-08-312023-10-10中国科学院软件研究所Database persistence organization optimization method
WO2024060944A1 (en)*2022-09-222024-03-28超聚变数字技术有限公司Key value storage method and system
CN118051643A (en)*2024-02-232024-05-17中国科学院信息工程研究所 A LSM data organization method and device for sparsely distributed metadata
CN119336955A (en)*2024-12-202025-01-21贝格迈思(深圳)技术有限公司 How to operate the learning index tree graph
WO2025066090A1 (en)*2023-09-282025-04-03北京超弦存储器研究院Cxl memory module, controller, method for accessing data and storage system

Citations (4)

* Cited by examiner, † Cited by third party
Publication numberPriority datePublication dateAssigneeTitle
CN102971732A (en)*2010-07-022013-03-13华为技术有限公司System architecture for integrated hierarchical query processing for key/value stores
WO2013174305A1 (en)*2012-05-242013-11-28华为技术有限公司Ssd-based key-value type local storage method and system
CN103559027A (en)*2013-10-222014-02-05北京航空航天大学Design method of separate-storage type key-value storage system
CN107066498A (en)*2016-12-302017-08-18成都华为技术有限公司Key assignments KV storage methods and device

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication numberPriority datePublication dateAssigneeTitle
CN102971732A (en)*2010-07-022013-03-13华为技术有限公司System architecture for integrated hierarchical query processing for key/value stores
WO2013174305A1 (en)*2012-05-242013-11-28华为技术有限公司Ssd-based key-value type local storage method and system
CN103559027A (en)*2013-10-222014-02-05北京航空航天大学Design method of separate-storage type key-value storage system
CN107066498A (en)*2016-12-302017-08-18成都华为技术有限公司Key assignments KV storage methods and device

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
LANYUE LU 等: "WiscKey: Separating Keys from Values in SSD-Conscious Storage", 《ACM TRANSACTIONS ON STORAGE》*
王洋洋 等: "基于SSD-SMR混合存储的LSM树键值存储系统的性能优化", 《计算机科学》*

Cited By (25)

* Cited by examiner, † Cited by third party
Publication numberPriority datePublication dateAssigneeTitle
CN111831423A (en)*2019-04-152020-10-27阿里巴巴集团控股有限公司 A method and system for implementing Redis in-memory database on non-volatile memory
CN110347336A (en)*2019-06-102019-10-18华中科技大学A kind of key assignments storage system based on NVM with SSD mixing storage organization
CN110321367B (en)*2019-06-282021-12-17北京奇艺世纪科技有限公司Data storage method, data query method, related device and electronic equipment
CN110321367A (en)*2019-06-282019-10-11北京奇艺世纪科技有限公司A kind of date storage method, data query method and relevant apparatus, electronic equipment
WO2021027340A1 (en)*2019-08-132021-02-18华为技术有限公司Storage method and apparatus for key value (kv) and storage device
CN112394874A (en)*2019-08-132021-02-23华为技术有限公司Key value KV storage method and device and storage equipment
US11868651B2 (en)2019-08-132024-01-09Huawei Technologies Co., Ltd.Key-value KV storage method and apparatus, and storage device
CN112394874B (en)*2019-08-132024-08-20华为技术有限公司Key value KV storage method and device and storage equipment
CN111026329B (en)*2019-11-182021-04-20华中科技大学 Key-value storage system and data processing method based on host management tile record disk
CN111026329A (en)*2019-11-182020-04-17华中科技大学Key value storage system based on host management tile record disk and data processing method
CN112395212A (en)*2020-11-052021-02-23华中科技大学Method and system for reducing garbage recovery and write amplification of key value separation storage system
CN113094372A (en)*2021-04-162021-07-09三星(中国)半导体有限公司Data access method, data access control device and data access system
US11537582B2 (en)2021-04-162022-12-27Samsung Electronics Co., Ltd.Data access method, a data access control device, and a data access system
CN113901273A (en)*2021-09-072022-01-07联想(北京)有限公司Data processing method and device
CN113900590A (en)*2021-09-282022-01-07重庆紫光华山智安科技有限公司Shingled disk storage method, apparatus, device and medium
CN113900590B (en)*2021-09-282023-01-31重庆紫光华山智安科技有限公司Shingled disk storage method, shingled disk storage device, shingled disk storage equipment and shingled disk storage medium
CN114490524A (en)*2022-01-202022-05-13中国科学技术大学High-performance distributed key value storage method based on master-slave copy data decoupling
CN114490524B (en)*2022-01-202024-07-30中国科学技术大学High-performance distributed key value storage method based on master-slave copy data decoupling
WO2024060944A1 (en)*2022-09-222024-03-28超聚变数字技术有限公司Key value storage method and system
CN116860722B (en)*2023-08-312023-11-14中国科学院软件研究所Database persistence organization optimization method
CN116860722A (en)*2023-08-312023-10-10中国科学院软件研究所Database persistence organization optimization method
WO2025066090A1 (en)*2023-09-282025-04-03北京超弦存储器研究院Cxl memory module, controller, method for accessing data and storage system
CN118051643A (en)*2024-02-232024-05-17中国科学院信息工程研究所 A LSM data organization method and device for sparsely distributed metadata
CN119336955A (en)*2024-12-202025-01-21贝格迈思(深圳)技术有限公司 How to operate the learning index tree graph
CN119336955B (en)*2024-12-202025-03-04贝格迈思(深圳)技术有限公司Operation method of learning index tree graph

Similar Documents

PublicationPublication DateTitle
CN109521959A (en)One kind being based on SSD-SMR disk mixing key assignments memory system data method for organizing
US11741053B2 (en)Data management system, method, terminal and medium based on hybrid storage
CN110825748B (en)High-performance and easily-expandable key value storage method by utilizing differentiated indexing mechanism
CN103019953B (en)Construction system and construction method for metadata
CN105912687B (en)Magnanimity distributed data base storage unit
CN103106286B (en)Method and device for managing metadata
CN101488153A (en)Method for implementing high-capacity flash memory file system in embedded type Linux
CN104899297A (en) Hybrid index structure with storage awareness
CN103176754A (en)Reading and storing method for massive amounts of small files
CN110309233A (en)Method, apparatus, server and the storage medium of data storage
US20140244606A1 (en)Method, apparatus and system for storing, reading the directory index
CN106599040A (en)Layered indexing method and search method for cloud storage
CN111352586B (en)Directory aggregation method, device, equipment and medium for accelerating file reading and writing
CN106354805A (en)Optimization method and system for searching and caching distribution storage system NoSQL
CN109933564A (en) File system management method, device, terminal and medium for fast rollback based on linked list and N-ary tree structure
CN117687970A (en)Metadata retrieval method and device, electronic equipment and storage medium
CN103049574A (en)Key value system and key value method for implementation of dynamic duplicates of documents
WO2024041376A1 (en)Distributed graph data processing system, method, apparatus and device, and storage medium
CN109828975A (en)A kind of extensive quick account book access system based on block chain
CN1255748C (en)Metadata hierarchy management method and system of storage virtualization system
Wang et al.Leaderkv: Improving read performance of kv stores via learned index and decoupled kv table
CN116821126A (en) Key-value pair storage method, controller and system based on multi-stage dynamic migration
CN114415971B (en)Data processing method and device
Cai et al.The Embedded IoT Time Series Database for Hybrid Solid‐State Storage System
US12248517B2 (en)Method, apparatus, device, and storage medium for data processing of graph database

Legal Events

DateCodeTitleDescription
PB01Publication
PB01Publication
SE01Entry into force of request for substantive examination
SE01Entry into force of request for substantive examination
RJ01Rejection of invention patent application after publication

Application publication date:20190326

RJ01Rejection of invention patent application after publication

[8]ページ先頭

©2009-2025 Movatter.jp