CN112379825B - Distributed data storage method and device based on data feature sub-pools - Google Patents
Distributed data storage method and device based on data feature sub-pools Download PDFInfo
- Publication number
- CN112379825B CN112379825B CN201910906536.2A CN201910906536A CN112379825B CN 112379825 B CN112379825 B CN 112379825B CN 201910906536 A CN201910906536 A CN 201910906536A CN 112379825 B CN112379825 B CN 112379825B
- Authority
- CN
- China
- Prior art keywords
- data
- optimization
- pool
- feature
- strategy
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/06—Digital input from, or digital output to, record carriers, e.g. RAID, emulated record carriers or networked record carriers
- G06F3/0601—Interfaces specially adapted for storage systems
- G06F3/0668—Interfaces specially adapted for storage systems adopting a particular infrastructure
- G06F3/067—Distributed or networked storage systems, e.g. storage area networks [SAN], network attached storage [NAS]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/06—Digital input from, or digital output to, record carriers, e.g. RAID, emulated record carriers or networked record carriers
- G06F3/0601—Interfaces specially adapted for storage systems
- G06F3/0602—Interfaces specially adapted for storage systems specifically adapted to achieve a particular effect
- G06F3/061—Improving I/O performance
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F3/00—Input arrangements for transferring data to be processed into a form capable of being handled by the computer; Output arrangements for transferring data from processing unit to output unit, e.g. interface arrangements
- G06F3/06—Digital input from, or digital output to, record carriers, e.g. RAID, emulated record carriers or networked record carriers
- G06F3/0601—Interfaces specially adapted for storage systems
- G06F3/0628—Interfaces specially adapted for storage systems making use of a particular technique
- G06F3/0638—Organizing or formatting or addressing of data
- G06F3/0644—Management of space entities, e.g. partitions, extents, pools
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Human Computer Interaction (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention relates to a data storage method and device, belongs to the field of data storage, and particularly relates to a distributed data storage method and device based on data feature sub-pools. The method comprises the following steps: a characteristic division step, wherein a fault domain algorithm is planned in advance according to the characteristics of the service data, and the algorithm can automatically divide the data into different characteristic types; a strategy making step, namely making a distributed storage data optimization strategy aiming at different characteristic types; a strategy loading step, wherein the storage equipment loads a data storage optimization strategy corresponding to the data characteristic type of the sub-pool according to the data characteristic type of the sub-pool; and an optimization executing step, wherein the object storage equipment executes the optimization operation according to the related action defined by the data storage optimization strategy. Therefore, the invention has the following advantages: the sequential writing performance of the node disk where the sequential characteristic pool is located is basically full bandwidth, and the disk where the random characteristic pool is located also basically reaches the limit random performance, so that the data operation efficiency of distributed storage is obviously improved.
Description
Technical Field
The invention relates to a data storage method and device, belongs to the field of data storage, and particularly relates to a distributed data storage method and device based on data feature sub-pools.
Background
Distributed storage is a data storage technology, and in short, data is stored in a plurality of storage servers in a distributed manner, and these distributed storage resources constitute a virtual storage device, and actually data is stored in each corner of the servers in a distributed manner.
In a traditional network storage system, a centralized storage server is used for storing all data, and the storage server becomes a bottleneck of system performance and is also a focus of reliability and safety. The distributed storage system shares the storage load by using a plurality of storage servers and positions the storage information by using the position server, thereby effectively improving the reliability, the availability and the access efficiency of the system.
Distributed storage systems are widely accepted in the global scope at present, and compared with the traditional storage system, the distributed storage system has the following application advantages:
high performance: the distributed storage can efficiently manage read cache and write cache, and supports automatic hierarchical storage, so as to improve the response speed of the system by mapping data in a hot spot area into high-speed storage.
A plurality of pairs of the technology: the distributed storage adopts a multi-copy backup mechanism and meets the requirements of users on different reliability by using modes of mirroring, banding, distributed verification and the like.
Disaster tolerance and backup: the distributed storage supports multi-time point snapshot backup, can simultaneously extract a plurality of time point samples and recover at the same time, reduces the difficulty of fault location, and ensures high data safety and availability by combining a periodic incremental backup mechanism.
Elastic expansion: due to the reasonable distributed architecture, the capacity can be estimated, the calculation, the storage capacity and the performance can be expanded flexibly, the expanded old data can be automatically migrated to a new node, the load balance is realized, and the single-point overheating is avoided.
Ceph, as a unified distributed storage system in the prior art, is different from other distributed systems in that it adopts a pause (controlled Replication Under Scalable hashing) algorithm to make the storage locations of data all calculated rather than querying a dedicated metadata server.
The method has the advantages that the disaster tolerance domain isolation is well considered by the Crush on the basis of the consistent Hash algorithm, and the copy placement rules of various loads can be realized, such as cross-machine room and rack sensing. Meanwhile, the Crush algorithm supports two data redundancy modes of copy and EC.
The Crush algorithm is one of the first cores of Ceph, and is the place where Ceph is the most attractive. The disaster tolerance domain isolation is well considered by the Crush on the basis of consistent hash, and the copy placement rules of various loads can be realized, such as cross-machine room and rack sensing. Meanwhile, the Crush algorithm supports two data redundancy modes of a copy and an EC, provides a fault domain isolation mode of a host, a server rack and a data center, can reasonably plan a fault domain according to the physical structure of a storage server, fully considers iterative deployment and capacity reduction and expansion of hardware in the actual production process, has quite good expandability as far as possible, and can still ensure good load balance and reliable availability under the condition of thousands of OSD. However, this is more of a theoretical point of view and no one has yet given the results of testing in a production environment on a scale of several PB. In summary, the brush algorithm is still one of the best data distribution algorithms tested in practice at present.
Ceph abandons the traditional centralized storage metadata addressing mode, and based on the number of data copies defined by an administrator, the physical storage positions of the copies are designated by a Crush algorithm to separate fault domains, so that the data distribution is balanced, the parallelism is high, and the strong data consistency is supported. The Ceph can tolerate various fault scenes, considers the isolation of disaster tolerance domains, can realize the copy placement rules of various loads, and automatically tries to repair the loads in parallel after the faults occur.
The Ceph performs pool division according to the performance of the equipment, divides the high-speed equipment and the low-speed equipment into pools, establishes a high-performance pool formed by the high-speed equipment and a low-performance pool formed by the low-speed equipment, places data in the high-performance pool for the service with high performance requirement, and places data in the low-performance pool for the service with low performance requirement. Because the Ceph is a distributed storage system, a data redundancy mechanism and a data distribution algorithm Crush of the Ceph are set according to Pool, if a Pool is formed by mixing high-speed equipment and low-speed equipment, the data distribution algorithm Crush of the Ceph requires that the lowest level of data distribution or failure must be a disk level, and regardless of the data redundancy technology, high reliability of data needs to be guaranteed, that is, data distribution is distributed to at least 2 disks or more, which may cause a barrel effect, that is, a piece of data must be written into all redundant data to be returned, and the performance is limited by the slow-speed equipment. Therefore, the performance barrel effect can be well solved by carrying out pool division according to the performance of the equipment. However: high-speed devices are more expensive than low-speed devices, and the capacity of high-performance pools is necessarily limited. On the other hand, low-speed devices have performance gaps with respect to high-speed devices, but if not optimized, the performance of the low-speed devices in the face of different types of data cannot be exerted.
In the prior art, logical pooling also exists, and the pooling scheme mainly performs pooling according to business logic, so that the advantage is that the pooling logic is relatively simple, and the influence of other factors on the performance of the storage pool is not needed. Although the logical pooling scheme is simple, the scheme has no capability of adapting to the requirements of the services on performance, and cannot sense the data characteristics of the services, which means that the scheme cannot perform the most reasonable optimization on different services, for example, some services have certain requirements on delay, but the requirements are not very high, and according to the pooling scheme, the services must be put into a high-performance pool, so that the construction cost of the whole storage is increased.
Disclosure of Invention
The invention mainly solves the technical problems in the prior art and provides a distributed data storage algorithm and a distributed data storage system based on data feature sub-pools.
The technical problem of the invention is mainly solved by the following technical scheme:
a distributed data storage method based on data feature sub-pools comprises the following steps:
a characteristic division step, namely automatically planning a fault domain algorithm in advance according to the characteristics of the service data, and dividing the data into different characteristic types;
a strategy making step, namely making a distributed storage data optimization strategy aiming at the characteristic types of different services;
a strategy loading step, wherein the object storage equipment loads a data storage optimization strategy corresponding to the data characteristic type of the sub-pool according to the data characteristic type of the sub-pool;
and an optimization executing step, wherein the object storage equipment executes the optimization operation according to the related action defined by the data storage optimization strategy.
Preferably, in the above-mentioned distributed data storage method based on data feature classification, in the feature classification step,
and automatically dividing the data into a sequence characteristic and a random characteristic according to the service io type.
Preferably, in the policy making step, two sets of data feature optimization policies for the sequential features and the random features are made in combination with the electrical characteristics of hardware, the hardware performance of the server, and the operating system at the OSD level.
Preferably, in the distributed data storage method based on data feature pooling, the random feature optimization policy algorithm includes one or more of the following operations: automatically adjusting IO concurrent processing threads according to the number of the physical CPUs; aggregating continuous small IO at a client layer, a software abstraction layer and a kernel driving layer; automatically calculating a buffer; starting a data caching module; writing data to the high-speed device; shortening an IO path; and reduces system interruptions.
Preferably, the above-mentioned distributed data storage method based on data feature pooling further includes: and a data monitoring step, namely expanding persistent records aiming at data feature sub-pools in a daemon process of the distributed storage monitor, automatically expanding the attributes of the data storage groups according to specific sequences and random features, and issuing the sub-pool characteristics to corresponding storage equipment through creating the data storage groups.
Preferably, in the above distributed data storage method based on data feature pooling, in the feature partitioning step, a data feature type parameter may be specified when a command line or a client creates a Pool.
A distributed data storage system based on pooling of data characteristics, comprising:
the characteristic division module is used for planning a fault domain algorithm in advance according to the characteristics of the service data and dividing the data into different characteristic types;
the strategy making module is used for making a distributed storage data optimization strategy aiming at different characteristic types;
the strategy loading module loads a data storage optimization strategy corresponding to the object storage equipment according to the data feature types of the sub-pools;
and the storage equipment automatically executes the optimization operation according to the related action defined by the data storage optimization strategy.
Preferably, the distributed data storage system based on data feature pooling further includes:
in the feature dividing module, data is divided into sequential features and random features according to service types, wherein the sequential features comprise streaming media data, and the random features comprise database types and system disk data.
Preferably, in the above distributed data storage system based on data feature pooling, two sets of data feature optimization strategies for the sequential features and the random features are formulated by the strategy formulation module at the OSD layer in combination with the electrical characteristics of hardware, the hardware performance of the server, and the operating system.
Therefore, the invention has the following advantages: the sequential writing performance of the node disk where the sequential characteristic pool is located is basically full bandwidth, and the disk where the random characteristic pool is located also basically reaches the limit random performance, so that the data operation efficiency of distributed storage is obviously improved.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a system diagram of the present invention;
Detailed Description
The technical scheme of the invention is further specifically described by the following embodiments and the accompanying drawings.
Example (b):
as shown in fig. 1, the distributed data storage method based on data feature pooling of this embodiment includes the following steps:
a characteristic division step, namely automatically planning a fault domain algorithm in advance according to the characteristics of the service data, and dividing the data into different characteristic types; in this step, the data feature type parameter may be specified when the command line or the client creates Pool.
A strategy making step, namely making a distributed storage data optimization strategy aiming at different characteristic types; in the embodiment, two sets of data feature optimization strategies aiming at sequence features and random features are made by combining the electrical characteristics of hardware, the hardware performance of a server and an operating system at the OSD level.
A strategy loading step, wherein the object storage equipment loads a data storage optimization strategy corresponding to the data characteristic type of the sub-pool according to the data characteristic type of the sub-pool;
and an optimization executing step, wherein the object storage equipment executes the optimization operation according to the related action defined by the data storage optimization strategy.
As a preferred mode, this embodiment further includes a data monitoring step, where persistent records for data feature pools are expanded in the distributed storage monitoring daemon, attributes of data storage groups are expanded according to a specific order and random features, and pool characteristics are created and issued to corresponding object storage devices through the data storage groups.
In this embodiment, data is divided into a sequential feature and a random feature according to a service type, where the sequential feature includes streaming media data, and the random feature includes a database class and system disk data.
The sequential feature optimization strategy includes one or more of the following operations: the algorithm automatically adjusts the memory cache, adjusts the memory allocation management unit and adjusts the size of each buffer.
The random feature optimization strategy includes one or more of the following operations: the algorithm distributes random small io to a high-speed disk according to the size of io issued by the service, so that delay is reduced, and iops is improved. And combining the continuous small ios within 5s, and issuing the combined small ios to the low-speed disk, so that the seek time of the low-speed disk is reduced, and the write bandwidth of the low-speed disk is increased as much as possible.
After the method is adopted, two sets of data feature optimization strategies aiming at the sequence features and the random features are worked out on the OSD layer by combining the electrical characteristics of hardware, the hardware performance of the server, the operating system and other factors.
Different optimization strategies exert the performance of respective data characteristics to the limit;
the Monitor layer of this embodiment needs to expand persistent records for data feature sub-pools to prevent failure of data feature sub-pools due to Monitor abnormality, and the Monitor needs to expand attributes of the PG according to a specific sequence and random features and issue sub-pool characteristics to corresponding OSDs by creating data distribution groups;
in this embodiment, a user needs to plan a Crushmap in advance according to the characteristics of service data, for example, static data such as video and audio can be classified into sequential characteristics, and services such as database and system disk can be classified into random characteristics; when Pool is created through a command line or a client, data characteristic type parameters need to be specified;
in this embodiment, when receiving a Pool creation instruction, the Monitor first creates Pool metadata, then enters into related operations for creating a PG layer, and finally notifies the OSD to create a PG instance;
in this embodiment, when receiving the PG creation instruction, the OSD loads a data feature optimization policy according to a data feature type carried by the PG, and executes a related optimization action according to a related action defined by the policy.
As shown in fig. 2, the distributed data storage system based on data feature pooling provided for this embodiment includes:
the characteristic division module is used for planning a fault domain algorithm in advance according to the characteristics of the service data and dividing the data into different characteristic types;
the strategy making module is used for making a distributed storage data optimization strategy aiming at different characteristic types;
the strategy loading module loads a data storage optimization strategy corresponding to the object storage equipment according to the data feature types of the sub-pools;
and the object storage equipment executes optimization operation according to the related action defined by the data storage optimization strategy.
In this embodiment, in the feature dividing module, data is divided into sequential features and random features according to service types, where the sequential features include streaming media data, and the random features include database types and system disk data.
In this embodiment, the policy making algorithm module makes two sets of data feature optimization policies for the sequential features and the random features at the OSD level in combination with the electrical characteristics of the hardware, the hardware performance of the server, and the operating system.
Wherein, the sequence characteristic optimization strategy is as follows: the algorithm automatically identifies continuous small ios within 5 seconds, automatically merges the small ios, writes the continuous io into low-speed setting, and improves io writing bandwidth;
wherein, the random feature optimization strategy is as follows: the algorithm automatically identifies the random small io and writes the random small io into the high-speed disk, the writing delay is obviously reduced, the written io is cached, and when random reading occurs, the hit data is directly read and written from the random feature pool, so that the reading time is shortened, and the delay is reduced.
After the technical scheme is adopted, the technical scheme of the embodiment brings the direct effect that the sequential writing performance of the disks of the nodes where the sequential feature pool is located is basically at the full bandwidth, and the disks where the feature pool is written along also basically reach the limit random performance.
The specific embodiments described herein are merely illustrative of the spirit of the invention. Various modifications or additions may be made to the described embodiments or alternatives may be employed by those skilled in the art without departing from the spirit or ambit of the invention as defined in the appended claims.
Claims (8)
1. A distributed data storage method based on data feature sub-pools is characterized by comprising the following steps:
a characteristic division step, namely planning a fault domain algorithm in advance according to the characteristics of the service data, and dividing the data into different characteristic types;
a strategy making step, namely making a distributed storage data optimization strategy aiming at different characteristic types;
a strategy loading step, wherein the object storage equipment loads a data storage optimization strategy corresponding to the data characteristic type of the sub-pool according to the data characteristic type of the sub-pool;
the optimization execution step, the object storage equipment executes the optimization operation according to the related action defined by the data storage optimization strategy;
in the strategy making step, two sets of data characteristic optimization strategies aiming at sequence characteristics and random characteristics are automatically made on the OSD level by combining the electrical characteristics of hardware, the hardware performance of a server and an operating system; the random feature optimization strategy includes one or more of the following operations: aligning the minimum unit of disk space management with the block size of a physical disk; adjusting the depth of the IO queue; adjusting IO concurrent processing threads according to the number of the physical CPUs; aggregating continuous small IO at a client layer, a software abstraction layer and a kernel driving layer; promoting a buffer; starting a data caching module; writing data to the high-speed device; shortening an IO path; system interruption is reduced;
the Monitor layer expands persistent records aiming at the data feature sub-pools, so that the data feature sub-pools are prevented from being invalid due to Monitor abnormity, and the Monitor needs to expand the attribute of PG according to a specific sequence and random features and issue the sub-pool characteristics to corresponding OSD (on screen display) through creating a data distribution group; planning Crushmap in advance according to the characteristics of service data, and specifying data characteristic type parameters when Pool is created through a command line or a client;
when receiving a Pool creating instruction, a Monitor first creates Pool metadata, then enters related operations for creating a PG layer, and finally informs the corresponding OSD to create a PG instance;
when receiving a PG creating instruction, the OSD loads a data feature optimization strategy according to the data feature type carried by the PG and executes related optimization actions according to related actions defined by the strategy.
2. The distributed data storage method based on data feature pooling of claim 1, wherein in said feature partitioning step,
and dividing data into a sequence characteristic and a random characteristic according to the service type, wherein the sequence characteristic comprises streaming media data, and the random characteristic comprises a database class and system disk data.
3. The method of claim 1, wherein the sequential feature optimization strategy comprises one or more of the following operations: adjusting the minimum unit of physical disk space management, adjusting memory cache, adding a network card MTU, adjusting a memory allocation management unit and adjusting the size of each buffer.
4. The method of claim 1, further comprising: and a data monitoring step, namely expanding persistent records aiming at data feature sub-pools in a daemon process of the distributed storage monitor, expanding the attributes of the data storage groups according to a specific sequence and random features, and creating and issuing the sub-pool characteristics to corresponding object storage equipment through the data storage groups.
5. The distributed data storage method based on data feature pooling of claim 1, wherein in said feature partitioning step, data feature type parameters are further specified when a command line or a client creates a Pool.
6. A distributed data storage system based on pooling of data characteristics, comprising:
the characteristic division module is used for planning a fault domain algorithm in advance according to the characteristics of the service data and dividing the data into different characteristic types;
the strategy making module is used for making a distributed storage data optimization strategy aiming at different characteristic types;
the strategy loading module loads a data storage optimization strategy corresponding to the object storage equipment according to the data feature types of the sub-pools;
the optimization execution module is used for executing optimization operation by the object storage equipment according to the related action defined by the data storage optimization strategy; the Monitor layer expands persistent records aiming at the data feature sub-pools, so that the data feature sub-pools are prevented from being invalid due to Monitor abnormity, and the Monitor needs to expand the attribute of PG according to a specific sequence and random features and issue the sub-pool characteristics to corresponding OSD (on screen display) through creating a data distribution group; planning Crushmap in advance according to the characteristics of service data, and specifying data characteristic type parameters when Pool is created through a command line or a client;
when receiving a Pool creating instruction, a Monitor first creates Pool metadata, then enters related operations for creating a PG layer, and finally informs the corresponding OSD to create a PG instance;
when receiving a PG creating instruction, the OSD loads a data feature optimization strategy according to the data feature type carried by the PG and executes related optimization actions according to related actions defined by the strategy.
7. The data feature pooling-based distributed data storage system of claim 6, further comprising:
in the feature dividing module, data is divided into sequential features and random features according to service types, wherein the sequential features comprise streaming media data, and the random features comprise database types and system disk data.
8. The distributed data storage system according to claim 6, wherein the policy making module combines the electrical characteristics of hardware, the hardware performance of the server, and the operating system at the OSD level to make two sets of data feature optimization policies for the sequential features and the random features.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910906536.2A CN112379825B (en) | 2019-09-24 | 2019-09-24 | Distributed data storage method and device based on data feature sub-pools |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910906536.2A CN112379825B (en) | 2019-09-24 | 2019-09-24 | Distributed data storage method and device based on data feature sub-pools |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112379825A CN112379825A (en) | 2021-02-19 |
CN112379825B true CN112379825B (en) | 2021-07-06 |
Family
ID=74585995
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910906536.2A Active CN112379825B (en) | 2019-09-24 | 2019-09-24 | Distributed data storage method and device based on data feature sub-pools |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112379825B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112905345B (en) * | 2021-02-23 | 2024-04-05 | 深圳市网心科技有限公司 | Task allocation method, distributed storage system and server |
CN113296706B (en) * | 2021-05-27 | 2024-04-09 | 上海仪电(集团)有限公司中央研究院 | Ceph system data cleaning method, device, equipment and medium |
CN115543222B (en) * | 2022-11-30 | 2023-03-10 | 苏州浪潮智能科技有限公司 | Storage optimization method, system, equipment and readable storage medium |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103425675A (en) * | 2012-05-17 | 2013-12-04 | 南京中兴力维软件有限公司 | Monitoring data storage method and device used for centralized monitoring management system |
CN106537359A (en) * | 2014-07-15 | 2017-03-22 | 三星电子株式会社 | Electronic device and method for managing memory of electronic device |
CN107807796A (en) * | 2017-11-17 | 2018-03-16 | 北京联想超融合科技有限公司 | A kind of data hierarchy method, terminal and system based on super fusion storage system |
CN108920095A (en) * | 2018-06-06 | 2018-11-30 | 深圳市脉山龙信息技术股份有限公司 | A kind of data store optimization method and apparatus based on CRUSH |
CN109947363A (en) * | 2018-12-11 | 2019-06-28 | 深圳供电局有限公司 | A kind of data cache method of distributed memory system |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2014031618A2 (en) * | 2012-08-22 | 2014-02-27 | Bitvore Corp. | Data relationships storage platform |
-
2019
- 2019-09-24 CN CN201910906536.2A patent/CN112379825B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103425675A (en) * | 2012-05-17 | 2013-12-04 | 南京中兴力维软件有限公司 | Monitoring data storage method and device used for centralized monitoring management system |
CN106537359A (en) * | 2014-07-15 | 2017-03-22 | 三星电子株式会社 | Electronic device and method for managing memory of electronic device |
CN107807796A (en) * | 2017-11-17 | 2018-03-16 | 北京联想超融合科技有限公司 | A kind of data hierarchy method, terminal and system based on super fusion storage system |
CN108920095A (en) * | 2018-06-06 | 2018-11-30 | 深圳市脉山龙信息技术股份有限公司 | A kind of data store optimization method and apparatus based on CRUSH |
CN109947363A (en) * | 2018-12-11 | 2019-06-28 | 深圳供电局有限公司 | A kind of data cache method of distributed memory system |
Also Published As
Publication number | Publication date |
---|---|
CN112379825A (en) | 2021-02-19 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US10042869B1 (en) | Method for data transfer between compute clusters and file system | |
US8307159B2 (en) | System and method for providing performance-enhanced rebuild of a solid-state drive (SSD) in a solid-state drive hard disk drive (SSD HDD) redundant array of inexpensive disks 1 (RAID 1) pair | |
US7971013B2 (en) | Compensating for write speed differences between mirroring storage devices by striping | |
US20220137849A1 (en) | Fragment Management Method and Fragment Management Apparatus | |
US7032070B2 (en) | Method for partial data reallocation in a storage system | |
US8639876B2 (en) | Extent allocation in thinly provisioned storage environment | |
CN112379825B (en) | Distributed data storage method and device based on data feature sub-pools | |
WO2013057764A1 (en) | Storage system | |
CN105657066A (en) | Load rebalance method and device used for storage system | |
CN103929500A (en) | Method for data fragmentation of distributed storage system | |
US8032784B2 (en) | Duplication combination management program, duplication combination management apparatus, and duplication combination management method | |
WO2015015550A1 (en) | Computer system and control method | |
US10564865B2 (en) | Lockless parity management in a distributed data storage system | |
CN103516549B (en) | A kind of file system metadata log mechanism based on shared object storage | |
JP2000099282A (en) | File management system | |
JP2007323224A (en) | Flash memory storage system | |
US20140075111A1 (en) | Block Level Management with Service Level Agreement | |
US20100306488A1 (en) | Performing mirroring of a logical storage unit | |
US8195877B2 (en) | Changing the redundancy protection for data associated with a file | |
CN102164165B (en) | Management method and device for network storage system | |
JP4261532B2 (en) | Logical disk management method and virtualization apparatus | |
JP2016118821A (en) | Storage management device, storage management method and storage management program | |
CN111736764B (en) | Storage system of database all-in-one machine and data request processing method and device | |
US7725654B2 (en) | Affecting a caching algorithm used by a cache of storage system | |
US6934803B2 (en) | Methods and structure for multi-drive mirroring in a resource constrained raid controller |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant | ||
TR01 | Transfer of patent right |
Effective date of registration: 20211220 Address after: Room 203, building 5, No. 13, Fuqian 1st Street, Tianzhu District, Shunyi District, Beijing 100037 Patentee after: Beijing Urban Construction Intelligent Control Technology Co.,Ltd. Address before: 100037 No. 5 Fuchengmen North Street, Xicheng District, Beijing Patentee before: BEIJING URBAN CONSTRUCTION DESIGN & DEVELOPMENT GROUP Co.,Ltd. |
|
TR01 | Transfer of patent right |