Object-based Storage-Alibaba Cloud Developer Community

1 Preface

I often see articles on object storage on the Internet, but there has never been any article that can make me understand what object storage is. Recently, I got a chance that the boss wanted us to tell him what object storage is. In order to make 5 pages of PPT for the boss, the author referred to a large number of documents and sorted them into articles (the following articles are for reference only, it is not an authoritative work. If you think there is something wrong, please correct it). 2. Object storage

2.1 The concept of Obeject-based Storage (personal arrangement, not authority)

is a new storage architecture. Object storage is different from Block and File-level storage. It uses an elastic and customizable data object as the basic storage unit, the index is based on the unique ID of the data object and forms a virtual file organization structure. Combined with cluster technology, the storage devices of a group of servers are integrated into the storage pool. Software data can be automatically distributed in the cluster, and the storage capacity can reach TB or more than EB, object storage features high I/O, convenient sharing, high scalability, high reliability, and high performance. 2.2 concept of OSD OSD is the short name of Object-based Storage Device, which is the short name of Object-Oriented Storage technology Device. It refers to the basic logical unit in the cluster that runs Object-oriented Storage technology. 2.3 Object concept

-Data refers to the user to access the Data record or data-MetaData is used to locate the real location, historical data, resource search, file Record etc. function-OID that Object ID IS the Data Object unique identification of (equivalent to human ID card ID) - Attributes Data description attribute-Object By the Data/OID/Metadat/Attributes Composition, is the minimum data storage unit defined by object storage. 2.4 Comparison between block storage and object storage

-The minimum storage unit of Block storage is a data Block, while the basic storage unit of Object storage is an Object.-Block storage searches for a unique data block based on inode, while object storage searches for a unique data object based on OID-the data block of block storage and the data block of object storage are both flat structures (I/O)

-Block storage blocks are divided into data blocks.and folder blocks. Folder blocks organize file structures by recording inode and repositioning data blocks.-similarly, object Storage relocates data objects through GID to organize virtual file structures (that is, directory tree) 2.5 Object Type

the ANSI T10 SCSI OSD standard defines four different objects:

-Root object, OSD itself-user object, object created by SCSI commands from applications or clients-collection object, a group of user objects, such as all input items. mp3 object collection or all objects-partitioned objects, user objects and collection containers that share security or spatial management features (such as coordination and keys) 2.6 System architecture

Object Storage is generally divided into three parts: Metadata Server Cluster-Metadata Cluster, Object Storage Cluster-Object Storage Cluster for locating the location where data is accessed, Object Storage Clients-data demander, go to the location where the data is located in the Metadata Server and log on to the Object Storage. 2.7 features of object storage -High I/O read/write performance, high-speed direct access to disks by SAN-convenient sharing, distributed sharing of NAS-high scalability, using ordinary x86 servers, supports 10 to 1,000 servers and supports TB to PB scale-out. -High reliability, no single point of failure, multi-data Copy, automatic management, automatic repair. -High Performance, balanced data distribution, and high parallelism. For objects storage and block storage, don't need the metadata server. 3. Comparison and analysis of storage types

3.1 block storage -Customized by professional manufacturers, the entire storage is equivalent to a large hard disk (users can format it into a file system as needed)-use data blocks as the minimum storage unit, and use inode to directly locate data (determine IO performance)-Use the FC-SAN protocol as the communication medium-the advantage is good I/O performance-the disadvantage is poor scalability, inconvenient data sharing, and high device price-scale up expansion mode, single-point fault tolerance through multiple controllers and multiple power supplies-representative products: SAN 3.2 File Storage -It is equivalent to a shared folder (Samba) or NFS, implemented by software and server hardware-using files as storage units and locating data in directory structure (determining IO performance)-Advantage is that data sharing convenience-shortcomings is the transport protocol overhead large, I/O performance difference-default scale up the extension way, with lvs can realize scale out extension way (storage does not have its own high reliability ability)-Representative products: NAS 3.3 object storage -A type of software-defined storage that is implemented by a storage software system and server hardware. Data objects are used as the minimum storage unit, and data is directly indexed by the data object ID (which determines the I/O performance).-It has the advantages of good I/O performance, low protocol overhead, convenient sharing, and moderate cost.-scale out is an extended mode, and the storage itself achieves high reliability, high performance, and high scalability.-it represents the Redhat Cept and OpenStack Swift of the product. 3.4 Comparison tables of three types of storage

Selected, One-Stop Store for Enterprise Applications
Support various scenarios to meet companies' needs at different stages of development

Start Building Today with a Free Trial to 50+ Products

Learn and experience the power of Alibaba Cloud.

Sign Up Now