IBM Cloud Object Storage is an object storage system consisting of software delivered on individual hardware elements that can scale to exabytes of data. The architecture of Cloud Object Storage is based upon information dispersal across geographies, presenting a single, global namespace for objects. As one of the first object storage systems deployed in cloud access environments, IBM Cloud Object Storage has continued to add capabilities and additional usage cases. Announcements in July 2020 increased performance by 300%, focused on AI usage, and added direct block storage for backend.
Cloud Object Storage stores objects that have been broken into slices across dispersed elements of storage that can be local or remote. Three elements are used in the system: COS Manager to manage the system, Accesser to break data into slices and disperse slices for storage and to retrieve data and recombine for host access, and Slicestor to provide storage control for attached storage devices. The system can be delivered as individual hardware components with the software preloaded or as OVF virtual machine files for execution on standard servers. HTTP/REST is used with either a custom API or Amazon S3 API for object access. Cloud Object Storage also has an interface to HDFS for Hadoop environments and supports WebDAV.
Regulatory compliance capability was added with support for Compliance Enabled Vaults with WORM mode, retention management, audit trails, and legal hold settings, either through GUI controls or through the REST interface using S3.
Cloud Object Storage stores objects using direct block I/O. A flat architecture is used with no central database. Multiple, simultaneous reads and writes can be in progress both from the client side and with the backend storage. Metadata is appended to the object data and stored in the storage nodes. The application manages the usage of the metadata.