High Performance
(1) The file storage is able to improve the concurrent access performance by employing the intelligent Stripe technology, the setting of single file across nodes and the storage system distribution.
(2) The smart data hierarchy is able to optimize performance and capacity balance by automatically performing storage tiering in flash pool, HDD pool, and tape based on relevant policies.
(3) The intelligent prefetching mechanism, which prefetches files by predicting file access patterns to reduce read and write latency.
(4) The distributed block-level lock management, consisting of file and directory levels, allows for maximum concurrent access.
(5) The said distributed storage supports client data cache, which can set different cache sizes according to the requirements at different nodes, so that enables the global cache improving the performance of cluster read and write.
(6) The ​Qos is capable of constraining background functionality such as reconstruction or asynchronous replication. Meanwhile, the flexible policy settings such as day-of-week and time-of-day are available.
(7) The high-availability write cache (HAWC) is capable of improving synchronous write performance, while the local read-only cache (LROC) is able to achieve the cache acceleration.
(8) Furthermore, the said distributed storage supports the GPU Direct(GDS) function, meaning that the data can be read from storage and put into GPU application memory, so that it can bypass the CPU memory for better performance.
High Scalability
(1) The cutting-edge distributed and fully symmetric architecture can achieve linear growth in performance and capacity, potentially supporting maximum cluster sizes of tens of thousands of nodes and IO throughput at the TB level per second, leading to super-high throughput and low latency access.
(2) The unified namespace supports the creation of a single file system and a single global namespace as well as a multi-site unified namespace to access data across different storage and locations at any time and anywhere.
(3) In case of the online business, the nodes can be added in or removed from the cluster, the disks can be added in or deleted from the file system, and the file system attributes (the maximum number of inodes, etc.) can be modified. After expanding the capacity of the whole storage system, the data can be balanced automatically or manually.
(4) The said distributed storage is capable of using the tiered storage capabilities of ILM to store all data in different media of a unified storage space, such as NVMe SSD, SAS SSD/HDD, and tape library, etc.
High Reliability
(1) The said distributed storage supports functions, such as quota, snapshot protection, synchronous replication, asynchronous replication and global data sharing, etc., to keep the system safe, stable and reliable.
(2) It also supports functions such as dual-active, synchronous replication and asynchronous replication to enable fault automatic switching and recovery, and ensure business continuity and data security.
(3) In addition, the said distributed storage supports various security protection modes to meet business and regulatory requirements by means of authentication, encryption, security and replication options.
(4) ​The system has no single point of failure, ensuring that the front-end operations do not stop and data is not lost in the event of a failure of not less than one storage node.
(5) The storage system can be reconstructed automatically in case of the failure of hard disks or nodes, ensuring the balance between the performance and availability of the storage system.
(6) ​The said distributed storage supports end-to-end consistency checks and error write checks, so that the results of disk operations are directly reflected to the user/client and the silent data corruption problems can be detected and corrected.
(7) Furthermore, the said distributed storage supports multi-path disk access, if one path access fails, it can automatically try other paths, meanwhile, it is able to support failure automatic switching at the node and disk levels, and its application is transparent to the upper level, thus ensuring high continuity of the application.
High Availability
(1) The said distributed storage allows a variety of applications or servers to access the storage simultaneously. It also supports the interoperability of file and object data, and enables the access protocols related to files, objects and big data in the same storage pool without duplicating data.
(2) The said distributed storage employs a fully redundant networking architecture to eliminate no single point of failure, ensuring that the service is not interrupted and the application is not aware when the hard disk or node is damaged.
(3) It also supports ACL access control and file audit log function to track the access to all nodes and file systems and events supported by all protocols executed by users.
(4) In addition, the said distributed storage supports different types of Quota management, wherein different Quotas can be set for different users and groups as required.
(5) ​Moreover, the said distributed storage employs AFM distributed routing and advanced caching techniques to achieve low latency in both read and write access of data stored anywhere.
Convenient Management
(1) Relying on rapid cluster deployment, the said distributed storage supports large-scale batch deployment, configuration boot settings and one-click upgrades, as well as aggregates various management tools and the intuitive GUI.
(2) The single point of administration can perform the administrative tasks of the system at any node in the cluster, and accordingly, the commands will be valid at all nodes.
(3) The management interface can manage the status of file systems, disks, and resource pools easily, and monitor the Dashboard to review the information related to health, alarm, capacity and performance.
(4) Relying on the monitoring and alarm system, the said distributed storage can quickly define problems and perform troubleshooting, thus enabling users to customize data distribution strategies, and supporting fault domain isolation.
(5) The said distributed storage is capable of automatically synchronizing configuration files and system information among each node, meaning that when the data is operated in one site, all other locations can synchronize immediately.