Unified Architecture

　　The distributed storage developed by the Inspur employs a fully symmetric and decentralized distributed storage architecture. It can simultaneously provide clients with file, block, object and big data storage services, thereby realizing a set of storage functions to support a data center that can dramatically improve data mobility, storage utilization and manageability.

　　(1) File Storage Service: capable of running CIFS/NFS/FTP/HTTP/HTTPS and other standard communication protocols, and implementing file data sharing and access in a variety of operating systems, as well as supporting a number of advanced features including snapshot, clone, auto-reduction, remote replication, erasure correction, and volume QoS, etc.

　　(2) Block Storage Service: capable of providing block storage according to the standards of the iscsi-based protocol, supporting Windows, Linux, UNIX, Vmware ESXi and other operating systems, and available for KVM, vSphere, Xen, Hyper-V and other virtualized platforms, also having the functions of snapshot, clone, reduction configuration, etc., thereby creating flexible storage space for application scenarios such as virtualization and cloud computing, and guaranteeing data reliability.

　　(3) Object Storage Service: capable of being compatible with S3/Swift protocols and providing various SDKs to satisfy the data backup and data retrieval requirements of big data businesses in different scenarios, in addition, supporting advanced features including multi-tenant, multi-source zero-copy, WORM, remote disaster recovery, metadata retrieval, quota, encryption and so on.

　　(4) Big Data Storage Service: suitable for the separation scheme of big data storage and computing related to the native HDFS, so that realize the on-demand configuration of storage and computing resources, in addition, capable of solving the problems of misallocation of resources for computing in memory of Hadoop and data space wasting effectively by virtue of its efficient erasure correction technology and excellent parallel processing capabilities, meanwhile, it can be compatible with Hadoop HDFS and other standard interfaces based on big data analysis and application for realizing cross-platform data sharing and access, so that can meet the needs of different applications.

　　(5) Distributed Database Protocol: capable of supporting the access and storage of distributed database protocol interfaces, quickly locating the specified files or objects by employing database protocol interfaces, and implementing management according to different attributes of files or objects, thus effectively reducing cluster stress.

　　(6) Unified Resource Pool: capable of providing NFS, SMB, HTTP, FTP, POSIX, S3 and other interfaces simultaneously, and supporting IPV6/IPV4 dual-stack networking without configuring third-party shared software and secondary exports, in addition, it can achieve the scale-out of the data storage pool module as the data storage sub-modules are relatively independent in the storage pool.

　　(7) It supports to provide file, block, object, and big data services simultaneously in the same storage pool to simplify management.

　　Powerful Performance

　　Based on employing our own original iTurbo intelligence engine innovation technology, the distributed storage provides clients with innovative performance acceleration solutions, and comprehensively improves storage system performance by virtue of small file aggregation, object aggregation, global caching, multi-channels, kernel client, intelligent metadata management, intelligent resource scheduling, new data distribution algorithm and other leading technologies.

　　(1) This storage service scales the performance linearly with the number of nodes to provide users with performance at the petabyte-scale by scale-out.

　　(2) It provides a number of performance optimization techniques, including cache acceleration, small file aggregation, object aggregation, and automated tiering to improve the storage performance dramatically.

　　(3) In addition, the said storage technology supports global read, write and cache, so that can improve cache hit ratio, and further improve cache efficiency significantly.

　　(4) ​It also supports CIFS multi-channel to considerably improve service performance and reliability, in addition, when a single connection channel fails, it can employ other channels to continue transmitting data without affecting the service.

　　(5) Moreover, the file-based automatic tiered storage feature is configured to enable the intelligence storage and dynamic migration of hot and cold data among different storage levels.

　　(6) The said storage technology is capable of employing the specified kernel for designated purpose through intelligent resource scheduling to reduce the switching of single task between different kernels; in addition, the cutting-edge data distribution algorithm it adopted can achieve IO balance for each node and each disk, resulting in a space utilization rate of over 95%.

　　(7)  It enables end-to-end NVMe full-stack optimization of full-flash storage systems, converting random I/O to sequential I/O and completeing dump file, resulting in double performance gains.

　　Stable and Reliable

　　The distributed storage developed by the Inspur provides users with powerful data protection and system availability based on industry-leading redundancy design with data availability of up to 99.9999%.

　　(1) It supports the deployment of 2-8 replicas and erasure code redundancy protection mode ranging from K+1 to K+4 with the maximum erasure code size up to 32+2, furthermore, it supports the deployment of K+M:1 advanced erasure for small-scale clusters, so that allows online change of protection mechanisms.

　　(2) In addition, this storage technology provides hard disk reconstruction speed of 1TB/15min, so that reduces the cluster degradation time significantly, and guarantees the stable operation of the front-end business.

　　(3) It also provides AI-based disk failure prediction techniques to perform scheduled disk reconstruction to reduce the risk of data loss and performance degradation.

　　(4) Moreover, the said storage technology provides triple data protection at hard disk level, node level and cabinet level to reduce the risk of data loss.

　　(5) ​It also offers a hard disk anti-misinsertion and unplugging service, meaning that if the reinsertion is completed within five minutes, the function as a data hard disk can continue without data reconstruction.

　　(6) Furthermore, the comprehensive and systematic sub-health detection and pre-processing can be performed to identify degradation modes of the running status of networks, nodes, hard disks and other resources quickly and accurately, leading to the execution of alarms and automatic shutdowns to ensure durable, stable and reliable system performance.

　　(7) It also supports the synchronous remote replication and the asynchronous remote replication, and is capable of volume- and directory-based remote replication, as well as incremental data synchronization.

　　(8) The said storage technology supports users to create snapshots of files according to the employed storage service, also supports the creation of snapshots of any directory.

　　(9) It supports DNS domain name configuration, meaning that the event load balance can be managed via IP address. Moreover, it supports automatic load balance of client connections.

　　Massive Space

　　In addition to providing users with massive storage space thanks to the excellent system architecture, the distributed storage developed by the Inspur can improve capacity utilization dramatically through a series of functions to provide clients with lower TCO.

　　(1) The said distributed storage supports the expansion of 3-10240 nodes, and a single cluster can provide users with capacity space at the EB-level.

　　(2) Thanks to the balancing at the node level or the disk level achieved by this storage technique, the capacity error rate among disks does not exceed 5 percent, the error rate among nodes does not exceed 1 percent, and the available capacity utilization is as high as 95 percent.

　　(3) Based on its optimized architecture, a single directory can support 10 million files and a file system can support 100 billion files in a file scenario, in addition, the storage capacity of a single namespace can be extended to the EB level, wherein, the maximum single file is ≥128TB, and is completely transparent to the client.

　　(4) ​Relying on the optimized design of its architecture, a single bucket can support billions of objects in the object scenario.

　　(5) The said distributed storage supports the expansion of capacity and performance without service disruptions by adding storage nodes to storage systems under the condition of no downtime. Wherein, expanding a node takes less than 1 minute, and the performance loss is less than or equal to 10 percent. Furthermore, it supports to automatically balance the data load of all nodes.

　　(6) The said distributed storage supports the reduction of capacity without service disruptions by removing storage nodes from storage systems under the condition of no downtime. Wherein, reducing a node takes less than 1 minute, and the performance loss is less than or equal to 5 percent.

　　All Cloud

　　The distributed storage developed by the Inspur provides external users with standard interfaces to various virtualization, container and cloud platforms to help users achieve efficient data circulation in hybrid cloud environments.

　　(1) The said distributed storage supports CSI container interface to satisfy the storage requirements of container cloud.

　　(2) It also supports Manila, Cinder, Swift, S3 and other protocols, and is capable of connecting to various private/public clouds.

　　(3) In addition, it can provide a restful API to connect various cloud management platforms, leading to the management of storage systems and facilitating the operation and maintenance of cloud platforms.

　　Convenient Management

　　The management interface of our distributed storage employs humanized GUI design to provide a simple and easy-to-use operational interface for users. Meanwhile, it realizes the management of multiple sets of clusters with the same Web interface through the Inview, a multi-device management software, and provides advanced functions of smart operation and maintenance, thus creating intelligent and efficient operation and maintenance experience for users.

　　(1) Our distributed storage enables the docking with the Inview, a multi-device management software, and provides a GUI management interface to support visualization of operation and maintenance without third-party software or plug-ins, meaning that the management of multiple storage clusters can be performed with the same Web interface.

　　(2) It supports the docking with the Inview to provide all functions of AIOPS, including disk failure prediction, performance trend prediction capacity trend prediction, and other advanced capabilities, wherein, the overall prediction accuracy reaches 99%.

　　(3) The said distributed storage is also capable of sending alarm information through SMS, email and so on.

　　(4) In addition, it supports the QoS based on the number of connections and bandwidth, etc.

　　(5) Moreover, it enables the quota management based on directories, users and groups, which can be integrated with AD and LDAP.

　　(6) ​ The distributed storage can also generate annual reports, monthly reports, daily reports, and additional statistical reports, as well as support monitoring analysis, trend prediction, performance comparison, and diagnostic analysis, among others, provide API data interfaces.

　　Safe and Reliable

　　The released version of our distributed storage is strictly compliant with the enterprise-class high-standard security baseline. It has passed thousands of security tests covering the business areas of data, system, communication and application, providing users with a secure data storage platform.

　　(1) The said distributed storage supports the WORM function to provide users with file write-protection at different granularity.

　　(2) It also supports enhanced ACL and provides five different authorities including read, writing, execution, deletion and visualization.

　　(3) Moreover, the said distributed storage meets the security baseline requirements, and provides features such as user authentication, multiple data encryption protection, and Web application reinforcement to ensure secure storage applications.