Concepts

The distributed storage service is not supported if you select compute virtualization during system initialization.

Storage cluster

UIS integrates UIS-ONEStor distributed storage components to manage the UIS hosts as a storage cluster. You can configure the native disks on the hosts (storage nodes) into replicated or erasure coded storage pools to provide distributed storage services that are scalable, hardware independent, and highly reliable.

UIS-ONEStor distributed storage components offer the following benefits:

• Easy to manage—Traditional RAID requires expensive server hardware, storage switches, and special components such as HBAs. RAID group configuration is complex and typically requires advanced expertise and skills. In contrast, the back-end distributed storage system of UIS enables you to use the native disks of dispersed hosts to provide storage services. In addition, it is easy to use by automating most of the storage management tasks.

• Easy to scale—Native disks of the hosts are virtualized into distributed storage pools, which grow linearly in performance and capacity as the cluster scales out.

• Highly reliable—Data redundancy policy is configurable on demand to ensure data availability and durability. You can configure replication and erasure coding as needed to distribute replicas or blocks of data across multiple servers, which can be in different racks for enhanced data protection at the rack level. The cluster can automatically recover data when one or multiple disks, nodes, or racks fail depending on the configured redundancy policy.

• Intelligent load balancing—The storage system automatically balances data across the storage system when storage capacity is added or removed.

Disk pools

In a storage cluster, you can create disk pools by adding part of or all of disks on a node to a disk pool. The disk pools are logically isolated from each other. You can select one of the following deployment modes for each disk pool: All HDDs, All SDDs, and SSD Caches+HDDs.

Figure-1 Disk pools

 

Data pools

Data pools are created logically in each disk pool. You can configure the replication or erasure coding redundancy policy for each data pool.

Figure-2 Data disks

 

Replication and erasure coding

Data pools define the following data redundancy policies:

• Replication—Protects data by creating multiple replicas of data. The replicas are preferentially distributed on different disks. Data is available as long as one disk is operating correctly with the replica on the disk being integrated. Data durability increases with the number of data blocks at the expense of storage capacity. If the replication redundancy policy is configured, the storage space usage is 1/N, where N is the number of replicas.

• Erasure coding—Protects data by breaking data into blocks, which are expanded and encoded with redundant data pieces and stored across a set of different locations or storage media. The system can recover the data from any combination of a smaller number of those blocks. For example, if n data blocks are encoded with redundant data pieces to generate m parity blocks, data is available when a maximum of m blocks (including data blocks and parity blocks) fail. Encoding refers to the process of generating parity blocks, and decoding refers to the process of restoring data blocks. If the erasure coding redundancy policy is configured, the storage space usage is n/(n+m).

Replication is mainly used in production systems with higher performance and reliability requirements. Erasure coding provides higher storage usage efficiency and potentially higher fault tolerance than replication. However, erasure coding is more complex, requires more computation, and potentially increases the time to recover from failure. Erasure coding policies apply to read-only data and cold data such as VM backup files and VM templates.

Figure-3 Replication and erasure coding

 

Shared file system

The system supports OCFS2 shared file systems, which can be block devices provided by distributed storage and logical storage volumes provided by a SAN.

A shared file system can be mounted to multiple clients for conflict-free concurrent access.

Figure-4 Shared file system