Huawei Storage Study Notes-8

RAID 2.0+ composition concept

Disk domain

A hard disk domain is a combination of a bunch of hard disks (which can be all the hard disks in the entire system). These hard disks are integrated and reserve hot spare capacity to provide storage resources to the storage pool in a unified manner.

In traditional RAID, the first step to provide hard disk space to business hosts is to create a RAID group. However, there are some restrictions and requirements for creating a RAID group: the types of hard disks in the RAID group are the same, the size and speed of the hard disks need to be consistent, and the number of hard disks in a RAID group is not recommended to exceed 12.

Another implementation method is used in Huawei RAID 2.0+ technology. In this method, the first step is not to create a RAID group, but to create a hard disk domain. A hard disk domain is a group of hard disks. A hard disk can only belong to one hard disk domain.

In a hard disk domain, the number of hard disks can be much larger, such as hundreds, and can contain up to 3 types of hard disks. Each type of hard disk is assigned to a storage tier, and SSD disks are assigned to a high-performance tier. SAS disks are assigned to the performance layer, and NL-SAS disks are assigned to the capacity layer.

If there is no corresponding disk type in the disk domain, the disk domain does not have a corresponding storage layer. The more critical difference is that the main function of the hard disk domain is to separate this group of disks from which group of disks, in order to achieve complete isolation of faults, performance, storage resources, etc.

When creating a disk domain, the RAID level has not been specified, that is, the data redundancy protection method has not been specified. In fact, RAID 2.0+ has more flexible and refined settings for data redundancy protection. The storage space composed of hard disks in the hard disk domain will be divided into smaller-granularity storage pools, as well as hot spare space shared within each storage layer.

The size of the hot spare space is automatically set by the system based on the hot spare policy (high, low, or none) set by the administrator for the hard disk domain and the number of hard disks in each storage layer of the hard disk domain. It is no longer like that in a traditional RAID group. The administrator designates a specific disk as a hot spare disk.

Storage pools & storage tiers

Storage Pool is a storage pool, which is a container for storing storage space resources. The storage space used by all application servers comes from the storage pool.

A storage pool can be divided into multiple Tiers based on hard disk types. Tiers are storage tiers. A collection of storage media with similar performance in the storage pool is used to manage storage media with different performance to provide different storage spaces for applications with different performance requirements.

A storage pool is created based on a specified hard disk domain. Chunk (CK) resources can be dynamically allocated from the hard disk domain, and Chunk Groups (CKG) are formed according to the "RAID policy" of each storage layer (Tier) to provide applications with RAID protected storage resources.

When creating a storage pool, you can specify the storage tier (Tier) type that the storage pool is divided into from the hard disk domain, as well as the "RAID policy" and "capacity" of this type.

The OceanStor storage system supports RAID1, RAID10, RAID3, RAID5, RAID50, and RAID6. The capacity layer consists of large-capacity SATA and NL-SAS disks. The RAID strategy recommends using RAID6 with double verification.

hard drive group

The OceanStor storage system automatically divides each hard disk domain into one or more Disk Groups (DG) based on the number of hard disks of each type. A Disk Group (DG) contains only one hard disk type.

Multiple CKs in any CKG come from different hard drives in the same Disk Group (DG). Disk Group (DG) is an internal object of the system. Its main function is fault isolation. It is automatically configured by the OceanStor storage system and is not reflected externally.

logical disk

Logical Drive (LD) is a logical disk, which is a hard disk managed by the storage system and has a one-to-one correspondence with the physical hard disk.

CK

Chunk, abbreviated as CK, is the hard disk space in the storage pool divided into several fixed-sized physical spaces. It is the basic unit of RAID.

CKG

Chunk Group, referred to as CKG, is a logical storage unit composed of CKs from different hard disks in the same DG according to the RAID algorithm. It is the smallest unit for the storage pool to allocate resources from the hard disk domain.

Extent

Extent is a fixed-size logical storage space divided on the basis of CKG. The size is adjustable. It is the smallest unit for hot data statistics and migration (data migration granularity). It is also the smallest unit for applying for space and releasing space in the storage pool.

An Extent belongs to a Volume or a LUN. The Extent size can be set when creating a storage pool and cannot be changed after creation. The Extent sizes of different storage pools can be different, but the Extent sizes in the same storage pool are the same.

Grain

In Thin LUN mode, Extent is further divided into finer-grained blocks according to fixed sizes, and these blocks are called grains. Thin LUN allocates space at grain granularity, and LBAs within grains are continuous.

Thin LUNs are mapped to LUNs in grain units. For thick LUNs, there is no such object.

Volume & LUN

Volume is a volume, which is an internal management object of the storage system. LUN is a storage unit that can be directly mapped to the host for reading and writing, and is the external manifestation of the Volume object.

A Volume object is used to organize all Extent and Grain logical storage units of the same LUN. Extents can be dynamically applied to release to increase or decrease the space actually occupied by the Volume.