Understanding Redundant Array of Inexpensive Disk (RAID)

Introduction

The term RAID is a common parlance in any Enterprise infrastucture.

So, what exactly is it ???

RAID stands for Redundant Array of Inexpensive (or Independent) Disk and makes two or more physical disks appear as one logical disk to the Operating System.

Great, but why do we need it ???

Enterprises use RAID for the following reason(s):

• Increased Capacity :: combines multiples inexpensive physical disks to form a larger capacity logical disk

• Increased Performance :: disk IO is spread across multiple physical disks thereby increasing read and write throughput

• Redundant and Fault Tolerant :: allows one to replace failed disk(s) and prevent data loss

RAID Terminology

The following are some of the commonly used terms in the RAID lingua franca:

• RAID Controller :: is the main component that manages how the data is written to and/or read from the array of physical disks

• Data Striping :: Striping distributes data blocks across the physical disks for better throughput through parallelism

• Data Mirroring :: Mirroring writes a given data block to two physical disks in parallel for data redundancy

• Parity :: Parity data block is created using the XOR operation on two data blocks from two physical disks. Using the Parity data block, RAID can recover the other data block if one of them is lost due to disk failure. The following illustrates how the Parity data block is computed:

  

  Figure.1

RAID Levels

The following paragraphs list and describe the most commonly used RAID levels:

RAID 0

This is the simplest RAID configuration.

• Requires at least two physical disks

• Uses data striping

• Does NOT use data mirroring

• Does NOT use parity

• No data redundancy

• Provides superior read/write performance

• Efficient logical disk space usage

The following visually illustrates RAID Level 0:

Figure.2

RAID 1

A given data block is stored on both the disk drives. When a disk drive fails, the other disk drive comes to the rescue.

• Requires at least two physical disks

• Does NOT use data striping

• Uses data mirroring

• Does NOT use parity

• Provides data redundancy

• Provides decent read/write performance

• Only 50% of logical disk space can be used

The following visually illustrates RAID Level 1:

Figure.3

RAID 5

A given data block is striped and a parity block is created & distributed amongst the disk drives. When a disk drive fails, it can be recreated using the remaining disk drives.

• Requires at least three physical disks

• Uses data striping

• Does NOT use data mirroring

• Uses distributed parity

• Provides excellent data redundancy

• Provides excellent read/write performance

The following visually illustrates RAID Level 5:

Figure.3

RAID 10

Thus far we have seen all the RAID Levels use an array of physical disks to form a logical RAID disk. Nothing is preventing a RAID Level from using the logical RAID disks.

RAID 10 is nothing more than a combination of RAID 1 and RAID 0. RAID 10 is also referred to as RAID 1+0.

• Requires at least four physical disks

• Uses data striping

• Uses data mirroring

• Does NOT use parity

• Provides superior data redundancy

• Provides superior read/write performance

The following visually illustrates RAID Level 10 (or 1+0):

Figure.4