1.6.6.1 RAID-0 - Block Interleave Data Striping without Parity

Striping of data across multiple disk drives without parity protection is a disk data organization technique sometimes employed to maximize DASD subsystem performance (for example, Novell NetWares data scatter option).

An additional benefit of this data organization is drive spanning. With data striped across multiple drives in an array, the logical drive size is the sum of the individual drive capacities. The maximum file size may be limited by the operating system.

 

 

 

 

 

 

￿

 

 

 

 

 

 

 

 

 

 

￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿

 

 

 

 

 

 

 

 

 

￿

Disk

￿

 

 

 

 

 

 

 

 

 

￿

Controller

￿

 

 

 

 

 

 

 

 

 

￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿

 

 

 

 

 

 

 

 

 

 

￿

 

 

 

 

 

 

 

 

 

 

 

￿

 

 

 

 

 

 

￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿

 

 

 

￿

 

 

￿

 

￿

￿

 

￿

 

 

￿￿￿￿￿￿￿￿￿￿￿￿

￿￿￿￿￿￿￿￿￿￿￿￿

￿￿￿￿￿￿￿￿￿￿￿￿ ￿￿￿￿￿￿￿￿￿￿￿￿ ￿￿￿￿￿￿￿￿￿￿￿￿

 

￿

xxxxx

￿

￿

xxxxx

￿

￿ xxxxx

￿ ￿ xxxxx

￿ ￿

xxxxx

￿

Block 0

￿￿￿￿￿￿￿￿￿￿￿￿

￿￿￿￿￿￿￿￿￿￿￿￿

￿￿￿￿￿￿￿￿￿￿￿￿ ￿￿￿￿￿￿￿￿￿￿￿￿ ￿￿￿￿￿￿￿￿￿￿￿￿

 

￿

yyyyy

￿

￿

zzzzz

￿

￿

￿ ￿

￿ ￿

 

￿

 

￿￿￿￿￿￿￿￿￿￿￿￿

￿￿￿￿￿￿￿￿￿￿￿￿

￿￿￿￿￿￿￿￿￿￿￿￿ ￿￿￿￿￿￿￿￿￿￿￿￿ ￿￿￿￿￿￿￿￿￿￿￿￿

 

￿

 

￿

￿

 

￿

￿

￿ ￿

￿ ￿

 

￿

Block n

￿￿￿￿￿￿￿￿￿￿￿￿

￿￿￿￿￿￿￿￿￿￿￿￿

￿￿￿￿￿￿￿￿￿￿￿￿ ￿￿￿￿￿￿￿￿￿￿￿￿ ￿￿￿￿￿￿￿￿￿￿￿￿

 

 

Disk 1

 

Disk 2

 

Disk 3

Disk 4

 

Disk 5

 

 

 

 

 

xxxxx = Blocks belonging to a long file

 

 

 

 

yyyyy and zzzzz = Blocks belonging to short files

 

 

 

 

 

 

 

 

Figure

11.

RAID-0 (Block Interleave Data Striping without Parity)

 

 

Data striping improves the performance with large files since reads/writes are overlapped across all disks. However, reliability is decreased as the failure of one disk will result in a complete failure of the disk subsystem according to the formula:

Mean Time to Failure of a single disk

Mean Time to Failure = ￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿￿

Number of Disks in the array

1.6.6.2 RAID-1 - Disk Mirroring/Duplexing

This approach keeps two complete copies of all data. Whenever the system makes an update to a disk, it duplicates that update to a second disk, thus mirroring the original. Either disk can fail, and the data is still accessible. Additionally, because there are two disks, a read request can be satisfied from either device, thus leading to improved performance and throughput. Some implementations optimize this by keeping the two disks 180 degrees out of phase with each other, thus minimizing latency.

However, mirroring is an expensive way of providing protection against data loss, because it doubles the amount of disk storage needed (as only 50% of the installed disk capacity is available for data storage).

24NetWare Integration Guide

Page 39
Image 39
IBM SG24-4576-00 manual RAID-0 Block Interleave Data Striping without Parity, RAID-1 Disk Mirroring/Duplexing