RAID Concepts
Module 4
After completing this module, you will be able to:
n Understand RAID concepts and level definitions.
n Understand the RAID levels offered in StorageWorks products.
n Understand the advantages of Adaptive RAID 3/5.
n Define JBOD and storage set.
n Understand the guidelines recommended for RAID configurations.
n Know the limits of subsystem configurations.
StorageWorks subsystems implement Redundant Array of Independent Disks (RAID) technology to protect data against individual disk drive failure and increase the speed of data access.
StorageWorks subsystems offer RAID levels 0, 1, 0+1, and 5. Information on other defined RAID Levels (2, 3, and 4) can be found in the Appendix.
Raid Level 0 uses disk striping. Data is divided into chunks and written across all the members of the array. This increases the speed of reads and writes because multiple physical disk drives participate in every I/O request. Striping is very fast when compared to the other RAID levels and the performance of an individual disk drive.
|
s |
Note RAID Level 0 is still frequently misunderstood. This “RAID” level provides no redundancy. If any member device fails, all the data on the set is lost. RAID Level “zero” provides “zero” redundancy |
Raid Level 1 uses disk mirroring. This RAID level provides redundancy simply by creating a copy on a second device of the information written on the first device. You can create and maintain multiple copies of the primary member.
RAID level 5 calculates and stripes parity information as well as data across all of its membership. If any one member fails, the missing data is regenerated from the surviving data and parity information.
StorageWorks products offer enhanced RAID 5 performance by virtue of the subsystem design and intelligence of the controllers.
Traditional implementation of RAID 5 suffers from what is termed a “write penalty” because each write requires a reading of old data and old parity before the new data can be written. With four I/O operations (two reads and two writes) required for each write, RAID 5 performance can suffer in write-intensive applications.
StorageWorks Products implementation of RAID 5 is faster than traditional RAID 5 because of these subsystem features:
·
Multiple
Internal SCSI Buses If a RAID set is spread across all six internal SCSCI
buses, the controller can write to all members of the array
simultaneously. This approaches the
speed advantage of synchronized RAID 3 operation.
· Intelligent Read and Write-Back Cache The controller coordinates read and write cache to increase the efficiency of traffic to the member devices. This approaches the RAID 3 requirement that all member devices remain equally active.
· [Deleted]
This is a combination of striping and
mirroring. StorageWorks subsystems
create mirror sets first. Those mirror
sets are then striped to create a 0+1 RAID configuration.
There are a few additional concepts relevant to subsystem configuration that are not strictly defined RAID concepts
This is not a RAID level. It is an informal way to describe the use of a single disk drive as a logical unit behind a RAID controller. JBODs are identified and initialized as logical units, but have no advantages beyond the normal performance associated with a single disk drive.
|
s |
Note There is no redundancy attached to JBODs. If a disk device configured as a JBOD fails, all the data on that device is lost. |
Storage sets are collections of physical devices that present themselves to the host system as large, single disk drives. Storage sets can be configured at any of the RAID Levels offered by the controller.
Storage sets can be viewed by entering the
“Show Storage” Command at the CLI prompt.