HP NetRAID-4M Configuration and Upgrade Guide (Release 5)

85
Chapter 10 Glossary
RAID-10, Multiple 48 Drive Containers
The controller design supports a total of 64 containers of which 24 are accessible to
the user and 40 are hidden. The first 48 drive RAID-10 uses up 24 accessible and 24
hidden. Therefore, there are not enough leftover containers to make another RAID-
10 in the same HP Server.
RAID member fault
An example of a RAID member fault is a failure of the first drive of a not-fault-
tolerant container (volume set or stripe set). If one member of non-redundant
container fails the entire container will be wiped out.
RAID rebuild
In RAID-5, rebuild is a process that uses parity to recreate and write data to a new
partition. In RAID-1, rebuild is a simple process of copying data from the master
partition to the slave partitions and verification that the both halves of the mirror
contain identical data.
RAID scrubbing
Used during container creation process to create parity stripes. Scrubbing a mirror
set copies the data from the master half to the slave half. Scrubbing a RAID-5 set
involves reading data, calculating, and writing parity. During scrubbing, the host can
access the container, but performance is degraded to about 80%, and the container
is not yet fault tolerant. The scrubbing process takes about 14 hours per terabyte.
Scrubbing with the clear option involves writing zeros data into the container
without the need for reading and calculation. Scrubbing with clear is much faster but
access to the container is disabled during this time.
RAID types
Volume set (JBOD) - Just a bunch of disks (JBOD). Multiple drives are combined
together (concatenated) into a large logical drive and filled one by one. Very
convenient if all your applications are written for C: and C: is running out of space.
No fault tolerance.
Stripe set (RAID-0) - Multiple drives are combined together into a large logical
drive and then filled evenly (striped) for better performance. No fault tolerance.
Mirror set (RAID-1) - Two drives are combined together and the data is duplicated.
Write to both (relatively fast because no need to compute parity) and read from
either, therefore high performance. Fault tolerance but with 50% overhead.