200
EMC Symmetrix DMX-3 Product Guide
Data Integrity, Availability, and Protection
Regeneration In the event that a RAID 5 hypervolume fails, the RAID 5 device
operates with the surviving members. It is then running in reduced
mode. The data on the failing/failed member is reconstructed by
XOR’ing the information from the surviving members in the rank.
Fast Copy Back After the failed disk is replaced, data is rapidly copied back from the
spare drive to the newly installed drive.
Normal mode
RAID 5 normal mode has the following functions:
◆ “Writing data in RAID 5 normal mode” on page 200
◆ “Reading data in RAID 5 normal mode” on page 201
Writing data in RAID 5
normal mode
RAID 5 requires that the controller execute two read-modify-write
sequences executed concurrently Table 29 and Figure 44 on page 201.
The Symmetrix DMX implementation of RAID 5 distributes the work
of computing parity between the disk director and the disk drives
using the XOR chip located on the disk drive and the disk-level
buffer.
Symmetrix DMX RAID 5 optimizes performance for large sequential
write workloads as there is no need to read the parity from disks.
Since many sequential tracks are written, they are all in Symmetrix
global memory. The parity is calculated in global memory and
information is written to the disks in one stroke without requiring the
use of an expensive disk-level read-XOR-write operation.
Table 29 RAID 5 write operation sequence in normal mode
Drive containing data Drive containing parity
1. Read old information
2. Compute XOR bit mask
3. Write new information
1. Read old information
2. Receive XOR bit mask
3. Write new information
To Next Page
Loading...
Need help?
General
