IBM Power Systems 775 Manual page 93

GPFS Native RAID automatically creates redundancy information, depending on the
configured RAID code. By using a Reed-Solomon code, GPFS Native RAID equally divides a
GPFS block of user data into eight data strips and generates two or three redundant parity
strips. This configuration results in a stripe or track width of 10 or 11 strips and storage
efficiency of 80% or 73% (excluding user configurable spare space for rebuild).
By using N-way replication, a GPFS data block is replicated N - 1 times, implementing 1 + 2
and 1 + 3 redundancy codes, with the strip size equal to the GPFS block size. Thus, for every
block or strip written to the disks, N replicas of that block or strip are also written. This
configuration results in track width of three or four strips and storage efficiency of 33% or
25%.
End-to-end checksum
Most implementations of RAID codes implicitly assume that disks reliably detect and report
faults, hard-read errors, and other integrity problems. However, studies show that disks do not
report some read faults and occasionally fail to write data, although it was reported that the
data was written.
silent errors phantom-writes dropped-writes
These errors are often referred to as , , , or
off-track writes
. To compensate for these shortcomings, GPFS Native RAID implements an
end-to-end checksum that detects silent data corruption that is caused by disks or other
system components that transport or manipulate the data.
When an NSD client is writing data, a checksum of 8 bytes is calculated and appended to the
data before it is transported over the network to the GPFS Native RAID server. On reception,
GPFS Native RAID calculates and verifies the checksum. GPFS Native RAID stores the data,
a checksum, and version number to disk and logs the version number in its metadata for
future verification during read.
When GPFS Native RAID reads disks to satisfy a client read operation, it compares the disk
checksum against the disk data and the disk checksum version number against what is stored
in its metadata. If the checksums and version numbers match, GPFS Native RAID sends the
data along with a checksum to the NSD client. If the checksum or version numbers are
invalid, GPFS Native RAID reconstructs the data by using parity or replication and returns the
reconstructed data and a newly generated checksum to the client. Thus, both silent disk read
errors and lost or missing disk writes are detected and corrected.
Declustered RAID
Compared to conventional RAID, GPFS Native RAID implements a sophisticated data and
spare space disk layout scheme that allows for arbitrarily sized disk arrays and reduces the
overhead to clients that are recovering from disk failures. To accomplish this configuration,
GPFS Native RAID uniformly spreads or declusters user data, redundancy information, and
spare space across all the disks of a declustered array. A conventional RAID layout is
compared to an equivalent declustered array in Figure 1-58 on page 80.
79
Chapter 1. Understanding the IBM Power Systems 775 Cluster