RAID 5 Reliability
Which one is more reliable - RAID 5 or RAID1?
From the disk failure point of view, these arrays are the same.
Both survive a single member disk failure.
From the recovery point of view, RAID 1 is much simpler.
For example, in case of a controller failure, RAID5 requires
using a RAID recovery software first
to determine RAID parameters and may then require the use of a data recovery software to retrieve data;
while for a RAID1, each of the mirror disks is most likely usable on its own.
Which one is more reliable - RAID 5 or RAID10?
From the reliability point of view, RAID 5 and RAID10 are the same because both survive a single disk failure.
However, RAID 10 is a little better since its performance doesn't degrade that bad when a disk fails; another aspect is that RAID10 can survive a
multiple disk failure with non-zero probability. Strictly, probabilities are not taken into account when speaking about reliability.
Reliability is something that is guaranteed. If you want something more reliable than a RAID5, get a RAID6.
How does a RAID5 provide fault tolerance?
RAID5 is a fault tolerant array because it stores not only data itself,
but also the values of the XOR function calculated from the array data, so called "parity".
If some of the data is lost, it is still possible to restore data from the rest of the data and parity by simple calculations.
What is RAID rebuild?
For different RAID layouts the term "rebuild" means different things.
For RAID 5 and RAID 6 rebuild is a process when a parity is calculated and written to the disks.
For RAID 1 and RAID 10 rebuild is a data copying to get a redundancy. During the rebuild all the array data is read.
Usually the rebuild is performed in a background, therefore you can access the RAID during the rebuild, albeit slow.
Until the first rebuild is complete, a RAID doesn't provide fault tolerance. For large arrays, the rebuild can take a several days.
What is a RAID5 write hole?
RAID5 write hole is a condition of a RAID5 array when some of the parity blocks do not match the corresponding data.
Most often a write hole occurs when a power failure happens during the write operation.
In this case a controller has time to write data only but not parity.
To avoid accumulating write holes, one should use an UPS (Uninterruptible Power Supply) or BBU (Battery Backup Unit).
Additionally, periodical array rebuilds correct accumulated inconsistencies.
|