A system administrator is tasked with setting up a new 2TB disk for a database server. The database requires high read/write performance and redundancy. The server has a hardware RAID controller, but the administrator wants to use Linux software RAID for flexibility. Which of the following RAID levels should the administrator choose to maximize performance while providing fault tolerance, assuming the disk will be part of a larger array in the future?
Trap 1: RAID 5
RAID 5 writes parity across all member disks, so every write incurs read-modify-write overhead and rebuilds are slow, which conflicts with the database's high read/write requirement. RAID 5 suits read-heavy file or web servers where capacity efficiency matters more than write throughput.
Trap 2: RAID 0
RAID 0 stripes without parity, delivering the highest throughput but zero fault tolerance; a single disk failure destroys the entire array, violating the redundancy requirement. It is the right choice for scratch or temporary data where performance matters and loss is acceptable.
Trap 3: RAID 6
RAID 6 uses dual distributed parity, so each write performs multiple parity calculations and rebuilds are slower than RAID 5, undermining the high write performance the database demands. RAID 6 is correct for large nearline arrays where surviving two simultaneous disk failures outweighs write speed.
- A
RAID 5
Why it fails: RAID 5 writes parity across all member disks, so every write incurs read-modify-write overhead and rebuilds are slow, which conflicts with the database's high read/write requirement. RAID 5 suits read-heavy file or web servers where capacity efficiency matters more than write throughput.
- B
RAID 0
Why it fails: RAID 0 stripes without parity, delivering the highest throughput but zero fault tolerance; a single disk failure destroys the entire array, violating the redundancy requirement. It is the right choice for scratch or temporary data where performance matters and loss is acceptable.
- C
RAID 6
Why it fails: RAID 6 uses dual distributed parity, so each write performs multiple parity calculations and rebuilds are slower than RAID 5, undermining the high write performance the database demands. RAID 6 is correct for large nearline arrays where surviving two simultaneous disk failures outweighs write speed.
- D
RAID 10
RAID 10 stripes across mirrored pairs, combining RAID 0 read/write throughput with RAID 1 redundancy. It tolerates one disk failure per mirror and, unlike RAID 5 or 6, avoids parity write penalties, meeting the database's performance and fault-tolerance requirements.