A server that hosts a write-intensive SQL database uses eight 10 K RPM SAS disks on a hardware RAID controller with 1 GB of battery-backed cache. According to the build book, the volume should deliver the highest possible write throughput while still surviving the loss of one disk in each mirrored pair. After a recent controller replacement and array rebuild, administrators notice that large batch inserts now take three times longer and the controller dashboard shows constant parity-calculation activity. The logical drive is currently defined as RAID 50 with a 256 KB stripe size. Which change would BEST correct the improper RAID configuration and restore the expected write performance without sacrificing fault tolerance?
Enable read-ahead caching on the controller and retain the current RAID 50 layout.
Convert the array to RAID 6 and add a global hot spare.
Keep RAID 50 but reduce the stripe size from 256 KB to 32 KB.
Recreate the logical drive as RAID 10 across all eight disks.
RAID 10 (striped mirrors) has no parity overhead, so write I/O is limited only by the spindle count and controller cache; it still tolerates one disk failure per mirrored pair. RAID 50, RAID 6, and other parity-based levels must read old parity, recalculate it, and write new parity on every write, introducing significant latency-exactly the behaviour now seen on the controller. Simply changing stripe size or enabling read-ahead does not remove the parity penalty, and RAID 6 adds even more write overhead. Re-creating the volume as RAID 10 therefore resolves the performance issue while maintaining redundancy.
Ask Bash
Bash is our AI bot, trained to help you pass your exam. AI Generated Content may display inaccurate information, always double-check anything important.
What is the main difference between RAID 10 and RAID 50 configurations?
Open an interactive chat with Bash
Why does RAID 10 perform better than RAID 50 for write-intensive workloads?
Open an interactive chat with Bash
What role does the stripe size play in RAID configurations, and why wouldn’t changing it improve performance in this scenario?