In RAID 4, the main performance bottleneck arises due to

RAID levels are distinguished mainly by how they lay out data and where they place redundancy (parity) information. RAID 4 uses block-level striping across all the disks in the array, but keeps all parity blocks for every stripe on one single, dedicated disk, unlike schemes that distribute parity across every disk in the array.

1. Every stripe's parity block lives on the one dedicated parity disk, never on any of the data disks.

2. Any write to a data disk must recompute and write the new parity value for its stripe.

3. That recomputed parity always lands on the same single disk, so writes from different stripes queue on it one after another instead of proceeding in parallel.

Checking this against the other RAID levels and the remaining options confirms the cause:

• RAID 5 rotates the parity block for each stripe across a different disk in turn, so no single disk is hit by every write, which is exactly why RAID 5 avoids this particular bottleneck.

• Disk mirroring duplicates a disk's full contents onto a paired disk and is the redundancy method used in RAID 1, not RAID 4.

• Fragmentation is a file-system-level scattering of a file's blocks over time, unrelated to the array's striping or parity layout.

• RAID 4 uses a fixed, uniform stripe size across the array rather than variable block sizes.

So the performance bottleneck in RAID 4 comes from having a single dedicated parity disk that every write in the array must update.