Storage and File Systems
This domain covers how capacity becomes usable space: block versus file versus object access, RAID trade-offs, provisioning models, file system choice and mount behaviour. The exam favours questions where the arithmetic and the failure mode both matter.
Core rules
- Block storage (iSCSI, FC) presents raw LUNs; file storage (NFS, SMB) shares a file system; object storage addresses opaque objects over HTTP.
- RAID 5 survives one drive failure, RAID 6 survives two, RAID 10 survives one per mirrored pair but costs 50% capacity.
- Hot spares reduce usable capacity and rebuild time is dictated by drive size and load, not by RAID level alone.
- Thin provisioning decouples presented capacity from physical capacity, so overcommit requires monitoring of real consumption.
- Snapshots are point-in-time metadata references, not independent copies; they are not backups and they grow as source data changes.
- XFS and ext4 are journaling file systems: journaling protects metadata consistency, not user data integrity.
- Inodes are allocated at format time; running out of inodes produces "no space left on device" while free blocks remain.
Worked example
Six 4 TB drives. RAID 6 usable raw = (6 − 2) × 4 TB = 16 TB. RAID 10 usable raw = 3 × 4 TB = 12 TB. Reserving one drive as a hot spare in the RAID 6 set leaves five data drives, so usable drops to (5 − 2) × 4 TB = 12 TB. Reasoning: parity count and spare reservation both subtract from the drive count before multiplication.
Traps
Mixing TB and TiB; forgetting the hot spare in capacity maths; treating snapshots as backups; ignoring thin-provision overcommit; checking only free blocks when inode exhaustion is the cause.