3.2.3. File System Types
💡 First Principle: A file system defines the rules for how data is organized, stored, named, and accessed on a volume. Different file systems make different trade-offs among performance, reliability, metadata capability, and maximum file/volume size. Choosing the wrong file system for a workload is a permanent mistake until reformatting (with data loss or migration).
File System Reference
| File System | OS | Max File Size | Key Features | Use Case |
|---|---|---|---|---|
| ext4 | Linux | 16 TB | Journaling, fast fsck, backward compatible with ext3 | General-purpose Linux |
| NTFS | Windows | 16 TB (4 KB clusters) / 256 TB (64 KB clusters) | ACLs, compression, encryption (EFS), journaling, quotas | Windows servers |
| VMFS | VMware ESXi | 62 TB (VMFS 6) | Cluster-aware, supports simultaneous multi-host access | VM datastores |
| ReFS | Windows Server | 35 PB | Integrity streams (checksums), auto-repair, designed for large volumes | Storage Spaces, Backup targets |
| ZFS | Linux, FreeBSD, Solaris | 16 EB | Copy-on-write, built-in RAID, checksums, snapshots, deduplication | High-value data, NAS |
⚠️ Exam Trap: Do not confuse a file system's maximum file size (above) with its maximum volume size — they are different numbers and questions exploit the gap. VMFS 6 caps a single file (VMDK) at 62 TB but a datastore at 64 TB. ZFS caps a file at 16 EB while a zpool can theoretically reach 256 ZB. ReFS is unusual in that both limits are the same (35 PB).
Key Characteristics to Know
- Journaling: Most modern file systems (ext4, NTFS, ReFS) maintain a journal—a log of pending changes. If the system crashes mid-write, the journal allows recovery without full fsck. Prevents corruption from unexpected shutdowns.
- ACLs (Access Control Lists): NTFS supports granular per-file/folder permissions. ext4 also supports ACLs but requires mounting with
acloption. - VMFS cluster-awareness: VMFS is specifically designed to be accessed by multiple ESXi hosts simultaneously—on shared block storage, this is what enables vMotion and vSphere HA (NFS datastores achieve sharing at the file level instead). Standard file systems like ext4 or NTFS would corrupt if accessed simultaneously by multiple hosts.
⚠️ Exam Trap: ZFS is not a file system you "choose for a workload" in the same way—it replaces the entire storage stack (RAID controller, volume manager, and file system) with its own implementation. Its copy-on-write architecture means it never overwrites existing data in place, which provides strong data integrity guarantees.
Reflection Question: A Windows Server is being configured as a backup target that will store large amounts of data and needs to detect and auto-repair data corruption. Which file system is most appropriate?
