Choosing the Right RAID Level for HP ProLiant Servers (RAID 1, 5, 6, 10)
RAID 0, 1, 5, 6, 10, 50, 60 on HPE Smart Array — what each costs you in usable capacity, how many drives can fail, and which level actually fits databases, virtualization, backups, and boot volumes. Plus which Smart Array controllers support which levels.
TL;DR — the fast answer by workload
- Databases / OLTP (write-heavy): RAID 10
- Virtualization (mixed): RAID 10, or RAID 6 if capacity matters more than peak write speed
- Backups / archives / file shares (capacity): RAID 6
- OS / boot pair: RAID 1
- Scratch / fully reproducible data only: RAID 0 (no redundancy — use sparingly)
The reason it's almost never "RAID 5 for everything" anymore is rebuild risk on large drives — more on that below.
---
What each level costs and protects
| RAID | Min drives | Drives that can fail | Usable capacity | Write profile |
|---|---|---|---|---|
| 0 | 2 | 0 | 100% | Fastest, zero protection |
| 1 | 2 | 1 (per mirror) | 50% | Great reads, safe writes |
| 5 | 3 | 1 | (n-1)/n | Write penalty (parity) |
| 6 | 4 | 2 | (n-2)/n | Heavier write penalty, safer |
| 10 | 4 | 1 per mirrored pair | 50% | Best random write |
| 50 / 60 | 6 / 8 | 1 per span / 2 per span | high | Striped parity at scale |
HPE Smart Array also offers 1ADM / 10ADM (three-way mirrors) on cached controllers like the P408i-a and P816i-a for maximum read performance and fault tolerance — at a 3x capacity cost. Supported levels per controller are listed in the HPE Smart Array SR Gen10 Controller User Guide.
---
Why RAID 5 fell out of favor for big drives
RAID 5 survives a single drive failure. The risk is the rebuild: with multi-terabyte nearline drives, a rebuild can take many hours to days, hammering the surviving drives — and if a second drive fails (or hits an unrecoverable read error) during that window, the whole array is lost.
That's why the modern guidance is:
- Small, fast drives, read-heavy: RAID 5 is still reasonable.
- Large nearline drives: use RAID 6 (survives two failures, covers you through a long rebuild).
- Write-heavy or latency-sensitive: use RAID 10 (no parity penalty, fast rebuild from a mirror).
---
Matching RAID to the workload
SQL Server / databases. Random writes dominate. RAID 10 gives the best write latency with no parity overhead. Pair it with a cached controller (P408i-a/P816i-a) so the flash-backed write cache absorbs commit bursts.
Virtualization (VMware/Hy-V). Mixed read/write across many VMs. RAID 10 if you can afford the 50% capacity hit; RAID 6 when you need more usable space and the workload isn't write-bound. For vSAN/Storage Spaces Direct, skip hardware RAID entirely — run the controller in HBA mode and let the software layer handle resiliency.
Backups, archives, media, file shares. Sequential and capacity-driven. RAID 6 on 7.2K nearline LFF drives is the sweet spot.
OS / boot. A simple RAID 1 mirror of two drives. Don't overthink it.
---
A few HP-specific gotchas
- RAID 6 needs a cached controller to perform well — the E208i-a (no cache) tops out at RAID 0/1/5/10. Step up to a P408i-a for RAID 6/60.
- Keep a hot spare for parity arrays (RAID 5/6) so rebuilds start automatically.
- Match drives within an array — same speed, interface, and ideally capacity. Mismatches drag the whole array down.
- Battery/cache health on refurbished controllers directly affects RAID 5/6 write speed — verify the FBWC pack.
---
FAQ
RAID 6 vs RAID 10 for a VM host? RAID 10 for peak write performance and fast rebuilds; RAID 6 when usable capacity matters more and the workload is read-leaning. Many shops run RAID 10 for the busy datastores and RAID 6 for bulk.
Is RAID a backup? No. RAID protects against drive failure, not deletion, corruption, or ransomware. Always keep real backups.
How many drives can I lose in RAID 10? One per mirrored pair — potentially several across the array, as long as you never lose both halves of the same mirror.
---
Pair the right RAID level with the right Smart Array controller and HPE SAS drives. Overview: HP Storage Solutions guide. Shop HP server storage.
Pro Disk Network is an independent reseller of genuine HPE hardware (not affiliated with HPE).