Every Dell PowerEdge or HPE ProLiant server with a RAID controller asks the same question when you build a volume: which RAID level? This guide explains what each common level does, how many drives it needs, how many drive failures it survives and how much usable space you get. The definitions and limits come from the Dell PERC 11 and PERC 12 user guides and the HPE Smart Array SR and MR Gen11 controller guides. The capacity figures are simple arithmetic you can repeat for your own drives.
Before you start
- RAID protects against drive failure. It does not protect against deleted files, ransomware or a site disaster, so keep separate backups (see the 3-2-1 backup strategy).
- Know your exact controller model. Supported RAID levels vary by model (table below).
- Plan arrays with drives of the same type. Dell and HPE do not allow hard drives and SSDs in the same virtual disk or array.
The building blocks: striping, mirroring and parity
Dell's PERC guide describes RAID with three techniques:
- Striping splits data into fixed-size pieces and writes them across several drives so they work in parallel. Striping alone gives no redundancy.
- Mirroring writes every block to two drives. If one fails, the other copy keeps the volume running. Dell notes that mirrored disks also speed up reads through read load balancing. Some HPE controllers also offer three-way mirrors, called RAID 1 Triple and RAID 10 Triple.
- Parity is redundant data calculated from the data blocks (HPE describes it as an XOR sum) and spread across all drives. RAID 5 keeps one set of parity and survives one drive failure. RAID 6 keeps two sets and survives two.
Nested levels combine these. RAID 10 stripes across mirrored pairs, RAID 50 stripes across RAID 5 groups and RAID 60 stripes across RAID 6 groups. Dell calls each group a span; HPE calls it a parity group.
RAID levels at a glance
| Level | How it works | Minimum drives | Drive failures survived | Usable capacity |
|---|---|---|---|---|
| RAID 0 | Striping only | 1 | None: one failure loses the volume | n × C |
| RAID 1 | Two-drive mirror | 2 | 1 | C |
| RAID 5 | Striping with single distributed parity | 3 | 1 | (n − 1) × C |
| RAID 6 | Striping with dual distributed parity | 4 | Any 2 | (n − 2) × C |
| RAID 10 | Stripe across mirrored pairs | 4 | 1 per mirrored pair | n ÷ 2 × C |
| RAID 50 | Stripe across RAID 5 groups | 6 | 1 per parity group | (n − g) × C |
| RAID 60 | Stripe across RAID 6 groups | 8 | 2 per parity group | (n − 2g) × C |
| RAID 1 / 10 Triple (HPE) | Three-way mirror | 3 | 2 per mirrored trio | n ÷ 3 × C |
Minimum counts come from Dell's "Minimum and maximum disks supported on each RAID level" table and HPE's "Selecting RAID for usable capacity" table. On PERC 11 and PERC 12, RAID 1 uses exactly two drives, RAID 0, 5 and 6 allow up to 32 drives, and RAID 10, 50 and 60 up to 240. A PERC RAID 10 with more than 32 drives must be spanned, with an even number of drives in each span. HPE adds RAID 10 drives in pairs and requires the drive count in RAID 50 or 60 to divide exactly by the number of parity groups.
Worked example: six 2.4 TB drives
Raw capacity is 6 × 2.4 TB = 14.4 TB. Usable capacity before file-system overhead:
| Level | Calculation | Usable | Survives |
|---|---|---|---|
| RAID 0 | 6 × 2.4 | 14.4 TB | No failures |
| RAID 5 | (6 − 1) × 2.4 | 12.0 TB | 1 failure |
| RAID 6 | (6 − 2) × 2.4 | 9.6 TB | Any 2 failures |
| RAID 10 | 6 ÷ 2 × 2.4 | 7.2 TB | 1 per pair (up to 3 if all in different pairs) |
| RAID 50, 2 groups of 3 | (6 − 2) × 2.4 | 9.6 TB | 1 per group |
| RAID 10 Triple (HPE), 2 trios | 6 ÷ 3 × 2.4 | 4.8 TB | 2 per trio |
| RAID 60 | Not possible | - | Needs at least 8 drives |
RAID 6 and RAID 50 both give 9.6 TB here, but they are not equal. RAID 6 survives any two failures. RAID 50 survives two only when they hit different parity groups: HPE states that all data is lost if a second drive fails in the same parity group before the first rebuild finishes.
Second example: eight 1.92 TB SSDs
Raw capacity is 15.36 TB. RAID 5 gives 7 × 1.92 = 13.44 TB, RAID 6 gives 6 × 1.92 = 11.52 TB, RAID 10 gives 4 × 1.92 = 7.68 TB, RAID 50 with two groups of four gives 6 × 1.92 = 11.52 TB, and RAID 60 with two groups of four gives 4 × 1.92 = 7.68 TB.
Why the operating system shows less
Drive labels use decimal units, where 1 TB = 1,000,000,000,000 bytes. Many tools report binary units, where 1 TiB = 1,099,511,627,776 bytes. So 12.0 TB appears as about 10.9 TiB. Also plan with the smallest drive if sizes differ: HPE requires replacement drives to be at least as large as the smallest drive in the array, and Dell's PERC 12 guide says a larger replacement is supported but its extra capacity may not be fully used.
Write cost: why RAID 5 and RAID 6 write more slowly
HPE's guide lists how many disk I/Os each host write causes, assuming performance is drive-limited and parity initialization has finished:
| RAID level | Disk I/Os per host write |
|---|---|
| RAID 0 | 1 |
| RAID 1 / 10 | 2 |
| RAID 1 / 10 Triple | 3 |
| RAID 5 | 4 |
| RAID 6 | 6 |
For example, 1,000 random host writes per second become about 2,000 disk I/Os on RAID 10, 4,000 on RAID 5 and 6,000 on RAID 6. HPE adds that reads perform similarly across levels (except small RAID 5 and 6 arrays), that full-stripe writes avoid most of the parity penalty, and that RAID 5, 50, 6 and 60 fault tolerance decreases as the array grows.
Not every controller supports every level
| Controller family | RAID levels |
|---|---|
| Dell PERC H755, H755N, H755 MX, H750 (PERC 11) | 0, 1, 5, 6, 10, 50, 60 |
| Dell PERC H355, H350 (PERC 11) | 0, 1, 10 |
| Dell PERC H965 series and H975i | 0, 1, 5, 6, 10, 50, 60 |
| Dell PERC H365 series | 0, 1, 10 |
| HPE Smart Array E-class SR Gen10 (for example E208i) | 0, 1, 5, 10 |
| HPE Smart Array P-class SR Gen10 (for example P408i, P816i) | 0, 1, 5, 6, 10, 50, 60, 1 Triple, 10 Triple |
| HPE SR Gen10 Plus and SR Gen11, 400 and 900 series | 0, 1, 5, 6, 10, 50, 60, 1 Triple, 10 Triple |
| HPE MR Gen11 200 series (MR216i) | 0, 1, 10 |
| HPE MR Gen11 400 series (MR408i, MR416i) | 0, 1, 5, 6, 10, 50, 60 |
Software RAID and boot devices support fewer levels. See hardware RAID vs software RAID and boot drives. To pick a level for a specific workload, read how to choose a RAID level.
Frequently asked questions
Is RAID 6 better than RAID 5 with a hot spare?
Dell's PERC guide states that RAID 6 protects against double disk failures and against a failure while a single disk is rebuilding, and that if you use only one array, deploying RAID 6 is more effective than deploying a hot spare disk. Hot spares are explained in global vs dedicated hot spares.
Can I change the RAID level later?
Sometimes. PERC 11 supports online RAID level migration for listed combinations, for example RAID 5 to RAID 6 by adding a drive, but RAID 50 and 60 cannot be reconfigured. Dell's unified guide states that RAID level migration is not supported on H965 and H365 series controllers. HPE Smart Array SR controllers can migrate RAID level with Smart Storage Administrator.
Warning: Changing a RAID level, or deleting and re-creating a volume to get a different level, puts data at risk. Back up first.
When is RAID 0 acceptable?
HPE describes RAID 0 as useful for fast storage of large amounts of non-critical data or when cost matters most. Dell warns that one physical disk failure fails a RAID 0 virtual disk and loses its data.
Checklist
- Confirm the controller model and the RAID levels it supports.
- Check the minimum drive count, pairs for RAID 10 and equal parity groups for RAID 50 and 60.
- Calculate usable capacity with the smallest drive size, then convert TB to TiB.
- Weigh write cost (2, 4 or 6 disk I/Os per write) against capacity.
- Prefer RAID 6 or 60 over RAID 5 or 50 when one more failure would be costly.
- Create the volume with the vendor tool, for example Dell PERC System Setup or HPE SSA, and keep backups.
Skyline can supply, configure and maintain Dell, HPE and Huawei servers under an annual maintenance contract (AMC). Contact us.
Comments
0 total · 0 threads