RAID 5, 6 or 10: what to choose for a storage system and why it depends on the drives
Servix technical team · Updated
In brief. For databases, ERP and 1C, RAID 10 is used; for archives and backups, RAID 6; and RAID 5 is not used on drives of 8 TB and up. The write penalty for RAID 10, 5 and 6 is 2, 4 and 6 operations; out of eight 8 TB drives, 32, 56 and 48 TB are usable. Rebuilding a 16–20 TB drive takes more than a day, and all that time RAID 5 is defenseless against a second failure; on SSDs, RAID 5 is acceptable.
Which RAID to choose for a storage system: the short answer
For databases, 1C and virtual machines — RAID 10. For file shares, archives and backups on HDDs of 8 TB and up — RAID 6. RAID 5 — only on SSDs or drives up to 4 TB. On Dell PowerVault ME5 and HPE MSA storage systems with 12 or more drives, distributed RAID is used — ADAPT and MSA-DP+: spare capacity is built in, and rebuilds run many times faster.
Task | Tier | Min. drives | Usable from 8 × 8 TB | Operations per write | Survives failures |
|---|---|---|---|---|---|
1C, DBMS, virtual machines | RAID 10 | 4 | 32 TB | 2 | 1 in each pair |
File server, archive, backup on HDD | RAID 6 | 4 | 48 TB | 6 | 2 |
SSD, mostly read workload | RAID 5 | 3 | 56 TB | 4 | 1 |
Dell ME5 / HPE MSA, 12 drives or more | ADAPT / MSA-DP+ | 12 | — | same as RAID 6 | 2 |
The RAID level is usually chosen as an afterthought: people take what the integrator suggested or what was on the previous array. Yet this choice determines three things — how much usable space will remain from the purchased drives, how fast writes will be, and whether the array will survive the failure of a second drive during a rebuild. Let's go through it in order, without textbook formulas but with numbers.
Three levels you encounter in practice
RAID 5 — data and parity are distributed across all drives; the capacity of one drive goes to parity. Out of eight 8 TB drives, 56 TB will remain usable. The array survives the failure of one drive.
RAID 6 — the same, but with double parity: the capacity of two drives is lost, but the array survives the simultaneous failure of two drives. Out of the same eight drives, 48 TB will remain.
RAID 10 — mirrored pairs combined into a stripe. Half of the capacity goes to mirrors: out of eight drives, 32 TB are usable. The array survives the failure of one drive in each pair — but not of two drives in the same pair.
The cost of writes: why RAID 10 is faster
The key parameter people forget is the write penalty. To write one block, the controller performs a different number of drive operations depending on the level:
Tier | Operations per write | Usable capacity from 8 drives of 8 TB | Survives failure of |
|---|---|---|---|
RAID 10 | 2 | 32 TB (50%) | One drive in each pair |
RAID 5 | 4 | 56 TB (87.5%) | One drive |
RAID 6 | 6 | 48 TB (75%) | Two drives |
The threefold difference in writes between RAID 10 and RAID 6 is not theory but what you see in monitoring under a database or active virtualization load. For reads, the difference between levels is small, so for archives and file storage, where writes are rare and reads are frequent, the write penalty hardly matters. For DBMS, VDI and any write-intensive workloads, it is decisive.
The main danger of RAID 5 is the rebuild
When a drive in RAID 5 fails, the array starts rebuilding onto a spare drive. To restore the data, the controller must read every block on all remaining drives. And here two risks appear that grow with drive capacity.
Time. Rebuilding a 16–20 TB drive on a loaded array takes a day or more. All this time the array runs without redundancy: a failure of a second drive means complete loss of the group's data.
Read error. Enterprise drives have a rated parameter — the probability of an unrecoverable read error, usually one sector per 10^15 bits read for NL-SAS and per 10^14 for SATA. When rebuilding an array of high-capacity drives, the read volume approaches these values, and the chance of hitting a bad sector exactly when there is no redundancy stops being theoretical.
The practical conclusion is simple: RAID 5 is not used for mechanical drives of 8 TB and up. The standard has become RAID 6 or its vendor equivalents with double parity, and RAID 10 for critical workloads.
Flash changes the picture
With solid-state drives the arithmetic is different. An SSD rebuild runs many times faster — there are no mechanics, and access time does not depend on data location. Accordingly, the vulnerability window is shorter, and RAID 5 on SSDs no longer looks reckless, especially on drives of moderate capacity.
But another concern appears — write endurance. The write penalty of RAID 5 and RAID 6 means that the drives take on more physical operations than the application sends. For drives with low write endurance (read-intensive), this shortens their service life. Therefore, for write-intensive workloads, either drives with higher endurance or RAID 10 with its minimal penalty are used.
A separate story is modern arrays that do not offer classic levels at all. In flash systems such as Dell PowerStore protection is implemented by a distributed scheme at the pool level: spare space is spread across all drives, and the rebuild runs in parallel onto all drives rather than onto one spare. This sharply reduces recovery time. There is no need — and no way — to configure the level manually; the platform does it itself.
What to choose for a specific task
File archive, video surveillance, backups on high-capacity drives. RAID 6. Few writes, capacity matters, and the failure of two drives during a rebuild is a real scenario. On arrays such as the PowerVault ME5012 or HPE MSA 1060 with LFF drives, this is the default choice.
Databases, ERP, 1C with active writes. RAID 10. Half the capacity is the price for predictable latency and a minimal write penalty. An attempt to save money with RAID 5 usually ends with buying more drives six months later and rebuilding the array anyway.
Virtualization cluster with a mixed workload. More often RAID 6 on the capacity tier plus RAID 10 or a flash pool for demanding machines. Splitting by storage tiers works better than trying to find one compromise level for everything.
Everything that goes on all-flash NVMe. The platform chooses the scheme itself; the administrator's task is to correctly estimate the required capacity taking data reduction into account and to reserve spare space.
What affects reliability more than the RAID level
A spare drive in the configuration. A hot spare should be in the array from day one. A rebuild that starts immediately after a failure and a rebuild that waits a week for a drive to be delivered are fundamentally different scenarios in terms of risk.
Monitoring and notifications. An array in which a drive failed a month ago and nobody noticed is a typical cause of data loss. Configured email notifications cost nothing and cover the main risk.
Backups. RAID of any level protects against hardware failure, but not against file deletion, ransomware or administrator error. Backups should be stored on another system, and preferably also on media that cannot be written to from the production network.
To select an array configuration with a calculation of usable capacity for the chosen level, write to us. See the available storage systems in the Storage systemssection; how to choose the platform itself is covered in the storage system selection guide.
Related roundups and tools: RAID calculator, File server and backup server, RAID controller guide, Drives and RAID controllers.
Sources
- Dell PowerVault ME5 Storage System Best Practices (H19551) — Dell Technologies
- Dell PowerVault ME5: ADAPT (distributed RAID) — white paper — Dell Technologies
- HPE MSA 1060/2060/2062 Storage Arrays Best Practices (a00105260enw) — HPE
Questions and answers
Can the RAID level be changed without losing data?
Some arrays support level migration, but it takes a long time and heavily loads the system; others require recreating the group and restoring from a backup. The level should be planned before buying drives — redoing it costs more than choosing the right drives from the start.
What is the minimum number of drives?
RAID 5 — from three drives, RAID 6 — from four, RAID 10 — from four and always an even number. In practice, it makes sense to calculate not the minimum but a reasonable group size: overly wide groups on high-capacity drives increase rebuild time and risk.
What is spare space instead of a spare drive?
Instead of a dedicated drive, modern systems reserve part of the capacity on all drives. On failure, recovery runs in parallel onto all remaining drives rather than sequentially onto one, and takes many times less time. This is exactly what makes large flash pools practical.
Related reading
RAID and the server disk subsystem: levels, fault tolerance, controllers
RAID levels for a server: how RAID 4 differs from RAID 5, how many drive failures RAID 6, 10 and 50 survive, why RAID 5 is dangerous on drives of 4 TB and larger. With prices.
Server drives: HDD or SSD, SAS, SATA or NVMe — what to buy for the DL380 and R750
Which drives to buy for an HPE DL380 or Dell R750 server: SAS, SATA, NVMe, PERC and Smart Array RAID controllers. Prices in UZS, three ready-made sets. Tashkent.
Dell PowerVault ME5 or HPE MSA 2060/2062: which storage system to choose for an office and virtualization
Dell ME5024 and ME5012 vs HPE MSA 2060/2062: 32G and 16G FC, iSCSI, SAS, ADAPT and MSA-DP+, licenses, choosing a storage system for virtualization with 2–4 hosts, and prices.
How to choose a storage system: SAN, NAS, all-flash and hybrid
Choosing a storage system step by step: block or file access, SAS vs. iSCSI and Fibre Channel, all-flash or hybrid, fault tolerance and growth. HPE MSA 2060 with FC — from 88.3 million UZS.



