RAID Levels Explained: RAID 1, 5, 6 and 10 for Indian Servers

Short answer: use RAID 1 for two-drive boot volumes, RAID 10 for databases and virtual machines, and RAID 6 for large hard-drive arrays such as file servers and backup targets. RAID 5 is the cheapest way to survive one failed drive, but it carries the most risk on big drives. RAID 0 has no protection. None of them is a backup.

This guide is for IT managers and business owners searching for RAID levels explained, RAID 10 vs RAID 5 or the best RAID for a server. It shows how each level works, how much usable space you get, why large drives raise rebuild risk, and which level fits which workload. Controller facts come from Dell's and HPE's own documentation and drive error rates from Seagate's datasheets, all read on 6 November 2026. Where a figure is our arithmetic, we say so. We do not quote prices here, because they change with stock.

What is RAID, and what does it not protect against?

RAID stands for redundant array of independent disks. It combines several drives into one volume so the server can keep running when a drive fails, and in some levels so it can read and write faster. The levels and their on-disk formats are standardised by the Storage Networking Industry Association in its Common RAID Disk Data Format.

RAID protects against a failed drive. It does not protect against accidental deletion, ransomware, file system corruption, a failed controller, a power surge, fire or theft. Those problems are copied to every drive in the array. Keep a separate backup, and see our backup and disaster recovery guide for a simple routine.

How do RAID 0, 1, 5, 6 and 10 work?

The table below uses the standard definitions and the minimum drive counts given in Dell's PERC documentation and the Wikipedia standard RAID reference.

• Striping splits data across drives so several drives work at once, which raises speed.
• Mirroring writes the same data to two drives, so either can serve it.
• Parity is extra calculated data that lets the array rebuild a missing drive. RAID 5 keeps one parity block per stripe and RAID 6 keeps two. Dell's guide describes RAID 6 as protecting against double drive failures and against a failure while a single drive is rebuilding.
• Nested levels combine the above. RAID 10 is a stripe of mirrors. RAID 50 and RAID 60 stripe across RAID 5 or RAID 6 groups, and Dell lists RAID 50 as needing at least six drives and RAID 60 as needing at least eight.

RAID 10 survives one failed drive in each mirror pair. If both drives of the same pair fail, the whole array is lost, so it does not survive every two-drive failure. RAID 6 survives any two.

How much usable capacity will you get?

Usable space comes from simple arithmetic: RAID 5 gives (n - 1) drives of space, RAID 6 gives (n - 2) and RAID 10 gives n divided by 2, where n is the number of drives. The table shows the result in terabytes as printed on the drive label.

Operating systems usually show a smaller number than the label, because drive makers count a terabyte as 10¹² bytes while many systems report in units of 2⁴⁰ bytes, roughly 9% smaller. Keep that in mind when you compare a quote with what Windows shows. Hot spares and any drives you keep outside the array come on top of these counts. Dell's guide notes that if you run only one array, RAID 6 is more effective than adding a hot spare, which is useful to remember when you plan the drive bays.

Why do big drives make rebuilds riskier?

When a drive fails, the controller rebuilds its contents onto a replacement by reading the surviving drives. In RAID 5 it must read every remaining drive in full. If it meets an unreadable sector on one of them while the array has no redundancy left, that data cannot be rebuilt from parity. Drive makers publish a worst-case rate for this, called the nonrecoverable read error rate.

These are the rates on the Seagate datasheets and product manual we read: the desktop BarraCuda SATA manual (Rev A, March 2025) lists 1 error per 10¹⁴ bits read, the Exos datasheets for 12 TB to 32 TB nearline drives list 1 sector per 10¹⁵ bits, and the older Exos 10E2400 10,000 RPM SAS datasheet lists 1 per 10¹⁶. Seagate's Exos X14 sheet labels the figure a maximum, so real drives may do better, but the specification is the only number the maker commits to.

The table below is our arithmetic. It multiplies the bits a rebuild must read by the datasheet error rate to give the expected number of unreadable sectors during one rebuild. It is a way to compare layouts, not a prediction for any specific drive.

Read it this way. A RAID 5 rebuild on four desktop-class 8 TB drives reads about 1.9 times the specified error interval, so hitting an error is likely on paper. The same rebuild on enterprise nearline drives is about ten times lower, and on 10K SAS drives it is far lower again. RAID 10 reads only the one surviving partner, which is why its rebuild reads less data than a parity rebuild. RAID 6 can correct an unreadable sector during a single-drive rebuild because it still has its second parity.

Rebuild time matters too. Even at an assumed sustained 200 MB/s, copying 8 TB takes about 11 hours of pure sequential work (8 TB divided by 200 MB/s), and a busy server takes longer. HPE's Smart Array guide lists a rebuild priority setting, so you can choose between a faster rebuild and keeping applications responsive. This is why the advice for large hard drives is RAID 6 or RAID 10, enterprise-class drives, and a tested backup.

RAID 5 or RAID 10: how do speed and cost compare?

RAID 10 writes each block to two drives, so a small random write costs two drive writes. RAID 5 must read the old data and old parity and write new data and new parity, which is four drive operations for one small random write, and RAID 6 needs six because it updates two parity blocks. This is standard textbook arithmetic and explains why parity RAID is slower on write-heavy work. Controllers soften it with a write cache and full-stripe writes, and HPE lists full-stripe writes among its Smart Array features. Reads are fast on all levels except a degraded array.

The trade is capacity. With eight drives, RAID 10 gives four drives of space and RAID 6 gives six, so RAID 6 holds 50% more data for the same drives. Choose RAID 10 when speed on mixed or write-heavy work matters more, and RAID 6 when capacity matters more. With SSDs, Wikipedia cites a study showing the RAID controller itself can become the bottleneck, so check the controller's rating when you build an all-flash array.

For a view on drive cost per terabyte see our Dell server hard drive and SSD price guide, the HP server storage price guide and SSD vs HDD for servers.

Which RAID level fits which workload?

This table is our own advice, built from the facts above. It is a starting point, not a rule for every application, so check what your software vendor recommends.

What do hardware RAID controllers, cache and batteries do?

A hardware RAID controller is a card or chip that builds the array and handles parity work. Dell's PERC 10 series (H740P and H840) lists RAID levels 0, 1, 5, 6, 10, 50 and 60, 8 GB of cache, a hardware XOR engine, hot-swap support and online capacity expansion. HPE's Smart Array SR guide lists the same seven levels plus RAID 1 Triple and RAID 10 Triple for its P-class controllers, and 0, 1, 5 and 10 for its E-class and S-class.

The write cache is where data and safety meet. Dell describes two write policies. In write-back, the controller confirms a write once the data is in its cache and writes it to the drive later. In write-through, it confirms only after the drive has the data. Dell's guide says write-back is used whenever the battery is present and in good condition, and it cautions that forcing write-back without a battery risks losing data if power fails suddenly. HPE's guide explains the newer design: flash-backed write cache uses flash devices and an energy pack so it needs power only for the time it takes to copy the cache from DRAM to flash, and the data is then posted to the drives on the next power-up.

The practical lesson for buyers is to check that a controller with write cache also has a healthy battery or energy pack. For prices and models see our RAID controller price guide and RAID controller price list.

Two maintenance features are also worth switching on. Dell's guide describes Patrol Read, a background scan that finds and fixes problems on drives before they fail, and Consistency Check, a background job that verifies and corrects mirror and parity data and that Dell recommends running periodically. Both use some performance while they run, so schedule them for quiet hours.

RAID controller, HBA or software RAID?

The right choice depends on your software. Windows Server, VMware and most business servers expect a hardware RAID controller. Storage software such as ZFS manages the drives itself and wants them presented individually. The OpenZFS documentation says it is best to use a host bus adapter instead of a RAID controller, and that using single-drive RAID 0 volumes to imitate an HBA is not recommended. ZFS has its own parity levels, raidz1, raidz2 and raidz3, which tolerate one, two and three drive failures, and its documentation says they avoid the RAID 5 and RAID 6 write hole, where data and parity become inconsistent after a power loss.

Controllers differ on pass-through. Dell's PERC 10 guide lists non-RAID pass-through mode as not available on the H740P and H840, while HPE lists mixed RAID and HBA mode on its E-class and P-class Smart Array controllers. So if you plan a ZFS, Proxmox or Ceph build, tell us before you buy and we will check which controller is in the unit. Compare the storage models in NAS vs SAN and our NAS, SAN and DAS comparison.

What mistakes should you avoid?

• Treating RAID as a backup. Keep a separate, tested copy.
• Using RAID 5 on large desktop-class drives. The rebuild arithmetic above shows why.
• Mixing very different drives. The array uses the smallest drive's size for every member, so a larger drive in a mixed set wastes space.
• Forgetting the controller's battery or energy pack. Write-back cache without protection risks data on a power cut. Our UPS sizing guide covers the power side.
• Ignoring alerts. A degraded array still works, so a silent failed drive can sit unnoticed until a second drive fails. Set up email alerts from the server's management tool.
• Planning capacity with no growth. Changing RAID level later is limited, so size for the next two to three years. Our server RAM and storage sizing guide helps.

Why buy RAID-ready servers from Serverwale?

Serverwale sells refurbished enterprise servers across India, tested before dispatch and backed by a comprehensive 1-year hardware warranty, with extended cover available at extra charge (see our policy page). Tell us your drive count and capacity target and we will suggest a RAID level, then tell you which controller and how many drive bays the unit has. Dell-based and HPE-based models are compared on Dell vs HP servers, and you can browse refurbished servers in India, add capacity with SAS hard drives, and cover the system after warranty with a server AMC. For model-level detail see the Dell R630, R730 and R730xd comparison and the HPE DL380 generation guide. Our guide on whether refurbished servers are reliable lists what to ask any seller, and our sister brand ProStation Systems explains the same ideas for new builds in its RAID levels guide. Current stock changes, so ask us what is available.

Frequently asked questions

Get a RAID plan for your server

Tell us your applications, how much usable capacity you need and how much downtime you can accept. We will suggest a drive count and RAID level, and tell you what controller is in the unit. Call or WhatsApp +91-87962-44410, message us on WhatsApp, or use the contact page. Serverwale, Burari, Delhi 110084.