What Is RAID? RAID 0, 1, 5, 6 and 10 Explained for Websites

What is RAID?

RAID (Redundant Array of Independent Disks) is a technique that combines several physical disks into one logical storage volume and spreads data across them by striping, mirroring or parity. In short, depending on the level you choose, it adds speed, protection against a disk failure, or both.

We wrote this guide for website and online store owners, and for developers who run their own VPS. Our goal is simple: after reading it, you should see what RAID solves and what it does not, so you can make better hosting and security decisions.

We are a digital marketing and web team, not a hosting company. So the explanations rest on the Red Hat Enterprise Linux RAID documentation, the Linux kernel md documentation and the mdadm manual. However, where we were not sure about a command, a version number or a default value, we left it out.

The path below is easy to follow: levels, a comparison table, the backup warning, hardware versus software, the hosting angle, a failure scenario and commands. Also, feel free to jump to the part that fits your situation.

What are the three building blocks of RAID: striping, mirroring and parity?

That is the technical answer to what is RAID, and it hides in three simple ideas. Every RAID level mixes these ideas in a different way, so you do not need to memorize the levels.

• Striping: RAID cuts data into small chunks and writes them to the disks in turn. Several disks work at the same time, but losing one disk can take the whole array down.
• Mirroring: RAID writes the same data to two or more disks. If one disk fails, the other keeps running.
• Parity: RAID stores extra information calculated from the data. If a disk disappears, it rebuilds the missing piece from the remaining data and the parity.

For example, mirroring is simple and recovers quickly, but it spends half of your raw capacity. Parity uses capacity more efficiently. However, it needs extra calculation on every write, so the balance of speed, capacity and safety changes from level to level.

There is also the hot spare. This idle disk waits inside the array and takes over when a member fails, which starts the rebuild. The Linux kernel documentation calls this state "recover": a spare takes the place of a failed or missing device.

What is RAID 0 and when should you use it?

RAID 0 stripes data across at least two disks. Usable capacity equals the sum of all disks, and there is no redundancy at all. For instance, the Red Hat documentation describes it as a performance oriented striped layout with no redundancy.

People who search for what is RAID often assume RAID 0 is a speed upgrade. In reality, it is the wrong choice for anyone who wants protection against failure. If a single disk dies, you risk losing every file on the array. Also, more disks mean more chances for one of them to fail.

So when does RAID 0 make sense? It fits data that is temporary and easy to recreate. For example, scratch space for video processing, build output or a cache you can download again.

Instead, keeping a store database, order records or customer files on RAID 0 is a bad idea. Moreover, "two disks, double the speed" does not hold for every workload. Also, a slow site usually suffers from code and queries, not from the disk layout.

What is RAID 1 and why is it the most common choice?

RAID 1 writes identical data to two disks. Specifically, the Red Hat documentation describes it as copying the same data to each member disk. Usable capacity then equals that of the smallest disk in the array.

The array can keep working with only one disk present. So when a disk dies, your site does not go down. The array simply enters a degraded state, and you need to install a new disk quickly.

The advantages of RAID 1 are easy to list:

• The setup and the logic are simple.
• Reads can spread across both disks.
• Recovery after a disk failure is relatively easy.
• Many teams pick it for the operating system disk.

The drawback is capacity cost. For example, two 2 TB disks give you 2 TB of usable space. Besides, RAID 1 is not a backup: a file you delete by mistake disappears from both disks at the same moment.

What is RAID 5, how many disks does it need and how much space do you get?

RAID 5 combines striping with parity. According to the Red Hat documentation, it needs at least three disks, gives you the total capacity minus one disk, and survives the loss of one disk. The same page calls it the most common type of RAID.

Example calculation: with four 2 TB disks in RAID 5, you get roughly 6 TB of usable space. Then one disk worth of capacity goes to parity. Instead, a set of mirrored pairs on the same disks would give you far less.

The weak spot of RAID 5 is write performance. For a small write, the controller reads the old data and the old parity, calculates new parity and writes both. Therefore, write heavy workloads such as a busy database can feel slower than you expect.

Also, RAID 5 tolerates only one failed disk. If a second disk causes trouble during the rebuild, you lose the array. With large disks, this risk matters more, as the rebuild section below explains.

Is RAID 6 really safer than RAID 5?

Yes, in one specific way. RAID 6 stores two independent parity blocks. According to the Red Hat documentation, it needs at least four disks, gives you the total capacity minus two disks, and survives the loss of any two disks.

That extra safety matters most on arrays with large disks. A rebuild can take hours or even days, and a second failure during that window is possible. Because of this, RAID 6 keeps the array alive through that period.

The price comes in two parts. First, two disks worth of capacity go to parity. Second, every write needs even more calculation than in RAID 5. In short, RAID 6 favors safety over speed.

Example calculation: six 2 TB disks in RAID 6 give you roughly 8 TB of usable space. The same six disks in RAID 5 would give roughly 10 TB. That 2 TB gap is the insurance premium you pay against a second failure.

What is RAID 10 and how does it differ from RAID 5?

RAID 10 combines mirroring and striping. In the classic layout, four disks form two mirrored pairs, and data stripes across those pairs. So you get part of the safety of RAID 1 and part of the speed of RAID 0.

Capacity is roughly half of the raw total, so plan for that. Example calculation: four 2 TB disks in RAID 10 give roughly 4 TB. The Red Hat documentation notes that Linux md is more flexible here: capacity depends on the configuration, and you can even build it from three disks. Hardware controllers may follow different rules.

The difference from RAID 5 comes down to a few points:

• RAID 10 has no parity calculation, so writes are usually more consistent.
• Also, after a failure, the rebuild only copies the partner disk. It does not have to read the whole array.
• However, capacity efficiency is lower than in RAID 5.
• Survival depends on which disks fail: if both disks of one pair die, the data is gone.

For that reason, many teams choose RAID 10 for write heavy work such as databases. Still, it is not always the right answer, because budget and capacity needs decide.

How do the RAID levels compare, and which one fits what?

The table below uses information from the Red Hat documentation and the example calculations above. Its capacity column assumes four 2 TB disks (two disks for RAID 1). Also, the last column shows a general tendency, and the real result depends on your workload.

Looking at what is RAID level by level, the conclusion is clear: no single level is best. For a small business site, RAID 1 is often enough. For a busy database server, RAID 10 stands out.

What is RAID's most important warning: is RAID a backup?

No. RAID keeps a system running through a hardware failure, while a backup stores your data in another place at another time. The Red Hat documentation also states that RAID is not a backup solution.

Here is what RAID does not protect you from:

• Files you delete or overwrite by mistake, because the change reaches all members at once.
• Logical damage such as ransomware or a broken update.
• File system corruption and application bugs.
• Physical events such as fire, flooding or theft, because all disks sit in the same machine.
• Admin mistakes, for example removing the wrong disk from the array.

So having RAID does not mean you have a backup. Keep your backup in a separate place, behind separate access, and test the restore. For the full picture, read our website backup strategy guide.

What is the difference between hardware RAID and software RAID?

In hardware RAID, a dedicated controller card or a chip on the motherboard does the work. Meanwhile, in software RAID, the operating system manages the disks itself. On Linux, this is called md, and you manage it with the mdadm tool.

What is RAID in practice also depends on this choice. According to the Red Hat documentation, software RAID performs well on modern systems without extra hardware cost. The page also says it can outperform hardware RAID. So the old belief that hardware always wins no longer holds.

The portability and visibility rows describe a general tendency. The exact behavior depends on the controller model, so check the vendor documentation.

Why should you ask your host what is RAID on their servers?

On shared hosting, for example, you never see the disks. The provider builds the storage in its own way, and the plan page rarely says whether RAID exists or which level it uses. Still, a disk failure decides whether your site goes offline or loses data.

So ask the provider directly. For a business site, asking what is RAID on their platform once makes a years long risk visible. We covered other selection criteria in our guide on how to choose web hosting.

Ask the provider for these answers:

• Does the provider protect storage with RAID, and at which level?
• How do disk replacement and rebuild work after a failure?
• Is there a separate backup of customer data, and where does it live?
• How fast does the provider answer a restore request?

Vague answers tell you something too. Still, remember this: a host that uses RAID does not give you a backup.

What does RAID mean on a VPS or cloud server?

On a VPS, your operating system sees a virtual disk. You usually cannot see whether RAID sits behind it. That is the provider's infrastructure decision, and it differs from one provider to the next. So building software RAID inside a VPS rarely makes sense.

A dedicated server is different. If you have several physical disks, you can often choose the RAID level yourself. Many providers offer this option during installation, so read your provider's documentation for details.

For block storage on cloud platforms, read the documentation and the contract. Durability promises vary by platform, and the platform documentation tells you whether a promise counts as a backup. The general rule does not change: plan an independent backup for your data.

If you manage your own VPS, you can use our IP lookup and WHOIS lookup tools. They help you confirm which provider owns your server, so you contact the right support channel.

What is RAID's effect on website speed?

It can matter, but for most sites it is not the main factor. In fact, most slowness comes from image size, JavaScript weight, inefficient queries and missing caching. Next to those, the disk layout usually stays in the background.

Still, in some cases it makes a difference. For example, a write heavy database can run slower than you expect on RAID 5. If the array is degraded or rebuilding, read and write latency can also rise.

Disk speed and IOPS are a separate topic, and we cover it in a sibling article, so we do not repeat it here. For the link between speed and search, see how site speed affects SEO. For stores, see page speed and sales.

In short, measure first, then blame the disk layout. A Lighthouse performance test is a good place to start.

What happens to a RAID array when a disk fails?

On a redundant level (RAID 1, 5, 6 or 10), the array keeps working when a disk fails. However, it enters a degraded state. Your data stays reachable, but protection against the next failure has dropped, and performance may change.

The process usually runs like this:

1. The system notices that a disk reports errors and drops it from the array.
2. The array keeps running in degraded mode.
3. You or your provider replace the bad disk, or a waiting hot spare steps in.
4. The new disk joins the array and the rebuild starts.
5. When the rebuild ends, the array returns to full redundancy.

Speed is the critical point here. The shorter the degraded period, the smaller the chance of a second failure. That is why alerts and a spare disk matter as much as the level you pick.

RAID 0 has no such scenario. When one disk dies, the array dies with it. So never keep production data on RAID 0.

Why is a rebuild such a risky period?

During a rebuild, the array rewrites the missing data onto the new disk. On parity levels (RAID 5 and 6), this needs most of the remaining disks to be read from start to finish. Then the bigger the disks, the longer this takes.

However, two risks stand out in this window. First, another disk may fail. Second, if a remaining disk has an unreadable sector, the data in that sector may not be recoverable. Disk makers also publish unreadable sector rates in their datasheets. The numbers vary by model, so we do not quote any here.

Also, disks bought together and run under the same load often age at a similar pace. So one failure can hint that others will follow soon. In that case, checking the health data of the other disks is sensible.

During a rebuild, do the following:

• Confirm first that you have a recent backup and that you can restore it.
• Then reduce extra write load and postpone heavy maintenance jobs.
• Watch the progress; on Linux you see it in /proc/mdstat.

How do you check the status of a Linux software RAID?

You can look at the state of a Linux md array with two commands. Both also appear in the Red Hat documentation. We used md0 as the array name, so replace it with the name on your own system.

cat /proc/mdstat
mdadm --detail /dev/md0

The first command prints a summary of all md arrays and, if one is rebuilding, its progress. The second lists the level, the member disks and the state of the array you choose. If you see the word "degraded", the array is running with a disk missing.

The kernel documentation also describes monitoring files under sysfs. For example, /sys/block/md0/md/degraded counts how many devices the array is missing. The sync_completed file shows the sectors finished in the current operation, followed by the total.

cat /sys/block/md0/md/degraded
cat /sys/block/md0/md/sync_completed

These commands only read, so they do not harm the array. Even so, on a production server make sure you typed the array name correctly before you run anything.

How do you replace a failed disk with mdadm?

The Red Hat documentation shows a three step flow for replacing a failed disk: mark the disk as failed, remove it from the array, then add the new one. For example, in the code below, /dev/sdb1 is the failing partition and /dev/sdd1 is the new one. Always verify your own device names.

mdadm --manage --fail /dev/md0 /dev/sdb1
mdadm --manage --remove /dev/md0 /dev/sdb1
mdadm --manage --add /dev/md0 /dev/sdd1

You swap the physical disk after the first two commands. On the new disk, you also need a partition layout that matches the old one. The details depend on your distribution and disk layout, so read its documentation.

After the third command, the rebuild starts and you follow it in /proc/mdstat. Then again, removing the wrong disk from a degraded array can destroy the whole array. So check the serial number and the device name twice before you run the commands.

A command that creates a new array also exists. Run it only on empty disks, because it can make the existing content of the member disks unusable. Do not try it in production.

How do you monitor a RAID array regularly?

The most dangerous state for RAID is a silent failure that nobody notices. A disk may have dropped out months ago, while everyone still relaxes because "we have RAID". So setting up monitoring matters as much as choosing a level.

The mdadm manual describes a monitor mode that polls arrays for events. In this mode, the events go to a program or arrive by email. According to the manual, only the Fail, FailSpare, DegradedArray, SparesMissing and TestMessage events send an email.

mdadm --monitor --scan --mail=root@localhost

The kernel documentation also describes a way to verify the array on a schedule. Writing check to the sync_action file starts a full check of redundancy. The kernel counts the inconsistencies it finds in mismatch_cnt.

echo check > /sys/block/md0/md/sync_action
cat /sys/block/md0/md/mismatch_cnt

The repair value also fixes what the check finds. Do not run repair if you do not know what it does. Some distributions already schedule regular checks, so read your distribution's documentation first.

When should you not set up RAID yourself and leave it to your host?

Knowing what is RAID does not mean you have to build it. The honest answer is that most website owners do not need to set up RAID on their own. On shared hosting, managed VPS or managed cloud, this layer is the provider's job, and you cannot touch it anyway.

We suggest you leave the work to your provider or an experienced system administrator in these cases:

• You move a live server with order or customer data to RAID for the first time.
• You must wipe existing disks to create an array.
• You plan to change the settings of a hardware RAID controller.
• Or you do not know which disk is healthy in a damaged array.
• You have no backup that you can restore.

If you do manage your own dedicated server, try the commands above on a test machine first. The first attempt in production is where a mistake costs the most.

We work on the web and marketing side, and we do not operate server infrastructure. If your infrastructure choice affects speed and conversions, we can define the technical needs with you as part of our web design service.

Which RAID questions should you ask a hosting provider?

A short list before you buy costs less than an outage later. Ask in writing by email, so the answer stays on record.

• Does the storage use RAID, and at which level?
• Do SSD and NVMe disks get the same protection?
• What is the response time and service commitment after a disk failure?
• Does the provider keep customer backups in a separate location?
• When did the provider last test a restore?
• Can you start a restore yourself, or do you need a support ticket?

In practice, a clear answer shows that a provider takes the job seriously. A vague or evasive answer makes your own backup plan even more important.

Also, server location and network quality are separate decisions. For location, read about ping, latency and RTT. For traffic volume, read about hosting bandwidth.

What are the most common RAID mistakes?

First, we listed the mistakes that documentation and general practice mention most often. All of them come from the thought "I have RAID, so I am safe".

• Treating RAID as a backup: A file you delete disappears from both disks.
• Not monitoring: The array runs degraded for months and nobody notices.
• Putting production data on RAID 0: One disk failure takes everything.
• Starting a rebuild without a backup: A second problem can appear during the rebuild.
• Removing the wrong disk: Device names can change after a reboot, so verify by serial number.
• Never testing a restore: You cannot trust a backup you never tried.

So even a person who knows what is RAID can fall into these traps. However, none of them is hard to avoid; they only take discipline. A simple checklist is enough. For the wider security picture, see our OWASP Top 10 guide.

Which RAID level should you choose?

The last step in answering what is RAID is picking a level. That choice depends on the value of your data, your budget and your workload. Our suggestions below are a general frame, not a guarantee.

• Operating system and a small business site: RAID 1 is simple and usually enough.
• File and archive storage: RAID 5 or RAID 6 uses capacity efficiently; with large disks, consider RAID 6.
• Write heavy database: RAID 10 often gives more consistent results.
• Temporary, reproducible data: RAID 0 is possible, but keep important data away from it.

Whichever level you pick, remember three things. Keep the backup separate, keep monitoring on and test the restore. Otherwise, even the most expensive level will not protect you.

If technical decisions for your site feel hard, you can review infrastructure and performance topics with our team as part of SEO consulting.