RAID, or a redundant array of independent disks, combines multiple physical drives into one logical unit. It improves performance, adds redundancy against drive failure, or both, depending on which RAID level you use. RAIDs protect against drive failures. They primarily use these technologies across different levels.
What is RAID?
RAID combines multiple physical drives either HDDs or SSDs, into one logical unit that the operating system treats as a single disk. Depending on the level you choose, RAID can improve read and write speed, protect against drive failure, or both.
RAIDs are not backups. RAID protects against one specific problem: a drive failing. It does not protect against accidental deletion, file corruption, ransomware, or a disaster that damages the whole system, because RAID faithfully copies or stripes data across every disk in the array. Most serious storage setups use RAID for uptime and a separate backup for actual recovery.
Striping, mirroring, and parity
Three mechanisms show up across almost every RAID level, so it's worth understanding each before looking at the levels themselves.
Striping splits data into blocks and spreads those blocks across multiple disks. Because several disks can be read from or written to at the same time, striping improves performance, especially for large files. Striping alone adds no redundancy: if one disk in a striped set fails, the data on that disk, and often the whole array, is gone.
Mirroring writes identical copies of data to two or more disks at once. If one disk fails, an untouched copy still exists on another disk. Mirroring is reliable, but expensive in terms of capacity, since you need at least double the storage to protect a given amount of data.
Parity is calculated redundancy information, stored alongside the actual data, that lets the array reconstruct a failed disk's contents from the surviving disks. Parity protects data using less capacity than mirroring does, but calculating it adds overhead to every write.
RAID levels
RAID levels combine striping, mirroring, and parity in different ratios. Here's how the common levels work.
RAID 0: Striping
RAID 0 stripes data evenly across every disk in the array, with no redundancy at all. Because multiple disks share the read and write load, performance improves, particularly for large files. RAID 0 needs a minimum of two disks and uses all of their combined capacity. It's also the riskiest level: if a single drive fails, the entire array's data is lost, since every file is split across all the disks.
RAID 1: Mirroring
RAID 1 mirrors data across two or more disks. Read performance improves, since read requests can be split across the mirrored disks, but write performance suffers slightly, since every write has to be repeated on each disk. RAID 1 needs a minimum of two disks and uses half their combined capacity, regardless of how many disks are mirrored. If one disk fails, the data is still intact on the others.
RAID 5: Striping with parity
RAID 5 stripes data across disks and adds a single set of parity information, distributed evenly across the array rather than stored on one dedicated disk. It needs a minimum of three disks and uses the capacity of all but one of them. Reads are fast, since multiple disks work in parallel, but writes carry a performance penalty, since parity has to be recalculated each time. RAID 5 survives a single drive failure.
One caveat: rebuilding a failed RAID 5 array means reading every remaining disk in full. On large, older drives, that process can itself trigger a second failure or an unreadable sector. This is one reason many admins now prefer RAID 6 for large arrays.
RAID 6: Striping with double parity
RAID 6 works like RAID 5, but calculates two independent sets of parity instead of one. It needs a minimum of four disks and uses the capacity of all but two of them. The trade-off for that extra protection is a bigger write penalty than RAID 5, since two parity calculations happen on every write. In exchange, RAID 6 survives two simultaneous drive failures. That matters most on large arrays, where a rebuild after one failure can take long enough for a second failure to become a real risk.
RAID 10 (1+0): Mirroring and striping
RAID 10 first mirrors pairs of disks, then stripes data across those mirrored pairs. It needs a minimum of four disks, in even numbers, and uses half their combined capacity. Because data is both mirrored and striped, RAID 10 combines strong performance with solid fault tolerance. It can typically survive multiple drive failures, as long as both disks in the same mirrored pair don't fail together. The trade-off is cost: you need twice the raw capacity to get your usable capacity, the same as RAID 1.
RAID 01 (0+1): Striping and mirroring
RAID 01 builds two striped sets first, then mirrors those two sets against each other. It also needs a minimum of four disks, in even numbers, and also uses half their combined capacity. The difference between the two is fault tolerance, not capacity. Because RAID 01 has only two groups, losing one disk from each group can take down the entire array. The same number of failures in RAID 10 would often be survivable instead. For that reason, RAID 10 is generally preferred over RAID 01 when both are available.
RAID levels: Comparison
| Level | How it works | Min. disks | Usable capacity | Survives | Best for |
|---|---|---|---|---|---|
| RAID 0 | Stripes data across every disk | 2 | 100% | No drive failures | Speed-first, non-critical data |
| RAID 1 | Mirrors data across disks | 2 | 50% | 1 drive | Small, critical volumes |
| RAID 5 | Stripes data with single parity | 3 | All but 1 disk's worth | 1 drive | General file and app servers |
| RAID 6 | Stripes data with double parity | 4 | All but 2 disks' worth | 2 drives | Large arrays, higher safety margin |
| RAID 10 | Stripes across mirrored pairs | 4 (even) | 50% | 1 drive per mirror pair | High-performance databases |
| RAID 01 | Mirrors two striped sets | 4 (even) | 50% | Less predictable than RAID 10 | Rarely the better choice vs. RAID 10 |
Hardware RAID vs. software RAID
RAID can be implemented in hardware or in software, and the choice affects both cost and flexibility.
Hardware RAID
Hardware RAID uses a dedicated RAID controller card with its own processor to manage the array, independent of the server's own CPU. Hardware RAID tends to offer strong, consistent performance and frees up the host CPU, but it costs more and can tie your array to that specific controller.
RAID implemented at the controller level is one approach to redundancy, but modern storage environments increasingly abstract it. All-flash array vendors (Pure Storage, NetApp, HPE) use their own RAID-DP or erasure coding schemes optimized for NAND characteristics.
Software RAID
Software RAID is managed by the operating system instead of dedicated hardware, using tools like Windows Storage Spaces or the Linux mdadm utility. It costs less and is more portable across hardware, since it doesn't depend on a specific controller.
Software-defined storage platforms (VMware vSAN, Ceph, Windows Storage Spaces Direct) use erasure coding policies rather than traditional RAID levels. In hyper-converged environments, the "RAID equivalent" runs in software across multiple server nodes rather than inside a dedicated array. These platforms still expose similar concepts like fault tolerance, rebuild time, capacity overhead, but the monitoring metrics and management paths differ from traditional hardware RAID.
RAID monitoring
Monitoring a RAID array helps you manage capacity, catch early warning signs, and avoid outages. A healthy-looking array can still be operating in a degraded, unprotected state after a drive failure. Three metrics matter most:
RAID array status: Monitoring the state of the RAID array gives you the best estimate of its health state. RAIDs have four states.
- Optimal: All the drives are healthy, full redundancy available.
- Degraded: One or more drives have failed, array still running but the protection is reduced.
- Rebuilding: A replacement drive has been added and the array is reconstructing parity or mirror data.
- Failed: Redundancy exhausted, data is at risk or lost.
A RAID in Degraded state is the most operationally dangerous condition in storage management. The array runs normally, no application errors occur, no latency spikes appear. But any drive failure can cause data loss.
Without monitoring, a Degraded RAID can remain undetected for weeks, especially if the failed drive's LED indicator is inside a rack that isn't inspected regularly. Monitoring helps detect any state change away from Optimal immediately.
Latency: During a RAID rebuild after a drive failure, read latency across the entire array increases significantly. In RAID 5 and 6, every read request during rebuild triggers a parity recalculation to reconstruct the missing data. The array is simultaneously serving application I/O and rebuilding, which doubles its workload. Monitoring latency during rebuild gives IT teams a measure of both application impact and rebuild progress. A latency spike that coincides with a status change to Rebuilding is expected but a latency spike on an Optimal array is not.
Capacity and utilization: Capacity monitoring for RAID must account for usable capacity, and not just raw capacity. A four-drive RAID 5 array with 4 TB drives has 16 TB raw but only 12 TB usable. A monitoring tool that shows 16 TB available is wrong. Alert thresholds should be set against usable capacity. Monitoring capacity alongside the comparison table above helps teams understand the real storage impact of their RAID level choice.
IOPS: IOPS degrades predictably during RAID 5/6 write operations (parity overhead) and during rebuild (double workload). Monitoring IOPS trends helps identify three conditions:
- A long-term IOPS decline suggests a drive is approaching failure and dragging array performance
- A sudden IOPS drop often coincides with a rebuild starting
- A sustained IOPS pattern that suddenly changes without a corresponding workload change usually indicates a drive or controller issue.
Battery: Many hardware RAID controllers include battery-backed write cache that hold recent writes in memory before flushing to disk. If the battery is failed or uncharged, the controller disables write cache and falls back into write through mode. This can reduce write performance by 50—70% on a parity-based RAID level. Without monitoring, this can be misdiagnosed as a storage bottleneck rather than a controller health issue.
RAID arrays also work alongside other storage concepts worth knowing. Multiple RAID groups are often combined into storage pools and divided into LUNs for individual applications. RAID-protected disks are frequently accessed by servers over a SAN, or delivered as file shares from a NAS.
For array-level monitoring, OpManager's approach applies regardless of whether RAID is implemented in hardware or software.
Monitor RAID with OpManager
ManageEngine OpManager is a storage monitoring software with powerful RAID monitoring capabilities. It comes with built-in support for 50+ RAID models from a diverse range of storage vendors like: Dell, HPE, Netapp, Hitachi, IBM, and Fujitsu.
OpManager auto-discovers storage devices and lists them in a monitoring inventory, classifying vendors, models, disk drives, controllers, virtual disks, storage pools, volumes, and LUN. For each device type, you can monitor capacity at the physical and logical layer, IOPs, throughput, and hardware metrics like temperature and power supply.
You can set up alerts and notifications to monitor the health and availability of your RAID arrays in real time. If a hardware or software issue arises, you'll be alerted so you can act quickly, rather than discovering the problem after the data is already at risk.
Moreover, OpManager's detailed reports can help you quickly determine the overall health and performance of your network and network devices. In addition to RAID monitoring, OpManager also extends support for tape libraries, FC switches, network devices, and servers.
FAQs about RAID
What are RAID configurations?
RAID configurations combine multiple hard disk drives or solid-state drives into a single logical unit. Depending on the level you choose, this setup can improve performance, protect against drive failure, or both, compared to using individual disks on their own.
Is RAID a backup?
Which RAID level is best?
What's the difference between RAID 10 and RAID 01?
What's the difference between hardware and software RAID?
How do I configure RAID in Windows or Linux?
What's the difference between RAID 1 and RAID 5?
Learn more about storage monitoring
Storage monitoring software
Learn how you can maximize storage uptime and performance with ManageEngine OpManager
Learn more →
By Visakh,
ManageEngine Team