Parity RAIDs: RAID-5 and RAID-6

Hello, my fellow knowledge-seeking friends. If you work closely with storage, are interested in the technologies, or have just started the journey to learn more about it, you most likely have heard the term RAID. For those who don’t know, let me quickly introduce you to it. RAID, or Redundant Array of Independent Disks, is a technology of combining multiple drives into a single storage system to improve performance, add redundancy, or both. Different RAID levels use different methods to store and protect data.

In this article, we will focus on Parity RAIDs – RAID 5 and RAID 6. These RAID levels use a special calculation called parity to store extra information alongside the actual data. This parity is used by the system to rebuild lost data if a drive fails, without keeping full copies of everything.

We are focusing on parity RAIDs because they offer a strong balance between usable storage space, performance, and fault tolerance – making them a popular choice for many business and home storage setups.

So, without any further introduction, let’s start our discussion.

What is RAID 5?

RAID 5 is a configuration that splits (or stripes) data across multiple disks and also stores parity information, which is distributed across all the drives instead of being kept on a single one. Parity allows you to rebuild the array and recover data if any one drive fails.

Now, a bit of history. RAID 5, as well as the whole concept of RAID, was introduced in the late 1980s. At that time, it was a solution to a common problem – the price of high-capacity drives. The idea was to combine several smaller, less expensive drives into a single logical unit for better performance and/or reliability. The RAID 5, in particular, emerged as a way to get redundancy without a sacrifice of 50% of storage with RAID 1 mirroring, while avoiding the performance bottleneck of storing all parity on a single dedicated drive as in RAID 4.

From its introduction through the 1990s, RAID 5 was rapidly adopted across different storage solutions. It was very valuable in cases when disk capacities were small and prices were high, making its storage efficiency a major selling point. However, over time, the capacity of drives has grown, which has made rebuild time and the risk of losing another drive during the rebuild a bigger concern. It led to interest in the successor of RAID 5 – RAID 6, as well as other configurations with improved fault-tolerant designs.

Still, RAID 5 never disappeared. In fact, it’s been making something of a comeback. The reason is flash storage (namely SSDs). With their low failure rates and incredibly fast rebuild times, the classic RAID 5 problem – losing a second drive before the first rebuild is finished – becomes much less of a threat. Today, many administrators see RAID 5 paired with SSDs as a perfect balance of performance, capacity efficiency, and failure protection. Though, admins should keep an eye on TRIM command support by SSD device and aligning RAID stripe size to minimize write amplification. We’ll cover this topic later.

RAID 5 Architecture

The minimal configuration of a RAID 5 array consists of three drives. Here is an example of what RAID 5 looks like:

As shown above, each set of data is split into parts and stored across all the disks, while parity information is distributed evenly among them. A RAID 5 array can withstand the failure of one drive without data loss. If a drive does fail, the array enters what’s called “degraded” mode, continuing to operate and providing access to data. Once the failed drive is replaced, the array begins the rebuild process, which temporarily reduces performance.

Now, let’s calculate the usable capacity of a RAID 5 array with a simple example:

• Drives: 3 × 2TB each
• Total raw capacity: 3 × 2TB = 6TB
• Usable capacity: (3 − 1) × 2TB = 4TB

In this case, 2TB of space is effectively used for parity information, which is distributed across all drives.

What Problems Does RAID 5 solve?

The main benefit of RAID 5 is protection against a single drive failure without wasting half the capacity on full duplication. It also provides a middle ground between cost, efficiency, and performance. For organizations that don’t want the expense of RAID 6 or more advanced setups, RAID 5 is often “good enough.”

Advantages and disadvantages of RAID 5

RAID 5 still offers one of the best trade-offs between capacity, redundancy, and cost. In a three-disk setup, two-thirds of the total space is usable, and as you scale up, efficiency rises further. Reads are generally quick since data can be pulled from multiple disks at once, and the economics are hard to beat: you only “lose” one disk’s worth of space to parity.

But the weaknesses are equally important to understand. Every write involves more work because parity has to be updated, and this creates a write penalty. So, not only does parity itself add overhead, even a small data change can trigger a block-wide read-modify-write cycle. That’s why write performance in RAID 5 is significantly lower than read performance.

There’s another subtle issue that often compounds this: stripe size alignment. RAID 5 arrays are rarely built with stripe sizes that line up neatly with filesystem blocks. Ideally, you want your stripe to span 2N disks, for example, 2 or 4 drives in RAID 0, or 3, 5, or 9 disks in RAID 5. But odd-sized RAID 5 setups are less common, and when blocks don’t align properly, almost every write ends up incurring extra read-modify-write cycles. This misalignment means that even writing a full block can still behave like a partial write, amplifying the overhead.

On top of that, RAID 5 redundancy only protects against a single disk failure. If another drive fails during rebuild, or even throws a bad sector, the entire array can collapse. And the larger the drives, the slower the rebuild, which increases the risk window.

These factors make RAID 5 less attractive for heavy-write environments or for very large HDD arrays, even though it remains strong for read-dominated workloads.

RAID 5 use cases

In general, RAID 5 is very good when you need to have a perfect balance between performance, storage efficiency, and redundancy. Now, let’s be a bit more specific:

• File and Application Storage: For many standard file and application servers, RAID 5 provides a solid balance of read performance, capacity, and fault tolerance. It is best suited for workloads where reads dominate over writes.
• Web Servers: Websites typically perform more reads than writes (for example, serving static files). RAID 5’s read performance makes it a suitable option.
• Media Streaming: RAID 5 can pull data from more than one drive at the same time, which is helpful for media streaming since it’s mostly about reading files, not writing them.
• Archives: RAID 5 is a good fit for data that changes rarely but must remain accessible at reasonable speed, particularly when maximizing available storage is important.

I hope you now have a strong understanding of RAID 5, so we can move on to discuss its successor – RAID 6.

What is RAID 6?

As we just pointed out, RAID 6 is a successor to RAID 5 and its natural evolution. RAID 6 is a fault-tolerant storage configuration that expands on RAID 5 by using two sets of parity information instead of just one. Like RAID 5, it stripes data across multiple disks for performance, but it calculates and stores two independent parity blocks for every stripe, distributed across all drives in the array.

RAID 6 Architecture

As the evolution of RAID 5, RAID 6 improves the architecture. The minimal configuration requires four drives to configure RAID 6. Here is an example of RAID 6 architecture:

Like in RAID 5, RAID 6 splits each set of data into parts and stores them across all the disks, and parity information is distributed evenly among them. A RAID 6 array can survive two drive failures without any data loss. RAID 6 has the same behaviour when it faces a drive failure; it switches to a so-called “degraded” mode. When a failed drive is replaced, the array starts the rebuild process, which temporarily reduces performance of the array.

Now, let’s calculate the usable capacity of a RAID 6 array with a simple example:

• Drives: 4 × 2TB each
• Total raw capacity: 4 × 2TB = 8TB
• Usable capacity: (4 − 2) × 2TB = 4TB

In this configuration, 4TB is used for parity information, spread across all drives.

What problems does RAID 6 solve?

RAID 6 directly addresses RAID 5’s biggest weakness. It can survive two simultaneous failures, it protects better against unrecoverable read errors (URE) during rebuilds, and it provides higher reliability for very large arrays. For enterprises and datacenters that cannot afford downtime or data loss, RAID 6 offers an additional layer of security.

Advantages and disadvantages of RAID 6

RAID 6 improves on RAID 5 by allowing the array to survive two simultaneous drive failures, but that extra layer of safety comes at a price. The most obvious drawback is write performance. Every write requires not one but two parity updates, so even a small change can turn into a block-wide read-modify-write cycle on every involved disk. This means the write penalty is heavier than in RAID 5, and the performance gap widens in write-intensive workloads.

The same alignment issue that plagues RAID 5 applies here too, and often in a worse way. If the stripe size doesn’t line up with filesystem blocks, which is common, since RAID 6 setups rarely use stripe widths that match 2N alignment, then writes are almost always misaligned. As a result, the array ends up doing unnecessary read-modify-write operations even when you’re writing a full block of data, amplifying the penalty further.

On top of that, RAID 6 always reserves the equivalent of two disks for parity, so usable capacity is reduced accordingly. With smaller arrays, that overhead can feel steep. Rebuilds on large disks remain slow and stressful for the system, and because dual parity is more complex to calculate, RAID 6 often needs a strong controller to keep performance acceptable under heavy load.

These trade-offs don’t erase the value of RAID 6, but they do explain why it’s best reserved for systems where uptime and data integrity matter more than raw write speed or maximum usable capacity.

Now that we have a better understanding of RAID 6, we can discuss its use cases.

RAID 6 use cases

RAID 6 is the choice when data safety outweighs capacity and write speed penalties. It’s especially suited to enterprise systems handling critical information, like financial databases or healthcare records, where downtime or data loss would be unacceptable. Backup storage benefits from RAID 6 too, since it adds protection during the time gap between backups. Archival systems and large multimedia libraries also use RAID 6 for reliability, as they hold vast amounts of data that must remain safe and readable over the long term.

RAID 5 vs RAID 6

Now, we better understand both RAID levels and can compare them directly:

The comparison comes down to priorities. RAID 5 is efficient, cost-friendly, and with SSDs it’s safer than it used to be, making it a solid fit for smaller arrays and read-heavy environments. RAID 6, meanwhile, sacrifices capacity and write speed for greater fault tolerance. It’s the safer choice for large HDD arrays or mission-critical systems where two drives could realistically fail before a rebuild is done.

Software and Hardware RAID

I also want to touch on the topic of hardware and software RAID, as both are very popular and capable of implementing RAID 5 and 6. Let’s look at each approach more closely and then make a quick comparison.

Software RAID

Software RAID is a method of creating a RAID system where all the RAID functions are handled by the operating system’s software rather than by dedicated hardware. The OS manages how data is striped, mirrored, or stored with parity across multiple disks. Software RAID uses the system’s CPU and RAM to perform RAID calculations, including parity generation for RAID 5/6.

Common tools include Linux MDADM, Windows Storage Spaces, GEOM (FreeBSD), Btrfs and ZFS (both are file systems with RAID capabilities, so arguably could be called software RAID).

Software RAID, like any other technology, has its advantages and disadvantages. Let’s look at both, starting with advantages:

• Low cost of implementation: Software RAID does not require any dedicated hardware to implement.
• Flexibility: Arrays can usually be migrated between machines running the same OS.
• Transparency: Software RAID is easier to monitor and configure at the OS level.

And disadvantages are:

• Performance impact: Uses CPU cycles and memory, which can affect workloads on busy systems.
• Fewer enterprise features: Lacks advanced caching and battery-backed write protection found in hardware RAID controllers.
• Scalability limits: Generally better for small to medium arrays, not very large enterprise deployments.

Good. Let’s check what hardware RAID is.

Hardware RAID

Hardware RAID is the opposite of Software RAID and is a method of creating a RAID where all the functions are handled by dedicated hardware. It uses a specialized RAID controller card (PCIe card, HBA with RAID features, or embedded server chipset) to manage all RAID operations. The controller has its own processor and often dedicated cache memory (sometimes battery- or flash-backed) to handle parity calculations, striping, and mirroring without relying on the host system’s CPU. Drives connect directly to the controller, and the host system sees the entire RAID array as a single logical disk.

Like with the software RAID, hardware RAID has its advantages:

• Performance: Offloads RAID work from the system CPU, improving efficiency, especially important for parity-based RAID levels (RAID 5/6).
• Advanced features: Write-back cache with battery/flash protection, hot-swap support, and background rebuilds.
• Reliability: Purpose-built for enterprise workloads, with consistent and optimized performance.

As with any solution, hardware RAID also comes with disadvantages:

• Cost: RAID controllers can be expensive, especially enterprise-grade models.
• Vendor dependence: Arrays are often tied to a specific controller model — if the controller fails, you typically need the same model to recover the array.
• Less transparent: Since the OS only sees the logical array, you may have less visibility into individual drives without vendor tools.

Choosing Between Them

As you can see, software RAID delivers flexibility and low cost by relying on the host system’s resources, making it attractive for smaller or budget-conscious setups, while hardware RAID justifies its higher expense with faster performance, enterprise features, and stronger reliability. The choice ultimately comes down to value: software RAID offers simplicity and affordability, whereas hardware RAID provides resilience and performance where uptime truly matters.

Alternatives?

Like with other industries, the storage/IT industry is driven by demand. So, what if RAID 5/6 performance is not enough? How can the write performance be increased without losing much efficiency?

The solution was to combine RAID 5/6 with RAID 0. As a result, we got RAID 50 and RAID 60. Both levels work in the following way: drives are combined in multiple RAID 5 or 6 groups that are then formed in a single RAID 0. Here are the examples of RAID 50 and 60:

The data striping of RAID 0 on top of RAID 5 or RAID 6 allows RAID 50 and RAID 60 to perform write operations in parallel across multiple groups, which increases overall write performance. The parity overhead still exists within each group, but RAID 0 striping distributes the load, making the process faster.

Another popular alternative is the Zettabyte File System (ZFS), which integrates volume management (like RAID) with the file system itself. ZFS features RAID-Z configurations, combining storage, redundancy, checksumming, snapshots, and repair in a unified approach.

For today, we’ll focus only on the RAID-Z aspect. ZFS supports the following RAID-Z levels:

• RAID-Z1: Equivalent to RAID 5 (single parity).
• RAID-Z2: Equivalent to RAID 6 (double parity).
• RAID–Z3: Triple parity (unique to ZFS, tolerates three simultaneous drive failures).

What Problems Does ZFS Solve as an Alternative to RAID 5/6?

ZFS addresses several weaknesses of traditional RAID:

Write Hole Problem: A power failure or crash during a write can leave parity and data out of sync, corrupting the array. ZFS is copy-on-write, it never overwrites data in place. New data and parity are written together, and only once the write is confirmed, then the old block is discarded. This eliminates the write hole problem.

Silent Data Corruption (Bit Rot): Traditional RAID doesn’t always detect silent corruption (a block that changes without an error being reported). ZFS uses end-to-end checksumming for every block of data and metadata. If corruption is detected, ZFS automatically repairs it using redundant copies (from parity or mirrors).

Rebuild Risks with Large Drives: Rebuilds on multi-terabyte drives take hours or days, increasing the chance of a second failure or URE. ZFS rebuilds only the used data blocks, not the entire drive. This makes rebuilds faster and reduces the risk window.

Separate RAID + Filesystem Layers: RAID and the filesystem are separate; RAID doesn’t know about filesystem data structures. This limits error recovery. Because ZFS integrates RAID functionality directly with the filesystem, it has full awareness of data layout and can self-heal at the file level, not just at the block level.

Need even more protection?

RAID was developed and introduced to provide the data redundant storage and protect it from drive failure. However, RAID provides only local redundancy. Nowadays, we rarely use a single host in the cluster. The norm for today is multiple node clusters or fleets of servers, or even fleets of clusters. Resources are shared not only within a single cluster but also across multiple clusters or fleets. One of those resources is storage, where all the data sits.

So, how do we protect data now? The challenge is not about hardware failures within a single server, but also about resilience against node failures. The answer lies in software-defined storage (SDS) solutions.

Within the joint DataCore-StarWind portfolio, there are two products designed for this purpose:

• StarWind VSAN: A lightweight, software-defined storage that is easy to deploy and manage. It is hardware agnostic. It delivers high availability by mirroring storage across servers without the need for expensive dedicated hardware, making it a cost-effective solution for both SMBs and enterprises.
• Datacore SANsymphony: Enterprise-grade software-defined storage solution focused on performance, security and large-scale efficiency, using intelligent caching and auto-tiering to optimize storage utilization while ensuring continuous uptime of businesses’ apps and services.

Both products address many of the resilience and performance limitations of classic RAID 5 and RAID 6 by adding advanced features such as synchronous replication, automated failover, and simplified management.

Conclusion

Parity RAIDs, RAID 5 and RAID 6, still matter. RAID 5, once thought risky with large HDDs, has found new life thanks to affordable SSDs that shrink rebuild times and reduce the danger of dual failures. RAID 6 remains the cautious choice, designed for systems that cannot risk downtime or data loss, even if that means giving up capacity and speed.

The right choice depends on your environment. RAID 5 emphasizes efficiency where budgets are tight and workloads are light on writes. RAID 6 emphasizes resilience for large arrays and critical data. And when combined with software-defined storage, both become building blocks in far more resilient, flexible infrastructures than classic RAID alone could ever provide.