A bit of thought about RAID

Given my present job role, I am often confronted with some assumptions people have about RAID arrays.

Thinking of those conversations, I decided to write a small and human-readable (I believe) article about this.

Let’s start by explaining what Hardware RAID is.
A RAID array, based in hardware requires a controller (AKA RAID Controler) physically installed in the system. Depending on your goal, and available hard drives, you can opt for several RAID levels—Please check about them here.

For not, let’s focus on what RAID can and cannot do for you.

What RAID will not be able to assure you?


Uptime of 100%. While a RAID based logical volume can offer some resilience and prevent some downtime, like many other systems, this is not a magical solution to have zero downtime issues. With RAID, the risk associated with hardware still exists. If a RAID Controler fails, you will have downtime to replace it. However, failures on such components are indeed less current than for example a hard drive failing, given its moving parts.


RAID is NOT a backup solution!

A frequently tested and properly planned backup solution will never be replaceable with a RAID array of any sorts.
You can suffer from data corruption, security incidents (ramson ware attacks for example), a myriad of human errors, and only a proper backup solution will rescue you by restoring a good set of data to replace the one you lost.

RAID will only protect you against hard drives fail on a running array an give you margin to replace the faulty drives (up to a limited number) without losing data or having downtime, but only this. I know I am repeating myself, but RAID is NOT a backup solution.

I would even advise that before considering RAID-based hardware, you should think ahead of a backup solution if you still don’t have one.


RAID does not always allow dynamically increasing in size the existing array.

If you need more space, in most cases, you cannot only add another drive to the array. Most of the times, it is true that you are going to have to start from the beginning, rebuilding and reinitializing the array.


Is RAID a good option for virtualisation and high-availability scenarios?

The answer is a straightforward No.

For such environments the right choice is based in SAN storage.

RAID levels, what are they and what they offer

RAID 0 (also known as Striping)

RAID 0 will use any number of disks and will merge them, creating one large logical volume. They are usually used to achieve high speeds. Such is possible because reading and writing happen from multiple disks at the same time.

However, this has a cost. RAID 0 does not offer redundancy at all. Loosing an individual hard drive will result in complete data loss.

To be honest, This RAID level is less reliable when compared on having a single hard drive, for example.

RAID 0 can be useful, for example for cache purposes, where speed is vital while losing data would not matter. In the short RAID 0 should only be used for SPEED scenarios.

RAID 1 ( Or mirroring)

In short, almost every use case of RAID 1 is where you have two identical hard drives, and they will have the data mirrored/copied.

The use of RAID 1 is mostly for redundancy. Imagine that you lose a hard drive. With RAID 1 you will remain with the system up, relying on the remaining hard drive.

The broken hard drive can be swapped most likely with no downtime (but exceptions may occur).

There is also some benefit on what matter increased read performance with this level of RAID; once data can actually be read off any of the existing hard drives in the array.

In parallel, there will be a slightly higher write latency due to the fact the data will be written to both hard drives.

RAID 5 and 6

RAID 5 requires at least 3 hard drives. RAID 6 requires at least 4.

RAID 5 and 6 use the concept of RAID 0 when it comes striping the data across multiple drives to increase performance. However, it introduces redundancy. For that, the parity of information is distributed across the hard drives.

I will not bore you with detailed technical aspects in this article. Instead, let’s focus on what you should know about RAID 5 and RAID 6 regarding losing hard drives.

With RAID 5 you can lose only one har drive. If you have more than one hard drive failing, your array is no longer healthy, and there is the loss of data.

RAID 6 allows you to have up to two failed hard drives before your array is compromised.

Both RAID 5 and RAID 6 will improve your read performance dramatically. Regarding writing to the hard drives, performance depends grandly on the RAID controller you chose for using and its computing capacities.

The fact that the RAID controller should be an independent component in your system is justified by the need to calculate the data parity and write it across all hard drives.

RAID 5 and RAID 6 are often good choices for file servers, standard web servers, archive solutions ( near and deep) and other systems where most of the I/O transactions are meant to be reading.

If you are planning to run heavy writing to the hard drive systems, these RAID levels of RAID are not the right choice.  As an example, a database server would not perform well on a system with storage based on RAID 5 or 6

A Note on performance and hard drive loss and size of drives

A RAID 5 or 6 system is going to have a considerable impact in performance in the eventuality of losing a hard drive. The performance will have to be sacrificed to assure that the environment stays operational. When replacing the failed drive, the rebuilt process will have to start. Such a process will use a significant amount of the total performance of the array. The rebuild times are getting longer and longer, in line with the hard drives getting bigger and bigger.

RAID 10

RAID 10 can run with a minimum of 4 drives.

It combines RAID 1 and RAID 0. Usually, this is the type of RAID you would look for if you’re looking for speed and still having redundancy.

Let’s take the example of a four hard drives configuration. Two hard drives will be mirrored holding half of the striped data and on the other two, also mirrored, the other half of the data. 

Practically such RAID level assures you that you can lose any single drive, and then perhaps a second drive, without losing data.

Bear in mind that like RAID 1, there will be only the capacity of half the drives, but read and write performance are improved.

Nested RAID levels

Also known as hybrid RAID. They combine two or more of the standard RAID levels. I promise I will write an article alone on these RAID Levels in the short future.

Close Bitnami banner
Bitnami