What is RAID why do I need it for a NAS?

Network Attached Storage Attached Storage is typically a type of server which can perform a multi-role job set. NAS systems can deploy many network based sharing protocols across a network be that NFS, SMB, iSCSI, FTP, AFP, SFTP, or any other storage server based role. Now to ensure data integrity defined under the CIA triad, I - Integrity, data being stored on a single drive has little to zero resiliency why? Because data stored on a single disk if that disk fails takes all your data with it. To achieve this resiliency it is done via a number of methods. The main way is via RAID drives, a RAID array is at the most basic understanding a group of disks which retain copies of data on those drives so that if a drive does in fact fail it will not make your data unreachable and will simply require a RAID array rebuild which does take time, but the reality is this is the better option than simply losing everything. Now lets get into the RAID levels and what they mean.

Words of Warning

RAID does have an effect on the life time of the disks placed into the array

This should be a fairly intuitive idea to grasp adding RAID drives impacts their overall life time they tend to use multiple data blocks, clones or parity checks across all drives if you have a drive that has a ~1 million hours before MTBF (Mean Time Before Failure) for every drive added to the array it gets divided by that number of drives. As such 2 drives becomes 500,000 hours then 250,000 hours and so on. More disks means a increasing probability window of failure, which is why you often do not use consumer class drives to run in RAID configurations as they simply do not have the disk endurance to ensure you don't lose your data

RAID is not a backup solution

You may be wondering if RAID is considered a backup solution? No it's not, it provides something called data resiliency and fault tolerance for data you have stored on your own network it does not act as a backup, never use a RAID NAS solution as a backup, that being said many platform based network attached storage solution can backup to a USB port which could be a flash drive, external hard-drive or external solid stage drive.

---

RAID types and their pros and cons.

RAID0 (Striping)

In this disk configuration disks are configured in a "stripping mode" this means data blocks from each write request is distributed across multiple disks which improves overall performance, more disks in a RAID0 configuration the faster the group of disks are. Now this is fine for systems which use a RAID0 drive as for example a caching disk the loss of the cache is no catastrophic and no real harm is done. However because the data is written across multiple disks it means that if you lose a "data block" which is shared between two drives or more that entire data block is gone and unrecoverable which is typically why RAID0 is not recommended for anything that requires any degree of data resiliency, one other thing to do RAID0 works a lot like something called a JBOD which is short for "Just a Bunch of Disks" and the total space of the RAID array will be the accumulated total of disk space summed together.

For example if you have 4 2TB Hard drives or Solid State drives, the total disk space available will be their total sum so 8TB in total again I cannot stress this enough RAID0 has zero data redundancy and resiliency once one drive is gone the entire array is gone.

RAID1 (Mirroring)

If you are after data resiliency then RAID1 is most desirable the reason being is data is mirrored across all disks, so if the 4 2TB drive example from before is used the data is "cloned" across all disks. This disk configuration requires minimum 2 disks because 1 is the primary the other is the secondary redundant disk, if one fails the second drive has a complete copy of the entire drive on the other. This is a great solution for someone who has very strict requirements on their data integrity and reliability, but it will only allow you to use 2TB across the board because data is cloned from 1 to all the 4 other drives giving you 3 extra copies of your data, also it should be noted if you have 4 2TB drives the total available space will be reduced to half the total capacity ability so 2TB becomes 1TB available, that being said RAID1 is very fast.

RAID5 (Single Parity with Striping)

Has a good balance of resiliency and speed, requiring a minimum disk count of 3 drives or more, data is stored in a parity based system where every data block written has to write a parity block across other disks. The parity block when an array drive fails allows the rebuild from the parity blocks but this takes 'rebuild time' that time can be up to a day. This also provides a good amount of data redundancy allowing a single drive to fail and the array to continue functioning.

RAID6 (Double Parity with Striping)

Much the same as RAID5, this uses multiple disks while also doubling the number of parity writes this offers high speed read performance but much slower write performance because parity checks have to be written to disk twice rather than once. This can suffer a secondary disk failure so this is the best option when you want to maximise array protection if you only have a small number of disks.

RAID10 (Mirroring RAID1 with RAID 0)

Has a high investment cost most customers won't choose this unless they absolutely require it. As a result RAID10 often needs a lot more disks beyond any of the previous RAID configurations up to groups of drives 4 or more at a time are grouped into RAID1 then 4x drives per RAID0 when a RAID0 array fails copying back from the other multi-nested RAID disks is possible.

RAID## (Nested)

There are RAID configurations beyond simple single digit RAID configuration this allows any number of disks to be configured in it's various nested configurations most customers will never need this or want this it's simply overkill for anyone who doesn't need absolute performance and deeper layers of redundancy beyond single digit RAID configurations, nested RAID often requires specialized RAID hardware controllers because software RAID is simply not up to scratch when dealing with nested RAID, the controllers also sometimes have caches and power backups as well for those caches in the form of batteries