hywards - Fotolia

How does RAID 5 or RAID 6 affect storage performance and capacity?

RAID provides workload resilience and protects against data loss, but not all levels of RAID are made alike. What are the storage tradeoffs for RAID 1, RAID 5 and RAID 6?

Workload resilience, data availability and efficient storage use are all crucial when working with critical workloads....

Without the proper security in place, a drive failure can completely wipe out your data. RAID spreads data out across multiple disks to protect against failures; should a disk failure occur, your data still exists at a second location, maintaining consistent computer speed and availability.

In the past, VMware Virtual SAN (vSAN) offered data protection in the form of RAID 0 or RAID 1, but more current versions of vSAN offer a larger number of RAID options. VSAN 6.2 supports RAID 1, RAID 5 and RAID 6. Each option carries a tradeoff in performance and storage capacity that an administrator should carefully evaluate before choosing to use RAID 1, RAID 5 or RAID 6.

Consider RAID 1 (mirroring), which basically copies one disk to a duplicate disk that mirrors the original. Each disk has a mirror, so the storage demands for RAID 1 are twice the requirements for the original data.

For example, a disk requires 200 GB to apply RAID 1 to a 100 GB disk. Both disks can run at full speed, and there is no parity or other encoding computations with which to contend. However, RAID 1 only protects against one disk failure.

It's important to note that vSAN 6.2 can handle a double mirror option, which enables using two mirrored disks for every data disk. This protects data from two simultaneous disk failures -- if an original and one mirrored disk fails -- but demands three times the storage capacity. For example, vSAN 6.2 requires 300 GB of capacity to protect against two faults on a 100 GB RAID 1 disk. 

The addition of RAID 5 support means that data is spread out across multiple disks and is interspersed with parity data. With four disks, this requires only an additional 33% storage capacity for the parity data. So a four-disk RAID 5 group storing 100 GB of data needs 133 GB of capacity to handle the additional parity, while still yielding the 100 GB of data capacity. RAID 5 recovers up to one disk failure, and the simple mathematical process means there is no perceivable performance penalty.

As with RAID 5, RAID 6 support enables you to spread data out across multiple disks, but a second layer of encoding, interspersed across the group, supplements the basic layer of parity data. With six disks, this requires an additional 50% storage capacity for the parity data. So, a six-disk RAID 6 group that stores 100 GB of data needs 150 GB of capacity to handle the additional encoding layers and to still yield the 100 GB of data capacity. RAID 6 recovers up to two simultaneous disk failures. Today's advanced processor instruction sets process RAID 6 erasure codes with no significant performance impact.

The general decision to use RAID 1, RAID 5 or RAID 6 in vSAN 6.2 depends on the organization's desired tradeoff between storage resilience and the cost of the added storage capacity necessary to support the RAID level.

There are other general considerations for RAID 5 and RAID 6 under vSAN. For example, RAID 5 and RAID 6 erasure coding are only available for all-flash disk groups, and the disks that the group involves must use on-disk format version 3.0 or later. Stretched vSAN clusters currently do not support RAID 5 and RAID 6. A valid vSAN license is required to enable RAID 5 and RAID 6 on a cluster.

Next Steps

Avoid these data storage capacity pitfalls

Get IT resiliency without purchasing new hardware

Create a money-saving storage capacity plan

Dig Deeper on Backing up VMware host servers and guest OSes