A fundamental keystone of data protection, Raid (redundant array of independent disks) has been around for decades.
The building blocks are very simple. Multiple disk drives allow data to be distributed via striping and mirroring, or sets of parity data allow a failing unit to be rebuilt by reference to data on healthy media.
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
But what has been the impact of solid-state media on the use of Raid?
We’ll look in detail at the Raid options available from the main flash array makers below.
Broadly speaking, however, when an all-flash array is a legacy product that’s been retrofitted with flash capacity then you find Raid levels that span the possible combinations of mirroring, striping and parity are available (ie, 1, 10, 5, 6), as they would have been in the HDD-equipped product.
Meanwhile, options are more limited on many of the newer startups’ all-flash arrays and some of the big six suppliers’ newer arrays, and seem to major on parity-based Raid levels (5 and 6) – sometimes their own flavour with branding to suit – and even in the case of NetApp, its own triple parity Raid.
But that’s a broad sweep. Let’s recap on the basics of Raid.
While Raid is implemented in many variants, it is all built on combinations of possible characteristics based around striping (which aids speed of access more than data protection), mirroring and parity.
So, for example, Raid 1 mirrors data between disks while Raid 10 stripes data across mirrored pairs, and these provide good read and write performance.
Meanwhile Raid 5 is based on parity for data protection/rebuilds and maximises capacity, but delivers less well on write input/output (I/O). Raid 6 boosts data protection over Raid 5, due to the extra parity data it stores.
Stripe set size is the number of disks that data is written to. With a bigger stripe set size, data is written to more drives and you’ll get better I/O performance.
Be aware, however, that large stripe sets can result in disk rebuild issues due to unrecoverable read errors.
And, of course, Raid rebuild times increase significantly as drive capacities grow.
Raid has continued to evolve and we have seen new protection methods that use the essential components of Raid, but that distribute data and parity information in new ways.
So is Raid relevant to the world of flash drives?
Well, yes. Despite their differences with spinning disk HDDs, flash drives do fail.
There is always a risk of component failure and eventually a flash drive will fail because they have limited write I/O capacity.
This means some protection is required to cater for failure scenarios, and suppliers that sell flash have implemented Raid in ways that vary from the standard levels to their own proprietary formats.
Raid in flash array products
Dell EMC offers all-flash and hybrid flash VMAX, Unity and SC series arrays. These all use standard Raid levels, with 5 and 6 in VMAX in configurations that include 7+1 parity disk, as well as 3+1, 6+2 and 14+2. Unity and SC series arrays offer Raid 5 and 6 plus Raid 10, or Raid 1/0 as Dell EMC calls it. The all-flash-only XtremIO uses its own XDP, which stands for Xtremio Data Protection, a so-called “modified diagonal parity Raid-6 variant”.
Fujitsu’s block-access Eternus all-flash arrays come as the small and medium sized enterprise-targeted AF250S2 that scales to 737TB and the enterprise-class AF650S2 that goes to nearly 3PB. Raid levels available are 0, 1, 10, 5 and the parity/striping Raid 50.
Hewlett Packard Enterprise’s (HPE) high-end 3PAR StoreServ SAN platforms come in three models that scale up to several petabytes in capacity. They can all be configured as all-flash or hybrid (as well as all-HDD) and offer Raid 1, Raid 5 and Raid 6 data protection. Early in 2017, HPE acquired all-flash startup Nimble Storage. Those arrays come in three sizes that range from a few TB of capacity to several hundred and have Nimble’s triple-parity Raid with claimed six nines availability.
Hitachi Vantara’s F-series all-flash arrays in its virtual storage product line come in four models that go up to capacities of around 2,300 drives. They support Raid 10, Raid 5 and Raid 6.
NetApp – NetApp’s all-flash AAF (All-Flash FAS) arrays provide for Raid DP (its implementation of Raid 6) or Raid TEC, which stands for Triple Erasure Coding. This is a feature of NetApp’s Data ONTAP 9 operating environment that provides for a third parity disk to ensure the safety and speed of disk rebuilds. Meanwhile, NetApp’s SolidFire arrays use their own Helix “Raid-less” data protection in which two copies of data are distributed through nodes in the cluster. Finally, NetApp’s EF series flash arrays use Raid levels 0, 1, 5, 6 and 10. E-Series arrays date back to NetApp’s acquisition of Engenio in 2011, whose arrays were designed for spinning disk.
IBM offers the all-flash FlashSystem and Storwize all-flash arrays, as well as the DS8880F products. The FlashSystem V900 offers between a few tens of TB and nearly 2PB with Raid 5. The FlashSystem 900 comes with 3D TLC NAND and Raid 5.
IBM adds so-called “variable stripe Raid”, which provides for bad data planes in the NAND chip to have their data shipped to healthy ones. Adding this feature to system-level Raid 5 is called Two Dimensional Raid by IBM.
IBM’s StorWize F-suffixed all-flash arrays come with Raid levels 0, 1, 5, 6 and 10. The V7000F supports 760 drives of up to 15TB (3,040 in a cluster) while the V5030F supports up to 1,520 in a cluster.
The DS8880F series all-flash arrays come with Raid levels 5, 6 and 10 with capacities that go over 1PB.
Kaminario offers the K2 all-flash array, which comes in capacity units of K-blocks that can take it up to around 4PB. Kaminario also favours dual parity in its K-Raid scheme.
Pure Storage’s FlashArray//M arrays come in a number of sizes from a few tens of TB to over 1PB. Raid HA is Pure’s take on things and is a dual-parity Raid scheme, so like Raid 6.
Western Digital-owned Tegile’s IntelliFlash arrays offer block and file access in nodes that range from a few hundred TB to nearly 2PB. The company’s datasheets aren’t too forthcoming but speak of “multiple Raid/mirroring options” that include dual-parity and two- and three-way mirroring.
Tintri’s key all-flash offerings are the EC series that range from tens of TB to tens of PB. Its T1000 offers all-flash for remote office/branch office locations. Dual parity Raid 6 is the only option in Tintri.