IDC just released their latest data showing that hybrid flash arrays (HFAs) accounted for around $10BN of the disk array market in 2014, growing to almost $14BN by 2018 (IDC presentation here). HFAs use a small mix of flash with conventional hard drives along with some hidden intelligence (usually embedded software) to ensure frequently accessed data is stored on the faster flash portion of the HFA. Our good friends at Techtarget have a nice article on the merits of HFA for those that wish to dig deeper here that adds to the IDC analysis.
HFA and AFA revenue forecast excluding “big data AFA” (Source: IDC DSI2015 presentation via SNIA.org)
For those that are wondering why this matters, the key point here is that this is the strongest indicator yet that mainstream storage architects are finding they don’t need all-flash storage as a replacement for their hard disks. In fact, the equally important message in the IDC report is that, on a revenue basis, AFA or all-flash-arrays, were only 1/10th the size of the HFA market in 2014.
Given AFAs are considerably higher in price (4x or higher than HFAs), this really does mean AFAs remain a smaller but important niche in the grand scheme of things. One day AFAs may become a mainstream offering as touted by our flash semiconductor friends, but fab capacity realities and technology challenges will prevent that from happening any time soon according to most industry experts, meaning HFAs will be around for some time, some saying at least 10 years.
What is not receiving a lot of publicity is the fact the hybrid storage revolution is silently moving to other places also, which Enmotus FuzeDrive is a key enabler for. For example:
Network attach server hybrids: flash and hard disk that is inside the server node itself or attached via JBODs (see our own case study here);
Hyper-convergence hybrids: direct or SAN attached (non-shared) storage inside the compute-storage nodes that is dedicated to individual server/compute nodes;
Bit data hybrids: webscale deployments of distributed flash in each of the scaleout nodes e.g. OpenStack, Hadoop.
The other missing but potentially interesting piece from our perspective, is that we are likely to see an increased adoption of hybrid AFAs (HAFAs) enabled by technologies such as MicroTiering i.e. AFAs that mix two types of solid state media, or even memory class storage with solid state, in a single storage offering to achieve a better cost point for AFA. So the story doesn’t stop at HFAs on the hybrid front.
Definitely a lot to look forward to in the coming years with respect to hybrids.