With the continued woes of the traditional storage vendors – such as NetApp laying off a whopping 12% of its workforce -- the SSD space is the shining light in an otherwise grim storage market outlook for 2016. While HDD-based products are hurting in the market, SSDs posted a notable increase in sales and look poised to accelerate the inevitable transition to SSD in the marketplace.
A few months ago, I predicted that 2017 would see SSDs achieving price parity with HDDs. Enterprise SATA SSDs bought through distribution are now priced well below both traditional box vendor SSD pricing. In fact, good-quality SATA SSDs also are cheaper than enterprise HDDs from distribution. These are MLC drives, but Google recently published a report that MLC drives are just as good as STC drives and a whole lot cheaper. With the write-life problem well behind us, this paves the way for SSDs to completely replace enterprise HDDs.
There is still substantial capacity growth for SSDs in the near future and 3D stacking appears to be bringing prices down, so I’m still willing and able to predict price parity in distribution for SSDs and bulk HDDs by the end of 2017.
Learn more about the changing storage landscape in the Storage Track at Interop Las Vegas this spring. Don't miss out! Register now for Interop, May 2-6, and receive $200 off.
There is however major collateral damage from this and it likely is the root cause of EMC’s merger with Dell and NetApp’s layoffs. The mainstay of storage for decades has been the RAID array, coupled to a Fibre-Channel SAN. This class of product has a problem dealing with the much faster solid-state drives.
Host interfaces are too slow, while controllers can’t handle the random IOPS capability of even a small number of SSDs. At the same time, we are moving towards low-cost, scale-out storage based on open-source software like Ceph. Add in ODMs expanding their customer base from large cloud providers to enterprises, together with pressure on hardware prices and appliances that fit the new, slimmed down architectures that suit SSD, and traditional storage vendors are under siege.
From an end-user perspective, we are looking at an extended period of price-cutting both at the box level and the drive level. Probably the first casualty will be EMC, which will face pressure on its 10x to 20x price markup as it merges into Dell with a more realistic 3x to 5x markup. Even Dell will find its margins to the end user under pressure as 2016 rolls into 2017.
Array sales for SANs are dropping across the board. By implication, this will impact Fibre Channel sales, even though a 32 Gb version is just entering the market. 25 Gigabit Ethernet also is entering the game, with the advantages of low cost, connection versatility and a working RDMA option. Ethernet is a common connection scheme for the front end of clouds and between servers in the cloud. Life gets much less complicated if this is the only network type to deploy and that is a major driving force to move off Fibre Channel for storage.
Vendors such as Mellanox are predicting a rapid up take of 25 GbE and with Google and AWS backing the initiative to bring it to market, 25 GbE looks set for fast adoption. Intel is discussing iWARP RDMA support at 25 GbE in chipsets, which helps to drive the Ethernet migration and opens up the possibility of the highly efficient NVMe protocol replacing iSCSI or Fibre Channel as the working protocol for storage, though this will take several years to achieve.
There are some brand-new factors on the storage horizon. HPE and Supermicro both have persistent NVDIMM options for servers. These are a relatively low-cost way to boost server performance for in-memory operations by a large factor, while providing an alternative to ultra-fast SSDs or flash cards. Intel also is entering the market, together with Micron. Their offering is called 3D X-Point and is billed as being much faster than flash. These products will begin to impact servers and storage later in 2016.
This coming year looks turbulent in the storage space. New ways of business and new products will sweep away a good portion of the current solutions. Their replacements should be much faster and versatile, and also cheaper, but be prepared to be flexible on configurations and vendor relationships.