How flash storage raced from glory to despair in 2013
And back again. For some...
Flash dramas abounded in 2013, from triumphant IPOs, disastrous IPO after-effects, profound strategy changes, and firms crashing out in flames.
Two CEOs went. Two companies crashed. Several were bought. All in all it was a hell of a year. Let's start with all-flash arrays.
Here's where a lot of the flash action was as all-flash arrays (AFAs) went mainstream. The legacy disk drive array suppliers all went into the all-flash array business as they built product to respond to Numbus Data, Pure Storage, Solidfire and Violin Memory arrays:
- EMC had its XtremIO product on limited availability but went GA late in the year. Rumours are that it's now the the AFA volume leader.
- HDS doubled the maximum capacity of its HAF (HDS Accelerated Flash) to 38.4TB. This is an all-flash array (ASIC, SSD + SW) inside a VSP, HUS VM or HUS 150 system.
- HP introduced its all-flash 3PAR 7450.
- Dell announced all flash versions of its EqualLogic iSCSI and Compellent FC arrays, saying its Compellent array software worked very well with flash, including automated data placement across two flash tiers (SLC and MLC) and disk tiers.
- We're still waiting on NetApp's ground-up design FlashRay with the EF550 appearing during the year, upgrading the EF540, and both being all-flash iterations of an E-Series array with a primarily disk-based OS.
- We're also still waiting on IBM for its FlashSystem renaissance, likely to happen later this month.
- Skyera announced a 21PB-in-a-rack monster costing $1.99/GB in August. The skyEagle packs 500TB of NAND into a 1U box, an amazing density that blows everybody else away.
Skyera skyHawk design.
Frankly, if Skyera's boss, Rado Danilak, and his team hadn't developed the SandForce controller business, bought by LSI in October 2011, people wouldn't have believed the density of this skyEagle box. It was a pre-announcement, with product delivery slated for the second half of 2014, possibly earlier.
In March IBM's new storage general manager, Ambuj Goyal, said that transaction data should move from IBM disk arrays to flash arrays; "Our play is the all-flash array for transaction processing, not hybrids." In his view the flash array controller should sit behind a [mature] disk array controller, like a DS8000 or Storwise one. Then, in April, Big Blue announced a $1bn investment to boost its flash business.
Flash arrays or appliances have become the preferred place in the enterprise data centre for storing random data with high access rates.
Gamely countering the flash hardware appliance trend VDI technology supplier GreenBytes got out of the flash hardware business and transitioned to selling its VDI software with no hardware, thinking that flash arrays were headed towards commodity status and making profits would be darned difficult.
The startup AFA momentum is with Pure and Solidfire, Violin having been hobbled - see below.
Flash in the cloud
In April Switzerland-based CloudSigma offered its IaaS services using all-flash storage.
November saw Amazon put out compute instances with flash storage in its cloud. December saw us writing about Performance Servers from Rackspace which were stuffed with flash.
Google also offered flash storage for compute-in-the-cloud instances in December.
That's it really; game over; flash has arrived as a server accelerator in cloud server instances from CloudSigma, Amazon, Google and Rackspace. Any other cloud compute provider has to match that or cut prices to suicidally low levels.
Flash came to the cloud with no fuss really but fuss there was amongst the companies building and supplying flash storage; major league fuss.
Acquisitions, investments and and crashes
One view is that many flash cowboys went and struck it rich while others just struck out. Here's a summary:
- Seagate invested $40m in PCIe flash card HW and SW startup Virident, and said it would sell Virident PCIe flash cards, only for WD to buy Virident later in the year. Oops
- WD invested in all-flash array startup Skyera in March
- Cisco went and bought all-flash array startup Whiptail for $415m in September and then confused us all by saying its technology would be integrated with UCS servers anthem emphasis would not be on selling stand-alone flash arrays
- Greatly troubled OCZ crashed and Toshiba bought its assets
- WD went on a buying spree:
- Buying VeloBit and
- Buying Virident and
- Buying sTec and
- Assigning the three to its HGST sub
- SanDisk bought SMART Storage and its FlashDIMM technology in June
- Avago, the re-named Avaya, is buying LSI, partly for its PCIe flash card business
- Hybrid flash/disk array startup Starboard Storage crashed and burned, with SGI picking up the assets
- EMC bought ScaleIO in July for its scale-out virtual SAN technology, using a server's direct-attached storage, including flash
The big news was effectively a no news item; despite encouragement from all quarters Seagate did not buy a flash technology company. Names mentioned included Violin, Fusion-io and Virident, with Seagate apparently turning down an offer to buy the latter. Disk in the cloud was more appealing to Steve Luczo's company than buying additional flash presence in servers, desktops or networked arrays.
This contrasts starkly with WD's pro flash product initiatives and HGST's emergence as WD's flash product arm, excepting Black2 hybrid disk+flash drives.
The surviving stand-alone all-flash hardware firms include Kaminario, which makes very fast flash arrays which top SPC benchmarks ratings, Nimbus Data, Pure Storage, Skyera and Solidfire and, of course leadership-challenged Violin Memory.
Nexenta and other storage SW-only suppliers also have all-flash arrays provided through their partnerships and channels.
Troubles? Oh, there's been troubles...
Some flash firms encountered significant troubles during the year. Let's start with Fusion-io
PCIe flash HW and SW pioneer Fusion-io bought hybrid flash/disk storage array startup Nexgen in April, after having bought ID7 and its turn-PCIe-flash_into_storage-arrays technology in March. But then, in a startling turn of events co-founder and CEO David Flynn was ousted/resigned in May by incomer, and former HP and Compaq bigwig, Shane Robison who became CEO in his place. The Robison creed was to make profits and not grow the company's technological prowess willy-nilly.
Fusion-io ioDrive Duo.
He spent the year re-fashioning Fusion-io, dealing with OEM channel problems and recruiting some fresh execs. The firm now looks to have the infrastructure in place to grow, but steadily and not like a rocket because there is a whole load of PCIe flash card competition out there: Intel, LSI, Micron, Seagate/OCZ, WD/Virident, Violin Memory and others.
Violin Memory and the Sloop Don B
Violin Memory, the flash array market leader, had an IPO last September with shares priced at $9.00 and the stock price subsequently crashed to a third of that because Violin could not make profits and its deepening losses outweighed its less than stellar revenue growth. There was a widely-held view that its native SW technology lacked deduplication and other storage efficiency, management and protection capabilities.
Partnerships with software suppliers Symantec and FalconStor were not felt to be satisfactory substitutes. Things came to a head with a $34m loss in the third quarter of 2013 and the pace of events picked up; SW CTO Jonathan Goldick left in early December with CEO Don Basile terminated by the board in mid-December. Chairman Howard Bain III stepped in as interim CEO and ship-steadier, and a new CEO signing is expected shortly.
We can note too that WD's purchase of sTec was a white knight acquisition for that troubled firm. OCZ's crash came after a tortuous struggle to remain a viable business in the face of catastrophic customer incentives programs offered during a previous CEO's reign. Fortunately Toshiba stepped in to pick up the pieces and joint Toshiba/OCZ SSD product has just been announced
What of this year, 2014?
There are, obviously, too many all-flash array suppliers and some consolidation will have to take place. Skyera's technology may make it a good buy for someone. It's hard to see how Kaminario, Nimbus Data, Pure Storage, Solidfire and Violin Memory can all survive and prosper against an onslaught from Cisco, Dell, EMC, HDS, HP, and IBM. There will, we predict, be blood.
Hybrid array happiness
Troubles in the all-flash array business contrasted vividly with the general health of the hybrid flash+disk array business.
Hybrid array suppliers, well, three of them, boomed in the shape of Nimble Storage, Tegile and Tintri. Alas, poor Starboard Storage put up a For Sale sign in March and was gone by the end of the year, purchased by SGI. Hopefully it will come back strongly this year.
Nimble Storage had a triumphant IPO in December while both Tegile and Tintri reported fast growth and infrastructure build out. We may see one or both of them IPO this year or early in 2015.
All three say their ground-up designed software gives them a sustainable competitive advantage versus the traditional disk storage array vendors. These, of course, stiffed SSDs in their drive bays and flash in their controllers but had to do so with operating systems designed for disk and systems priced higher than the hybrid start-ups.
Something has to give and we may see an acquisition in this area this year, 2014, as an incumbent supplier decides a ground-up hybrid design is the way to go.
Flash storage is hot but it's not easy money for the startups. Flash foundry capacity is limited and will stay limited, foundries costing $10bn a pop. Crudely speaking though, flash is the new disk, disk is the new tape (for nearline and backup data) and tape is the archive. The flash is located in the same premises as the servers, sometimes inside the servers, and all three storage media devices can be located in public clouds.
But there isn't enough flash being made and security of chip supply is coming to the fore, as is the growing presence and influence of NAND foundry operators, like Samsung and Toshiba in the business of packaging flash chips into products. We can point to the Seagate Samsung relationship here as well as the Toshiba-Violin nexus, Toshiba's OCZ purchase, and Toshiba foundry partner SanDisk's purchase of SMART.
There may be more of this in 2014. We might also see:
- Servers adopt flash DIMMS
- A PCI flash card supplier shakeout
- One or two all-flash array startups collapsing or struggling to survive,
- Mainstream storage vendors selling boatloads of flash arrays to their customers
And we should certainly see NetApp's ground-up designed FlashRay all-flash array appliance as well as having Cisco's flash array strategy clarified and EMC's ScaleIO system become more visible and ... it will just carry on at a frenetic pace because ... Flash. Is. Fast. And. Furious. ®