Storage suddenly seems to have got interesting again. As the interest moves from increasing spin speeds and applying ever more intelligent means to get a disk head over the right part of a disk in the fastest possible time to flash-based systems where completely different approaches can be taken, a feeding frenzy seems to be underway. The big vendors are in high-acquisition mode, while the new kids on the block are mixing things up and keeping the incumbents on their toes.
After the acquisitions of Texas Memory Systems (TMS) by IBM, Whiptail by Cisco and XtremIO by EMC in 2012, it may have looked like it was time for a period of calm reflection and the full integration of what they had acquired. However, EMC acquired ScaleIO and then super-stealth server-side flash company DSSD to help it create a more nuanced storage portfolio capable of dealing with multiple different workloads on the same basic storage architecture.
Pure Storage suddenly popped up and signed a cross-licensing and patent agreement with IBM, with Pure acquiring over 100 storage and related patents from IBM, with Pure stating that this was a defensive move to protect itself from any patent trolling by other companies (or shell companies). However, it is also likely that IBM will gain some technology benefits from the cross-licensing deal. At the same time as the IBM deal, Pure also acquired other patents to bolster its position.
SanDisk acquired Fusion-io, another server-side flash pioneer. More of a strange acquisition, this one – Fusion-io would have been more of a fit for a storage array vendor looking to extend its reach into converged fabric through PCIe storage cards. SanDisk will now have to forge much stronger links with motherboard vendors – or start to manufacture its own motherboards – to make this acquisition work well. However, Western Digital had also been picking up flash vendors, such as Virident (itself a PCIe flash vendor), sTec and VeloBit; Seagate acquired the SSD and PCIe parts of Avago – maybe SanDisk wanted to be seen to be doing something.
Then we have Nutanix: a company that started off as marketing itself as a scale-out storage company but was actually far more of a converged infrastructure player. It has just inked a global deal with Dell, where Dell will license Nutanix’ web-scale software to run on Dell’s own converged architecture systems. This deal gives a massive boost to Nutanix: it gains access to the louder voice and greater reach of Dell, while still maintaining its independence in the market.
Violin Memory has not been sitting still either. A company that has always had excellent technology based on moving away from the concept of the physical disk drive, it uses a PCI-X in-line memory module approach (which it calls VIMMs) to provide all-flash based storage arrays. However, it did suffer from being a company with great hardware but little in the way of intelligent software.
After its IPO, it found that it needed a mass change in management staff, and under a new board and other senior management, Quocirca is seeing some massive changes in its approach to its product portfolio. Firstly, Violin brought the Windows Flash Array (WFA) to market – far more of an appliance than a storage array. Now, it has launched its Concerto storage management software as part of its 7000 all flash array. Those who have already bought the 6000 array can choose to upgrade to a Concerto-managed system in-situ.
Violin has, however, decided that PCIe storage is not for it – it has sold off that part of its business to SK Hynix.
The last few months have been hectic in the storage space. For buyers, it is a dangerous time – it is all too easy to find yourself with high-cost systems that are either superseded and unsupported all too quickly or where the original vendor is acquired or goes bust leaving you with a dead-end system. There will also be continued evolution of systems to eke out those extra bits of performance, and a buyer now may not be able to deal with these changes through abstracting everything through a software defined storage (SDS) layer.
However, flash storage is here to stay. At the moment, it is tempting to choose flash systems for specific workloads where you know that you will be replacing the systems within a relatively short period of time anyway. This is likely to be mission critical latency-dependent workloads where the next round of investment in the next generation of low-latency, high-performance storage can be made within 12-18 months. Server-side storage systems using PCIe cards should be regarded as highly niche for the moment: it will be interesting to see what EMC does with DSSD and what Western Digital and SanDisk do with their acquisitions, but for the moment, the lack of true abstraction of PCIe (apart from via software from the likes of PernixData).
For general storage, the main storage vendors will continue to move from all spinning to hybrid and then to all flash arrays over time – it is probably best to just follow the crowd here for the moment.