Toshiba teams up with Violin for an enterprise storage revolution
'Leading the MLC charge'
Comment Toshiba has completed a funding round that will see it investing $20m in Violin Memory and bringing its multi-level cell (MLC) flash technology to the table.
Violin recently gained a new CEO, Don Basile, and wants to replace the hard drive storage of tier one enterprise data with flash array storage, teaming up with Toshiba and its MLC tech. Basile says: "Toshiba is leading the MLC charge. [There is] the possibility of really driving advanced MLC technology into spindle-sprawl enterprises in the next three to five years."
Violin's main product is its externally-attached 1010 Memory Appliance, offering up to 5TB of capacity. A series of coming announcements will extend its connectivity options, such that they will encompass Fibre Channel, FCoE, 10GigE and iSCSI for external attach and PCIe, SATA and SAS for local attach. Basile wants to move it, he says "into a primary storage tier replacement for multiple hard disk drive (HDD) arrays". He says Violin customers are already buying it with "hundreds of terabytes being deployed at a time".
The primary competition he sees comes from Oracle and its Exadata 2 product: "[It's] the most feature-rich application of flash. We think Larry's got that really on the right path, but only part of the data is in flash; the majority is still on rotating media." Violin wants to put it all on a combined DRAM cache and flash array.
In Basile's view Violin will take HDD array aggregation technology, like RAID, hot spares, mirroring and so forth that you get with high-end arrays from EMC, IBM and NetApp, "and build it into silicon," Violin's switch. Of course the Violin controller, aka switch, does the ECC protection, wear-levelling and so forth that all SSD controllers do "but it's the aggregation technologies that make it really exciting".
He says that its this aggregation technology that enables Violin to have sustained I/O speeds and not just peaks of speed: "That's where array aggregation technology is different from memory component technology. Data centres want predictability from their storage devices." He says EMC and Oracle are partially there with their enterprise flash storage, "but Violin does it at lower cost, in less space and at higher speed".
Violin is moving to couple the 1010 with software, like the already-announced FalconStore deal, and so have an appliance model. We will probably see more, with Violin providing the acceleration for software doing block or file storage, and should expect announcements relatively soon around database, data warehousing and email combinations of Violin's appliance and packaged software.
Turning to the 1010's capacity, Basile said: "We will likely announce 10TB and 20TB models with shipment by June." A 2010 goal is to achieve 40TB in a single appliance and perhaps 80TB in a 3U enclosure with roughly four to six times current 1010 appliance speeds, meaning one to two million IOPS per appliance. He said: "You need massive scale in primary storage [and] that's where Violin is driving."
What about pricing? Basile wants pricing to be on par with enterprise HDD storage array pricing to facilitate plug-replacement, and suggests Violin could achieve per-GB prices below enterprise arrays from EMC, IBM and NetApp - think V-Max and DS8000. "With the Toshiba strategic supply relationship we can out-perform enterprise HDD arrays and undercut their prices ... We expect to use Toshiba MLC technology quite soon. We will ship this technology this calendar year."
Basile reckons that the reason around 90 per cent of enterprise data centre applications haven't been virtualised yet is because the back end storage performance lags behind the compute and network performance available from relatively cheap, multi-core Nehalem-type processors and 8Gbit FC/10GigE networking with its low latency and high bandwidth. He says: "We think our memory array technology is the third leg of this triangle."
What do we think? Price, performance and predictability are key. Obviously no CIO is going to ignore a plug-in, equivalently or lower-priced flash array that drastically outperforms existing Symmetrix, DS8000 or FAS 6000 arrays with clockwork consistent predictability and equivalent array functionality plus integration with VMware and other hypervisors. Can Violin deliver? It's got $20m of funding, a new CEO and other senior execs driving the company forward, plus an existing reputable product it can develop.
But it's just a 40-person start-up out to take data centre enterprise tier one data storage into the leading edge. If it can deliver on the product, get its aggregation ideas sorted and implemented, and get the pricing right, then it will find open doors and cause all sorts of grief in Hopkinton, Sunnyvale and other storage array HQ buildings. If not then it's goodbye - Toshiba will look around for another partner. ®