Fusion-io Adds Supermicro as Partner, Expands With IBM
Having conquered the big three of enterprise server vendors, IBM, Hewlett-Packard, and Dell Fusion-io, the company that uses flash memory to turn ordinary servers into super-fast storage systems has landed number four. Supermicro, the white-box manufacturer that sells its systems primarily through distributors and resellers says it has started offering Fusion-io technology in its server lineup.
At the CeBIT trade show taking place in Germany today, the pair demonstrated a server using a new Fuion-io product, the ioDrive Octal, in a Supermicro server that can do 1.4 million random input-output operations per second.
Why team up with Supermicro? For its reach. Bob Emmett, manager of OEM sales at Fusion-io told me Supermicro has relationships with scores of resellers, private-label hardware vendors and other companies that reach places that IBM, Dell and HP don’t. “They’re the anonymous supplier in a lot of cases, because the product isn’t being sold directly under the Supermicro name,” he said.
One example: When the Lawrence Livermore National Lab announced it was using Fusion-io memory to build the world’s highest-performance storage array no one mentioned that the gear in question was actually manufactured by Supermicro, and resold by another company, Appro. It’s not a huge business, but it is a growing one: Supermicro reported $721 million in sales last year, up from $506 million the year before.
And if that weren’t enough good news for the fast-moving startup that last month sought to pick a fight with storage giant EMC over who sold the most flash memory to enterprise customers in 2010, there’s more from IBM. Big Blue has doubled the number of Fusion-io based adapters for use in its System x servers it sells with with Fusion-io tech inside it, from four to eight, and boosted the number of IBM systems that can work with Fusion-io adapters to 12. Not bad when you consider that both IDC and Gartner said last week that IBM finished 2010 as the world’s leading sever vendor by revenue.
As I’ve noted before, flash memory inside a server, is by itself not big deal. Fusion-io’s difference is in how it uses flash to put actively used data close to the microprocessor, which in conventional systems spends as much as 80 percent of its time waiting around for other parts of the computer to get data to it that it can work out. That’s a lot of wasted computing cycles, that when added all up amount to a lot of money spent. Get the data closer to the chip and you eliminate a lot of that wasted time. Always eager to squeeze more productivity out of the gear they guy, CIOs are naturally lining up to buy Fusion-io-based servers or use it in other gear, such as a trading system at Credit Suisse.
No surprised it has a lot of interest among venture capitalists. Last April the Salt Lake City-based outfit landed a $45 million in series C round led by Meritech Capital Partners, with Accel Partners, Andreessen Horowitz, New Enterprise Associates, and Triangle Peak Partners also investing.