How do you sex up an input-output system that will blow away data center bottle necks? Have Verne Troyer announce it at a groovy DEMO party with drink serving go-go girls!
Fusion-io, announced its new auto commit memory system, which allows a company to execute one billion input-output commands per second (called IOPS by techies) in a rack of eight servers. Previously, an I/O rate that high would require a roomful of servers, the company said. This plays into Fusion-io’s goal of shrinking the data system and making it as efficient as possible. Fusion-io chief scientist and Apple cofounder Steve Wozniak described it as, “least costly, most efficient, fastest operating.”
The company, which went public in June 2011, has grown based on its ability to deliver newer, faster ways of moving data from storage to server CPUs, as people learn more about the information at their finger tips. Fusion-io’s flash chips are more reliable than traditional solid-state disks (SSDs) and are much faster. The company released its first chip, at the DEMO conference in 2008, and released its newest iteration, the ioDrive 2 in October 2011. The ioDrive 2’s purpose is to speed up web page load time, while not overheating your servers.
For David Flynn, chief executive of Fusion-io, the company’s main purpose is to deliver useful products to the cloud computing generation, and move away from the outdated hard drive storage still used in many data centers.
The New York Stock Exchange uses Fusion-io’s technology. Not coincidentally, it is also the exchange Fusion-io stock trades on.
This new technology is called auto commit memory, which, according to Fusion-io’s blog, “is a new memory type that uses the underlying flash to present a persistent memory directly to applications.” It’s an extension to Fusion-io’s ioMemory architecture, and promises to significantly reduce latency and overhead in servers when transferring data.
For the press conference, the company set up a rack of eight HP ProLiant DL370 servers, each equipped with eight ioDrive2 Duos, and displayed a meter on-screen that they said showed the rack was executing one billion I/Ops. Fusion-io was able to execute a million in 2009. The next year it did the same with HP’s servers. The auto commit memory extension will be available in April of this year.
“The company culture starts from the beginning,” said Wozniak at the event. “One of my big jobs was recognizing what was beautiful about the various products. What’s in that rack there, last time we were here, would have filled this room.”
The company is able to achieve this by making chips that speed the process of moving data from short term to long term memory by putting data closer to the processors which will call for it first. The average processor utilization today is less than 20 percent, says Flynn. So by making memory faster, he says, Fusion-io’s technology can potentially improve server performance fivefold, by getting CPU utilization closer to 100 percent.
Wozniak, Flynn and Fusion-io chief marketing officer Rick White announced the product at the DEMO event, Enterprise Disruption: An Evening of Change and Innovation in San Francisco.
Check out the announcement video with Verne Troyer above. And don’t miss our gallery of photos of Woz and special guest Leonard Nimoy.