IBM scientist Ioannis Koltsidas’ phone rang on the afternoon of February 4. His manager was on the other line.
“Pack your bags, you and Nikolas [Ioannou] need to fly out to the IBM lab in Almaden and build a half-petabyte storage system out of Flash. It's for a demonstration at a critically important meeting on 28 February.”
“But I don’t think that’s ever been done before,” commented a nervous Ioannis.
His manager shot back, “It’s possible, right?”
Ioannis, “Theoretically, yes.”
“Then what are you waiting for?”
On the 12-hour flight from Zurich to San Francisco, the two scientists plotted out the fastest way to install and setup the two racks -- each filled with 240 terabytes of Flash provided by Texas Memory Systems (an acquisition IBM completed in October 2012), as well as 10 IBM Power 730 Express servers.
After the team unpacked and prepared the racks, Ioannis and Nikolas, together with their Almaden colleagues John Palmer, Hyojun Kim and Clement Dickey had the system working in 48 hours.
“This demonstration marks a tipping point for transactional workloads. It's the first time Flash storage has outperformed hard disks in all aspects, including capacity and performance density, and cost per Input/Output Operations Per Second (IOPS) and energy efficiency," Ioannis said.
By the numbers, the two achieved a remarkable feat: the IBM Flash System 820 achieved more than 6 million IOPS running an IBM DB2 workload on IBM Power servers.
“I think the important metrics here are capacity and performance density, as well as energy efficiency when compared to hard disks. If you consider the IOPS, what we achieved in two days with two racks would require 630 racks of hard disks. That’s roughly two years of planning and installation and they would require a room larger than the size of two basketball courts."
“In terms of energy our system runs on 19 kilowatts compared to 4.5 megawatts with high capacity hard disks, a 236 fold improvement," Nikolas said.
The demo left a big impression on IBM’s important guests, which is good news for banking, online markets, business analytics, media streaming, modeling and cloud computing clients due to its unique combination of performance and efficiency.
