NAND flash explained: SLC vs. MLC, part 2Date: Apr 17, 2012
In SLC vs. MLC, part 2, a segment of his presentation entitled “Making the case for solid-state storage” from Storage Decisions, Dennis Martin, president of Demartek, discusses SSD error correction, energy consumption compared with traditional disk drives, and SSD cost per IOPS. During the presentation, Martin said that error checking and correction technology can resolve reliability problems that can occur in solid-state storage, particularly in devices that rely on multi-layer cell technology, which pushes more cells into smaller space.
“The problem is that as you get smaller, your endurance goes down, and it just becomes more difficult. What you have to do is compensate for that by increasing error-correction code (ECC) in these devices so you can compensate for the lower life by having more ECC to correct for it,” said Martin.
Another issue that can be corrected with ECC is read disturb, he said.
“So if you read over the same bits a lot, I mean, really a lot, because of the physics of the way NAND works, sometimes this read that is happening over here will flip the guy next to him. That’s why you have ECC, right?” said Martin.
He also compares the energy cost of running SSDs versus traditional hard disks, noting that when SSDs are deployed in a data center, those costs would be reduced. And while hard drives continue using power when idle, SSDs drastically reduce that drain.
“You’re going to see when it’s idle, the flash is consuming no power at all, it turns out. There’s an FPGA on there that keeps things running, so there’s a little power consumption there. But idle SSDs are nearly nothing [for] power,” said Martin.
Martin said that SSDs do come up short in cost-per-gigabyte when compared to hard drives, making the traditional storage devices more attractive for buyers on a budget. But he said in his presentation that when you compare costs of using SSDs and the cost per IOPS for those units, SSDs are cheaper.