Hard disk capacity is cheap but slow. Flash drives are fast but costly - and we don't make enough flash to replace hard drive capacity. Hybrids - flash + disk - is industry's answer. Will it be yours?
Even with recent price drops in anticipation of Toshiba's new Fab 5 coming on line, the cheapest flash GB costs 30x the largest disks. And for large reads and writes, disk performance is competitive.
So disks aren't going away.
But the performance benefits of flash SSDs for other workloads are stunning. The entire industry, from drive vendors to the largest array vendors, are racing to find the sweet spots where performance is maximized at minimum cost.
One of those spots is the hard drive.
Hybrid tea leaves Seagate already has hybrid drives with their Momentus XT. I used them in notebooks and found they provided more performance, but the stingy 4GB of flash limited their value.
But industry rumors claim that Seagate is getting ready to roll out enterprise hybrids: 10k drives with 8 to 16 GB SLC flash caches and better optimized data placement software. The sales pitch: "get SSD-like performance for little more than a 15k drive price."
Toshiba is also looking to enter the market. Commenting recently Hironobu Nishikori, VP of Toshiba's Semiconductor & Storage Products said,
. . . optimization of data placement is what we are struggling with most now. Especially, it is now a great challenge to realize a high storage capability while considering security and power saving issues. . . .it is a great advantage to have both HDD and NAND flash memory businesses.
Toshiba invented flash and then doggedly invested in the technology for 20 years before its "overnight" success in the last decade. They can do as much as Seagate to exploit the hybrid architecture. And if they want to grow their shrinking disk market share, they need to.
Will it fly? What isn't clear is market acceptance. In recent research by my firm, only 7.5% of respondents planned to migrate their applications to SSDs. And, of course, even fewer consumers are using them.
Which means the market is still wide open for a hybrid solution. What will a successful hybrid look like?
- Performance. The hybrid needs to be within 20% an SSD in everyday performance. That means the flash cache has to be large enough and smart enough to identify and store 80% of what the average user accesses everyday. 4GB is too small, 8GB will satisfy casual users, and power users will need 16-32GB. Linux servers will want 8-16GB; Windows servers 16-32GB.
- Price. Flash is expensive and so is the integration work. Vendors need to make a profit while consumers need to see value. The price uptick is easiest to justify at higher price points, so stick large and/or fast drives.
- Cost+ pricing. The temptation is to look at SSD pricing and come in at, say, half that. Don't do it. Disks are the default option for most consumers and enterprises - and you want to keep it that way. Take some extra margin, but don't go crazy. This is a new segment and you're in it for the long haul.
- Reliability. Use MLC for consumers - they don't care. But enterprise buyers are wary of endurance problems so put SLC in those. Use the flash as a cache, not as a tier, for data protection, so the drive keeps working even if flash fails. And keep your standard warranty.
- Software. Ship with a standard set of caching rules, but add a tuning utility. It will be a selling point and gives sysadmins extra motivation to understand what your shiny new hybrid can do.
The Storage Bits take Flash SSDs are great - if you can afford them. Hybrids are the Next Big Thing in storage devices because they can give some of the key benefits of SSDs - such as fast boots and app starts - without sacrificing the capacity that our music, photos and videos gobble up.
The disk drive is far from dead. Hybrid drives will keep them spinning for decades to come.
Comments welcome, of course. I expect to see announcements starting next month. No NDAs were violated in the writing of this post.