SSD Caching: Enterprise Philosophies Drive Desktop Performance
Last year, at Intel’s Developer Forum, I talked to a company called GridIron Systems. It was showing off an appliance called the TurboCharger, which uses a ton of solid-state storage to help accelerate Oracle databases. The 30 TB warehouse it had on the show floor consisted of eight appliances and 36 hard drives in a single cabinet pushing 12 GB/s of sustained throughput and more than 1 million IOPS. Achieving those numbers previously required 3000 short-stroked hard drives. GridIron would likely be happy to go into depth on the hardware/software combination it uses to achieve those impressive figures. But at the appliance’s core, it’s essentially caching the most-used data in flash to keep it as close to system memory (smallest, fastest storage tier) as possible.
You can, of course, apply the same technology to less enterprise-oriented environments, though it loses some of its allure as you shift from SMB-class workloads to the desktop. Caching is super attractive to businesses because it doesn’t require big infrastructure modifications. You can put SSDs in front of a large array using an intelligent controller and the benefits get folded in automatically.
That’s not as big of a concern in the client space, where most folks can decide whether to install an app on an SSD or a conventional hard drive, depending on its performance profile. Nevertheless, Intel is enabling SSD caching on the enthusiast-oriented Z68 Express chipset, giving power users one more option for storage.
Intel reckons that, while an SSD boot drive facilitates great performance, data not on the solid-state storage sees no benefit. Caching done right ensures anything you access frequently enjoys a performance improvement.
Cache: What Is It Good For?
I’ve been playing with SSD caching for a while now. A couple of weeks ago, I asked some of the enthusiasts who like to talk tech on Twitter if they’d rather use a flash-based boot drive and hard disk or a hard drive accelerated by SSD caching. The unanimous vote was in favor of the separate boot drive. Enthusiasts like to have control, it seems, especially when that control yields superior performance.
Boot drives can get annoying, though. I consider an 80 GB Intel X25-M the absolute minimum-capacity SSD for a primary drive, and that’s a $175 investment. I’ve been there, run out of space, and moved on. My production workstation centers on a 160 GB drive—a $400 proposition. If your budget for storage is $200, you’re looking at either a 2 TB Barracuda XT or an 80 GB SSD, not both.
The other way to go would be a 40 GB Intel X25-V for $100 and a 1.5 TB Barracuda for $70. But there’s no way I’d try to manage a 40 GB SSD as a boot drive. Windows 7 on its own eats of 14 GiB of the user-accessible 37.1 GiB. After getting drivers and PCMark Vantage installed, I was already down to 18.6 GiB. Throw Office and a game on there, or any other combination of frequently-used apps, and you’re pretty much tapped! Even then, everything installed to the hard drive is subject to the performance constraints of magnetic media.
Caching makes it possible to install everything on the 1.5 TB disk, and then transparently throw in the SSD to accelerate recently-read data. A 40 GB drive is ample for that task. As a result, the technology will actually appeal to the contingent of enthusiasts who don’t have enough cash for a large-enough SSD and user storage, but can still afford to get a foot in the door with a nice big hard drive and an affordably-small SSD.