Sign in with
Sign up | Sign in

TCO: Buzzword No Longer

The Other Specs: An SSD Is More Than Its Throughput
By
Brought to you by What's this

It should be obvious by now that client and enterprise drives are completely different animals. And while consumers may be concerned about ROI and TCO in an abstract sense, for IT departments, they’re very real, and very important, especially with regard to performance, power, and endurance.

Every now and then, we’ll see people make a case for using client drives for enterprise purposes. They’ll point to consumer SSDs’ speeds and feeds and use that as the justification for giving these drives business-class seating, so to speak. Such SSDs are fast, sure, but they’re not really up to the task of true, heavy-duty enterprise workloads. And when intense data center workloads have to be sustained over several hours, the difference is even more glaring.

“Performance for SSDs is so much about IOPS, but it’s incomplete when you talk about IOPS without associating response time,” says Sameer Siraj, enterprise product line manager at Seagate. “Think about a car that can go 150 MPH around the race track and another car that can go 100 MPH, but the 150 MPH car has to stop every three laps for a pit stop. Is it really fast over the course of 15 laps? Probably not, because the other guy is maintaining 100 MPH without any stops. So if I’m running at 100,000 IOPS, but my read/response time goes up and down, I’m not getting the benefit of really high performance because the user has to wait. Well-managed drives can have high performance with extremely low response times, meaning that when the user sends a command or requests data, he can get it really fast. Guys who don’t manage the media well can claim fast performance, but when drive loads get really high, the response time craters.”

The relationship between performance and ROI is not something you can adequately measure by running Iometer or CrystalDiskMark for 10 minutes. This quick sampling of performance might paint a good picture of performance in a client system, but it’s useless for gauging true enterprise ROI. If you want to measure an SSD’s performance with a typical enterprise workload, running SPC is a much better approach. This test should cast those purportedly fleet-footed consumer SSDs in a new light. An enterprise drive can do more work (measured in IOPS) over the long haul, as in non-stop days or weeks, than a client drive (and in some cases, multiple client drives).

An SSD’s endurance and life span also contribute to its TCO. We don’t want to give the idea that enterprise drives are immortal and will continue working well after the walls of your data center start to crumble; all drives will eventually wear out. However, before you make a substantial investment in an SSD, find out how much work the drive will do before it requires replacing.

Generally speaking, you can count on an enterprise SSD to do 2X to 3X more work of an average consumer SSD of the course of their lives. The biggest reason for this is inside the drives, as enterprise SSDs use more robust components and typically manage data better, both of which lead to longer NAND endurance. Just as important, high-quality components should also reduce the likelihood of premature drive failure. This is a dreaded double whammy: In addition to the added cost of replacing the failed drive, an IT staff has to suddenly deal with crippling downtime, which can easily exceed the cost of the drive itself. This is part of the formula you have to use when adding up TCO. Enterprise drives will cost more at the cash register, but they’ll pay for themselves on the back end by lasting longer, doing more work, and requiring less support.

Power consumption figures into to TCO, as well. It’s tempting to look at a client SSD’s reported power consumption (often 2W to 3W or less) and think that it’s far more efficient than an enterprise SSD. For example, Seagate’s Pulsar SSD is rated at 9W, which is what you’d expect from a hard drive. Now, what seems lame at first glance is anything but when you measure the Pulsar’s consumption based on IOPS per watt over time. The extra power the Pulsar pulls keeps a greater number of NAND chips active, so you get a steady stream of data with not bottlenecks.

And don’t forget the issue of total work done. When it takes two or three client drives to perform the same amount of work as a single Pulsar SSD over a month of 24x7 operation, for example, then you can understand how an enterprise drive that uses 9W is actually more efficient than client drives that seem to consume less power. It all goes back to TCO.

Buying an enterprise SSD, whether it’s 10 or 1,000, shouldn’t be a snap decision. Before you invest thousands of dollars into upgrading your storage, be sure to find out exactly what you’re buying. There are too many important features in an SSD that don’t jump off the spec sheet. Interoperability and TCO are a couple, and they absolutely have to be a part of your overall decision. A quality enterprise SSD should deliver on both fronts.