There's More To Enterprise SSDs Than Just Speed

Solid-state storage is becoming an increasingly attractive option for enterprises in the ongoing struggle to provide more efficient storage but even as the economics of the technology improves, executives still grapple with justifying its expenditure, reportedly. On the whole, nothing much has changed in the last few years on that front. The big problems, as it pertains to storage, continues to be a case of too much data and not enough money.

“If you can use solid state for some portion of your I/O, then perhaps that allows you to make better economic use of the rest of your system,” says Mark Peters, senior analyst, Enterprise Strategy Group in Milford, Mass. “Solid-state is all about economics in a good sense; there are two ways to look at it that makes economic sense and that’s exactly why (companies) are beginning to adopt it.”

One of the advantages of solid-state is that it’s very measurable and precise and its life can be determined. It’s been suggested solid-state storage can last for up to 40 years. While that’s unproven, there is something to be said for the longevity and reliability of solid-state.

“One of the nice things about solid-state, which can be RAIDed, is you know how long each individual cell will last. The lifespan of any solid-state is determined by how many times you write to a particular cell,” Peters explains. “In heavy-duty enterprise environments, you’ll now find more manufacturers talking about doing full writes (using every cell on a solid state drive 10 times a day for three to five years) so they’re getting quite precise on the expected lifespan of the product and that helps tremendously.”

Jim Bagley, senior analyst, Storage Strategies NOW in North Carolina, says solid-state has been making serious inroads in enterprises in recent months. “The uptake into the enterprise for solid-state has been double what we forecast a year ago. Year-on-year, the enterprise implementation has doubled, and that was a big surprise as we were forecasting about a 50% to 55% compound growth rate,” he says. “It’s probably going to keep up at that rate for another year, and there’s a couple of things driving that, including the virtualization boom which creates I/O bottlenecks, and the standard rotating media in the shared networks can’t keep up with that demand.”

Organizations that are purchasing solid-state storage are doing so typically to solve speed problems, remarks Jim Handy, analyst, Objective Analysis in Los Gatos, Calif. “Storage needs are growing like crazy and backs are up against a wall and the decision is made to try (solid-state),” he says. “That’s where SSDs are getting a lot of favor. It gets its speed because it has no moving parts. Some people try to imply hard disk drives (HDDs) are not reliable because they’ve got moving parts, but they’re actually extraordinarily reliable.”

Where an SSD provides real power savings, Handy continues, is when it’s used to replace an array of disks. “That solves heat problems and it improves reliability for when you have less things (to manage) then you have less things that are going to fail on you.”

Moore's Law is definitely in effect when it comes to solid-state storage, and as the technology and market mature, the storage world is moving toward a place where SSDs are not just silicon disks but full solid-state systems, says Kurt Marko, a regular contributor to Network Computing and InformationWeek, as well as the author of a recent storage report. SSDs are replacing disks in high-throughput, high-IOPS systems, which are already very expensive, says Marko. To get the throughput the applications need, enterprises typically need 10 to 12 SSDs in parallel, which provides an order or two of magnitude in improved performance. The price/performance is driving that adoption, he says.Data from the report, 'Research: State of Storage 2012', finds the vast majority of enterprises currently using or planning to use SSDs are using them for databases (61% of the 166 respondents who noted they are using or evaluating SSDs), followed by usage for improving overall server performance (57%) and automated tiered storage (34%). Other notable reasons included technical applications (29%), reducing power consumption (27%), video/multimedia editing (21%) and other transaction-heavy software (26%).

On the subject of server consolidation, Handy says some companies are taking databases and spreading them out over a number of servers. This practice is known as “sharding” (the art of splitting a big database into a number of smaller sets to allow it to be processed in parallel). An SSD reduces the need to shard. Despite this, some hesitation remains, he says.

“Solid-state is still a new technology, and there’s still an uncertainty as to whether or not it will cause problems down the line from the executives’ perspective,” he adds. “There is a lot of places that haven’t adopted SSDs as there are always concerns about adopting a new technology.”

But that could change if considered from a cost-per-Gigabyte basis. The cost of solid-state is dropping.

“There’s a lot of flash memory coming into the market and that’s driven the cost-per-gigabyte down,” Bagley says. “The flash memories are approaching what the real expensive, high-end Fibre Channel 15,000 RPM drives cost on a cost-per-gigabyte basis.

“Part of that has been driven by the shortages of hard drives because of the flooding in Thailand. These things have come together to kind of make a perfect storm for flash.”

Learn more about Research: State of Storage 2012 by subscribing to Network Computing Pro Reports (free, registration required).