Mechanical hard drives with spinning disks are doomed to extinction, thanks to solid state flash drives that are becoming cheaper and offering greater capacity by the month. At least that’s how some in the data storage industry see it.
Outwardly, there’s a convincing logic to this argument, especially when you consider what’s happened in other markets where devices with moving parts faced competition from solid state electronics. Televisions, telephony and radio equipment, clocks, automobile ignition … the list is endless, and in every case it’s ended up with the same result: solid state electronic devices have won because they are cheaper to make, more reliable, and offer similar or (usually) superior performance.
So when it comes to storage planning, it’s sensible to at least consider when flash-based solid state drives (SSDs) might take over from conventional hard disk drives (HDDs). Right now, SSDs are significantly more expensive per gigabyte than HDDs, and while they offer very fast read speeds, they suffer from slower write speeds, and from the limited number of times flash cells can be written to before they wear out.
But flash memory prices are falling rapidly, perhaps by 50 percent to 60 percent a year, and SSD technology is also improving, so write speeds are likely to increase and memory wear-out is likely to become less of a problem. For example, companies such as California-based SandForce promise technology innovations that will ensure flash cells effectively last 80 times longer than is common now, with write speeds far closer to levels achievable for reads.
As prices drop and the capacity and performance of SSDs improves, it’s likely that first a few, and then an increasing number of HDDs of different types will be replaced by their solid state siblings. But the complete extinction of HDDs is unlikely for many years, if ever, for reasons we’ll get to in a moment.
Fibre Channel Could Go First
So what type of HDDs are likely to be replaced first? David Vellante, a former IDC analyst and founder of the Wikibon project, believes that the first to go will be high-performance Fibre Channel (FC) drives, which are usually bought for their high performance and low access times. He argues that since flash memory prices are falling much faster than HDD prices, the price differential between SSDs and FC HDDs — which is currently 15 times greater for SSDs — will drop to a multiple of just three in less than three years, and possibly considerably less than that. At that price, SSDs with their faster read speeds will make the competing FC HDDs obsolete, he believes.
Mark Peters, an analyst at Enterprise Strategy Group, agrees with at least some of this assessment. “I’d say that Fibre Channel drives are first in the firing line,” said Peters. “In general, SSDs will be more attractive than FC drives if they are not too much more expensive.”
But Peters believes that SSDs will have an impact in the FC drive market much sooner than Vellante anticipates. “I think SSD sales will take off next year,” he said.
That’s because some IT departments will be willing to pay a significant premium over FC disk prices for SSDs that offer higher performance. This should not be too much of a surprise — users have always paid more for disk storage than tape, and FC drives rather than lower-performance drives. For applications that require the highest possible I/O performance, why shouldn’t they pay more for SSDs?
But Peters warns against looking at storage media such as FC drives and SSDs solely on a price per gigabyte basis. “If you have a 500GB FC disk and you are only using 200GB, then what is the price per gigabyte? Your effective price per gigabyte is more than twice as high,” he said.
In any case, price per gigabyte is often not the relevant metric to be looking at when considering switching to SSDs.
“Companies should also be looking at price per I/O, or price per millisecond of access time, or cost per unit of power a drive consumes, depending on their circumstances,” Peters said. And that means that you end up with something like hard drives for capacity, and SSDs for I/O.
Of course, it won’t all be plain sailing for SSDs. There are two sides to I/O: reading and writing. While SSDs have a clear advantage when it comes to read speeds, what about the write side of things? And let’s not forget about the limited life of SSD memory cells.
“I think these problems are overblown,” said Peters. “We’ll overcome poor write speeds with techniques like caching using DRAM, and will be able to get around wear problems with techniques like wear leveling and over provisioning.” In any case, conventional hard drives often fail in a matter of months, and those that don’t are typically replaced every two or three years anyway.
SSD’s Obstacles: Tiering, Inertia
There are other problems to consider if you are adopting a “hard drives for capacity, SSDs for I/O” strategy. How, for example, do you make sure the right data is on the right medium? To get the full performance benefits of SSDs and the cheap storage benefits of spinning disks, you may end up needing a whole new software layer to help move data around in a tiering approach, Peters warns. But there are companies like Compellent (NYSE: CML) and Sun (NASDAQ: JAVA) working on the problem.
Another thing that may slow the advance of SSDs is the fact that there is an enormous installed base of FC drives around the world. The “stickiness” of any given technology shouldn’t be underestimated when there is lots of it about. Look at tape storage — it is still a multi-billion dollar business, and holding its own too. IT departments are rightly cautious when it comes to making changes and abandoning investments, so SSDs may be adopted far slower than the economic case dictates.
One thing is pretty much certain, however: delays won’t be caused by vendors dragging their feet.
“All the major vendors are looking at SSD technology — it’s not just a couple of them that are interested in pushing it,” said Peters.
Indeed, business from the likes of EMC (NYSE: EMC), Sun, IBM (NYSE: IBM) and HP (NYSE: HPQ) have made STEC (NASDAQ: STEC) the early winner in the SSD sweepstakes. Its shares were up 31 percent in a single day earlier this week after the company’s growth rates exceeded Wall Street estimates once again — and in the worst recession since the Eisenhower Administration at that.
Going forward, STEC will face greater competition from newer entrants like SandForce, Intel (NASDAQ: INTC), where the technology has attracted the attention of co-founder Gordon Moore, and Fusion-IO — the startup that managed to lure Apple (NASDAQ: AAPL) co-founder Steve Wozniak out of retirement.
With all that excitement, perhaps analysts’ claims are not that far-fetched. While conventional hard disk drives may not be obsolete in the foreseeable future, it appears certain that many of them will have been replaced by SSDs by the time 2012 rolls around.
Follow Enterprise Storage Forum on Twitter