SSDs are unquestionably faster and more energy efficient than traditional hard drives. But are they more trustworthy?
Conventional thinking holds that that absence of moving parts makes SSDs more resistant to failure. However, recent research is starting to question that view, as well as the technology’s reputation as being easier to work with and manage.
Data recovery firm Kroll Ontrack recently surveyed more than 500 enterprises and found that more than half had experienced data loss with SSDs, although the vast majority reported minimal failure. And while root causes of the failures varied from device malfunction and human error to broader system area corruption, nearly a quarter of the respondents said data recovery from SSDs was “nearly impossible.”
Meanwhile, Tom’s Hardware Guide recently completed a detailed examination of SSD vs. HDD operation, concluding that SSDs do, in fact, significantly improve performance. However, they did not show themselves to be any more reliable even if they were said to have high read/write endurance. Also note that since SSDs generally replace multiple HDDs in the typical enterprise, failure rates will diminish but the consequences of those failures will likely be greater.
You also need to consider the fact that SSD failures can produce a cascade effect in certain RAID configurations, according to SAP specialist John Appleby. When his firm, Bluefin Solutions, implemented a RAID5 array to support a growing VMware environment, they quickly realized that a single drive failure can overload the remaining drives before the first one can be replaced. Write endurance is generally lower in RAID5, a situation that is compounded by the practice of putting database logs and data on the same disk group, something that isn’t mentioned in the product brochures.
And sometimes, failures occur not in the drive itself, but in related systems. Firmware, for example, can make or break a drive, as Intel learned this year with its SSD 320 drive. The company recently released a fix for a bug that first caused the drive to crash and then prevented a reboot. Up to now, frustrated users had to use the SSD Toolbox to perform a secure erase, although at the expense of losing stored data.
It shouldn’t come as any surprise that SSDs are not immune to the immutable laws of physics, particularly that pesky one about matter moving from order to disorder. Solid state has a vital role to play in the enterprise, but it still needs to be handled in a manner that complements existing data center infrastructure.