Flash in the pan not the end of HDD
There's plenty of life left in the HDD market despite the ongoing noise about flash, says Gavin McLaughlin
Some are saying that 2013 will be the year the market says RIP to HDD. According to current hype, flash storage will have taken its place, resolving all customer data storage issues, saving money, time and power. But to misquote Mark Twain, the report of the demise of HDD has been greatly exaggerated.
Flash is a great tool for some purposes but it has serious limitations, and it definitely shouldn't be seen as the answer to every storage question.
Some people claim HDD is a homogeneous technology. However, the truth is there are differences in terms of performance, reliability, data integrity and overall ruggedness. Treating all HDDs the same, however, means they can be used incorrectly and explains how and why drives might fail or seem to fail.
Some people claim HDD is unreliable. The 'pull' rates of drives in the field, which depend on drive type, service load, environmental conditions and storage software, might suggest they are failing at quite a high rate.
Typically, manufacturers aim for one million hours as mean time between failure (MBTF). Mission-critical enterprise HDD is rated at full-speed operation for 24 hours a day, where the heads are moving constantly. They generate the most heat and have the most wear, depending on the environment.
Nearline or "business critical" HDD can be used 24 hours a day at 30 per cent without undue wear and an increased failure rate – again, assuming the right environment.
PC-grade HDD is what people think of when they talk about SATA drives. They're the cheap drives at an electronics store that make your customers wonder why they cost so much more through the channel.
Enterprise drives are built for performance and high utilisation; nearline drives are meant for back-up and archive; and desktop drives are for PCs or external backup drives at home or in a small office. If you install the wrong ones for the wrong purpose, the failure rate is going to rise.
Hard drives fail partly because of heat and vibration. And this can expose the business to data loss, where there's no redundancy or the time to restore redundancy.
Failure rates in this case go well past the design specifications, and I believe they're untenable for most businesses, although many storage vendors try various sticking plaster solutions such as fancier algorithms or making more copies of data, increasing the cost of the service to the user.
Some people claim RAID-6 is the answer to everything. I believe in the right RAID level for the right uncorrectable error rate (UER) of the drive or drives involved in the data protection scheme.
Combinations of RAID-1, RAID-5, or even RAID-6 can be employed, but all have costs which must be factored in. If you can prevent disk failure with an elegant solution, you shouldn't have to fork out to cover RAID-6. Far too many tenders and RFPs are demanding RAID-6 – probably unnecessarily.
So flash is still being pushed and pushed by vendors. Buying into this could prove an expensive mistake. An all-flash array will be overkill for most businesses, which don't need extreme I/O performance.
For most, hybrid storage with real-time tiering for acceleration will be a more suitable option.
HDD is not old school, and it has been around for more than 50 years for a reason. HDD is still the most cost-effective storage mechanism for most data workloads. For workloads that do require accelerated storage, intelligent hybrid arrays are still more reliable and use less power than flash arrays.
Gavin McLaughlin is solutions development director at X-IO