Flash storage is commonly viewed as a premium technology only practical for special situations where very high speed data read/write access is required. The main reason for this is that the upfront cost of flash per Gbyte is still above that of disk storage. However, flash is no more a special kind of disk than disk is a special kind of tape. It is a totally different, and a price comparison based solely on purchase price per GB, and the mindset behind such a comparison, misses those differences.
One major difference between flash and disk is in the way it fails over time. First, in most situations flash lasts longer than disk in real time, so while it is more expensive up front, it can make at least part of that up in longevity. And that difference actually increases in less intensive storage situations, where data is not constantly being accessed and rewritten, where flash is never considered because of the purchase premium.
The reason for this is that disk drives tend to fail catastrophically, with little warning, usually when a bearing breaks. One minute all the data is there, the next it is inaccessible. This is a major reason for data backups on multiple disks and for the popularity of RAID. This, however, effectively multiplies the price of disk storage. For instance, if you want to store 20 GB of data safely, you really need at least 40 GB of storage divided evenly on two disks, which also multiplies the number of controllers, disk racks, etc. This also means that the most important measure of disk life is hours of operation, whether the disk is actually being used all that time or simply spinning idly through the night. Since disk farms are seldom shut down, even when not in use, this leads both to waste of power and excess wear, both of which effectively increase the cost of disk.
Flash, in contrast, typically fails sector by sector. The relevant measure of its life span is number of reads and writes to each physical address on the chip. Thus, when data is not actually being accessed, the flash storage experiences virtually no wear, while the disk drive ages every minute. This is a major difference that needs to be considered in terms of the overall lifetime cost of storage.
And the failure of a few flash sectors does not mean that the entire unit needs replacing, while the failure of one bearing in a disk drives ends the life of that entire drive. Of course sectors still do fail without warning at the end of their lives, but other sectors on the same flash may have long lives ahead of them. Thus, while data still needs backup, that copy can be made to different sectors of the same flash unit with at least as great a degree of safety as provided by RAID. This equates to a much simpler architecture, with greater overall dependability.
Disks, of course, can lose individual sectors, for instance to head crashes, which are often caused by power failures. As a result, data centers commonly invest in UPS not only for their computers but also for their disk farms. Flash, having no equivalent of a disk head, is much less vulnerable to power interruptions. Data that is being written to a sector may be lost, but anything already in the flash unit is unlikely to be damaged. And also unlike disk, once the power returns, flash is instantly available. Thus depending on the application, a bank of flash storage may not require a UPS.
Flash also uses less power than disk under equivalent loads and can easily be powered down when not needed, and restarted almost instantly when it is, saving more power. That power saving is doubled because it means flash generates less heat and therefore requires less cooling.
Action Item: When considering flash storage, look at it as a different technology than disk, with its own advantages and disadvantages, and look at total lifetime cost, not just purchase price.
Footnotes: Enterprise Flash Drive Cost and Technology Projections