For the past three years, consulting firms and others have raised questions about the ability of NAND flash memory to handle large number of "writes" over time, more commonly known as the NAND write limitation issue. These questions raise concerns about the reliability, cost and durability of flash memory and solid state technology in general as it becomes more popular for enterprise systems.
But is there really a write problem that limits NAND in enterprise computing? This hasn't been a major problem in consumer devices like cellphones and digital music players that have been the largest market for flash memory to date. But as NAND flash makes its way into more enterprise storage products, these issues grow in importance.
"Write limitations in NAND are not really an Achilles' heel to end users," says Wayne Adams, chair of the Storage Networking Industry Association (SNIA) and an executive with EMC Corp. (NYSE: EMC) "When you buy a car, it is optimized for a certain oil-fuel mix, but you don't know all the engineering that went into the product to get it to perform. What's going on with NAND is analogous. Engineering algorithms are being applied to improve NAND write performance in the same way that similar technology is employed in hard drives to address their inherent limitations in moving data around. There are always certain parts of media that are ineffective, and we engineer to overcome those limitations."
The current thinking is that, natively, single level cell NAND (SLC) generally supports 100,000 writes, and multiple-level cell NAND (MLC) supports approximately one tenth of the writes that SLC can achieve.
Through a variety of wear-leveling algorithms intended to "spread the wear" over a NAND device, manufacturers incorporating NAND in their products engineer solutions for the write deficiency, extending the lifetime of NAND to the three- to five-year life cycle that hard drives enjoy -- and that IT decision makers look for because it matches their three- to five-year asset depreciation cycles.