People familiar with enterprise-class SSDs aren’t surprised that they never “took over” the data center as advertised. In fact, given their cost and reliability issues, it seems likely they will be relegated to a fairly small portion of overall enterprise storage needs for some time.
But that doesn’t mean SSDs are dead in the water, developmentally speaking. As recent advances indicate, SSDs have only just begun to tap their enormous potential, with each successive generation promising to take on more responsibility in highly complex data environments.
Many of the recent gains have centered around the I/O interface that connects the SSD to the wider world. SMART Modular Technologies, for example, recently launched an updated version of the XceedIOPS device with a 6 Gbps SATA interface that delivers up to 60,000 random IOPS and a sequential rate of 520 MBps. This is on a 1.8-inch form factor ranging from 50 to 400 GB.
And even as throughput is increasing, cross-platform interoperability is improving as well. The Open NAND Flash Interface (ONFI) working group recently published its 3.0 standard, effectively doubling NAND interface transfer rates to 400 MBps. The standard uses non-volatile DDR2 technology that will allow upcoming NAND controllers to either increase performance or maintain existing performance with half the channel count, lowering both the cost and footprint of future SSDs.
Action is also happening in Flash memory itself. Intel recently announced a new 25 nm device aimed largely at notebook and PC applications, where it will provide dramatically increased density at lower cost. The SSD 320 Series, as it is known, will range from 40 to 160 GB and supports the 3G SATA II interface to double sequential write speeds to 220 MBps.
That’s not the end of the story, however, as Intel and Micron say they are preparing to release a 20 nm chip within the next month or so. The move would represent a significant improvement in Flash technology, essentially doubling densities at the chip level. However, we’ll have to wait and see how well the companies overcome the significant engineering challenges when manufacturing circuitry that small — literally, thicknesses measured in atoms. Expect to see some fairly advanced error correction and signal processing on a working device.
All this talk about advanced Flash architectures and high-speed interfaces almost makes it seem like today’s cutting-edge technology is already on the way out. And in a way, it is. Despite the high upfront cost of SSDs today, their shelf-life is likely to be rather short. That’s why it’s probably best to reserve capital budgets for SSDs only where they are truly needed, say, in high-speed transactional or database environments.
There may come a time when SSDs can be comfortably deployed in more general-purpose settings, but that certainly won’t be tomorrow.