The question of how great a role solid-state disks (SSDs) will play in the enterprise is taking on a sense of urgency as reports of the technology's expanding capabilities go hand-in-hand with growing doubts about its reliability.
IBM this week pulled back the covers on its Project Quicksilver effort to reveal that a 4 TB rack of Fusion-io SSDs was able to provide a sustained data-transfer rate of 1 million IOPS, with response times of less than 1 millisecond. That puts it well within the range of high-performance, enterprise-class technology. In fact, that's about 2.5 times faster than any known disk system, according to IBM. The company also says it could have the technology ready for the channel within a year, although it will probably take a little longer while IBM works on the reliability issue.
But we probably won't see any significant gains in reliability with common NAND-based SSDs, according to this interview with analyst Bob Merrit of Convergent Semiconductors. With DRAM and SRAM, no moving parts means virtually unlimited erase and rewrite cycles. But single-level cell (SLC) NAND has been known to wear out in the tens of thousands of cycles -- even less for multi-level cell (MLC) -- which simply won't hold up in high-level enterprise environments. More than likely, enterprise-class SSDs will have to await the development of new phase-change and resistive memory technologies.
The may come as a let-down to the growing number of enterprise managers who are already looking forward to replacing multiple hard disk drives to cut both capital and operating costs. A new study by Objective Analysis pegs the growth rate for SSDs at 151 percent a year through 2013, with the biggest demand coming from transaction-processing applications.
It's that prospect of SSDs replacing only a portion of the hard disk base that has apparently pushed vendors like Intel into the market. The company plans a 2009 launch of an MLC device with a capacity of 320 GB aimed at enterprise server and storage environments. The company has been investigating SSDs for some time and only recently concluded that reliability is strong enough to roll out a commercial product, although the company has not said whether it intends to present its technology for high-end applications.
Like any new technology, SSDs will have to ease their way into the enterprise through experimentation first and then to low-level operations. They are most certainly worthy of consideration as an alternative to hard disks, but it still could be a while before we see them working mission-critical applications.