EMC's commitment to solid-state disk drives this week sure had a lot of people talking, if not necessarily about the end of standard drives then at least about a viable alternative for near-line storage in the enterprise.
But the question remains: Is it just the price of SSDs that are limiting their ability to perform at enterprise-class levels, or are there technical/performance considerations as well?
First, let's see exactly what was said at EMC World this week. The company said it will deploy SSDs throughout its enterprise-class storage arrays, with the expectation that the increased demand will cause prices to fall to the level of Fibre Channel drives in the 2010-2011 time frame. Computerworld quotes EMC Executive VP Dave Donatelli saying:
The market for flash is coming down significantly faster than rotating drives right now. Our stated corporate goal is we're trying to drive it down as fast as we can.
Two key considerations for EMC are the improved throughput that flash technology offers and its ability to enhance data deduplication through improved search.
But it would be a mistake to assume that EMC thinks SSDs are the only drive of the future. In fact, The Register quotes Donatelli quite clearly:
We are by no means saying that spinning disks are dead -- we see mixed environments. The disk drive will be around for a long time, particularly ATA drives.
That is clearly the case when it comes to swapping out tape-based backup systems. The company unveiled a set of three Disk Library systems that pair spinning-drive technology with dedupe, drive spin-down and other techniques to increase capacity while cutting power requirements. The top-level DL 4000 system scales into the petabytes with 1 TB SATA drives spinning as low as 5400 RPM.
Meanwhile, it seems that Intel is determined to get into the SSD market this year. The company anticipates a third-quarter launch of its Client X18-M line, a 1.8-inch drive for the Centrino notebook and the Client X25-M 2.5-inch device that could find its way into enterprise servers. Initial releases are expected to be 80 GB units with Serial ATA interfaces, with the potential for 160 GB versions by the end of the year and 250 GB in early 2009.
This is all well and good in terms of price and availability, says Amyl Ahola, CEO of flash drive maker Pliant Technology in his Enterprise Storage Sense blog. But the fact is that SSDs still don't come close to meeting the MTBF and error-rate specifications demanded in the enterprise. What's needed is an advanced error-correction technique that will preserve the integrity of the drive even after multiple erase/write cycles, and without hampering I/O performance to boot.
I'm wondering, then, if it's safe to assume that Pliant Technology is close to announcing just such a system?