At last week’s USENIX conference UCSD researcher Laura Grupp presented a paper that attracted a lot of attention. The paper, which had a somewhat misleading title: The Bleak Future of NAND Flash Memory predicts that NAND-based SSDs of the future: “may be too slow and unreliable to be competitive against disks of similar cost in enterprise applications.”
The paper was co-authored by UCSD researchers and Microsoft Research.
Of the three articles I have read about this paper two focus completely on the “End of NAND” which was just an underlying assumption used by the researchers to determine a point in time when NAND would stop being used. It was an input to the paper, not its conclusion.
The team’s conclusion is a simple one – if NAND endurance, program speed, retention, error rates, disturb, etc. continue to degrade along current trends as newer higher-density parts are introduced, then future SSDs using today’s SSD architectures are going to significantly underperform today’s SSDs, and may be no more appealing than HDDs.
It’s an interesting conclusion, but I would argue that today’s SSD architectures are very immature and that architectural improvements will more than offset the downward trends in NAND flash performance.
Objective Analysis will soon be publishing an update to our popular Enterprise SSD report. Keep your eye on the Objective Analysis Reports page for its release, or become a subscriber to Objective Analysis Alerts.