Questionable report questions enterprise flash’s future

A new report claims that SSD adoption is doomed to stall in the enterprise, but uses highly questionable metrics for that conclusion.








As NAND flash shipments continue to boom, the expectation has been that the technology would continue to gain ground in both the consumer and enterprise markets. A new report at the Enterprise Storage Forum challenges that assumption, arguing that NAND’s benefits have been vastly overstated. The author makes some excellent points about the headwinds that the SSD market faces, including high prices, and data retention challenges at lower process geometries. Intel and Micron have had to work hard to bring each generation of enterprise NAND up to snuff, which is one reason why new products are introduced on consumer hardware first.
But the performance metrics introduced in this story as purported evidence for why the SSD case in enterprise has been overstated
As NAND flash shipments continue to boom, the expectation has been that the technology would continue to gain ground in both the consumer and enterprise markets. A new report at the Enterprise Storage Forum challenges that assumption, arguing that NAND’s benefits have been vastly overstated. The author makes some excellent points about the headwinds that the SSD market faces, including high prices, and data retention challenges at lower process geometries. Intel and Micron have had to work hard to bring each generation of enterprise NAND up to snuff, which is one reason why new products are introduced on consumer hardware first.
But the performance metrics introduced in this story as purported evidence for why the SSD case in enterprise has been overstated are laughably terribly. If you have to invent an entirely new metric for performance measurement (drive price divided by GB/s of bandwidth), you derive that measurement differently for different drives, and you measure power consumption in W/GB/s of bandwidth as opposed to a simple wall socket test, you’re not going to walk away from the comparison looking real smart.

Where SSDs do and don’t make sense

Multiple reports and documented studies over the past few years have compared enterprise hard drive performance against SSDs, which is why I’m surprised to see a supposed takedown of the topic rely on such half-baked analysis. What performance studies consistently demonstrate is the real-world performance gap between SSDs and HDDs is directly dependent on concurrent data requests. The following graph from a Dell performance comparison illustrates this well:
. If you have to invent an entirely new metric for performance measurement (drive price divided by GB/s of bandwidth), you derive that measurement differently for different drives, and you measure power consumption in W/GB/s of bandwidth as opposed to a simple wall socket test, you’re not going to walk away from the comparison looking real smart.