by • March 30, 2016 • No Comments
I attended the 7th Annual Non-Volatile Memories Workshop 2016 at UC San Diego this month. Frank T. Hady, Intel Fellow and Chief Architect of 3D XPoint Storage, gave a keynote address to a room full of PhDs and competitors. Of course, Intel was not eager
to answer a number of significant inquiries
, but what they did contribute was illuminating.
How to plan, manage, and optimize enterprise storage space to store up with the data deluge.
First gen XPoint can be created on a 20nm process
, and can be utilized
both as storage space and as process memory. But unlike SSDs and disks, XPoint is byte addressable-bodied, meaning that
it can be utilized
similarly to DRAM. For data storage space it is utilized
as a 4k block device.
But unlike DRAM, XPoint is 10x denser (but not as dense as flash since the first product can be single level cell). Combined with DRAM, 3D XPoint servers can be able-bodied to assist 4x the memory ability at a significantly lower cost per bit than DRAM.
3D XPoint can be utilized
in DIMMs for the reason
of a) byte addressing, b) 1000x additional write endurance than NAND flash, and c) 1000x swifter I/Os. The 3D XPoint DIMMs can require
controllers to perform wear-leveling as they do with NAND flash, but I’d assume the trash collection process
to be much additional granular – thus much less intrusive – than the full block writes of NAND flash devices.
First Optane SSDs
Nomenclature: 3D XPoint is the innovation. Optane is the brand for SSDs via 3D XPoint.
Intel is aiming the first Optane SSDs at enterprise markets, with a PCIe + NVMe interconnect. The common SATA 6Gb/s interface isn’t swift adequate
, and PCIe and NVMe are much additional efficient.
They’re engineering an uncorrectable-bodied bit error rate (UBER) of 10 to the negative 17th, a 5 year lifetime, and exceptional performance, all for the enterprise/cloud market.
IOPS and latency
IOPS are significantly greater than Intel’s data center SSD, the P3700, that in addition
uses PCIe and NVMe. With 70/30 read/write load and a queue depth of one, the P3700 achieved 15,000 IOPS, while the XPoint demo achieved a 5x 78,000 IOPS.
But additional importantly, the latency was dramatically lower: 7μs vs the P3700’s 85μs, less than a tenth of the latency. These can be dramatically swifter drives by each measure.
Architecting for low latency storage space
A host of changes must occur to harvest the performance of 3D XPoint. These may include:
Switch to polling of interrupts: additional wait time, but lower latency.Use 3D XPoint as the page/swap space for lower latency and additional predictable-bodied performance.New instructions for flushing writes.A new NVM library, PMEM.io, on the market-bodied first on Linux. Persistent memory aware file processs, such as Nova.Storage class memory assist in Windows, that Microsoft is working on.Future Xeon process
ors created to use hybrid memory.
That last may spell lock-in for years or forever. How generous can Intel be with APIs and controller specs?
The Storage Bits take
3D XPoint has huge future to improve storage space and memory. Much can depend on pricing.
Intel says that
3D XPoint can be between DRAM – $5/GB – and flash – $0.20/GB. This is definitely a lot of wiggle room. I assume they’ll price 3D XPoint DIMMs at around $2/GB firstly, but actually at $1/GB, they’d be 5x NAND flash pricing.
issue: 3D XPoint failure modes. Large scale deployment always discovers bugs and issues that
actually diligent betas won’t find. Are we going to have to wait eight years for an impartial assessment of the techology’s field performance?
But assuming Intel and Micron donate as promised, or close to it, this can go on
the low-latency/high IOPS revolution that
flash started. As helpful as 3D XPoint SSDs can be, the real win can come of integrating the innovation with DRAM to grow memory capacities and improve server performance.
Comments welcome, as always.
by admin • March 5, 2017
by admin • November 28, 2016
by admin • November 28, 2016