Posted 20 November 2015 by Chris Evans
This week has seen two announcements from flash array vendors that are moving to support 3D-NAND technology in their products. On Tuesday Pure Storage announced that 3D TLC NAND would be supported in their FlashArray platform from 1Q2016. The new drives will increase capacity from around 30-40TB to 60TB per rack unit (depending on the model) with pricing around $1.50/GB (effective capacity).
HP Enterprise have announced that the 3PAR platform will also support 3D-NAND drives, although it’s not clear whether these will be TLC. Availability is from December 2015 and again, HPE are quoting a $1.50/GB (effective) price point.
Kaminario has been supporting 3D TLC NAND for a few months now and according to Chris Mellor at The Register, so has SolidFire. Dell has been using the technology too, with street prices down to $1.66/GB.
Why have vendors moved to this technology? There have been a number of reasons:
- Managed Endurance – vendors have been able to bring MLC levels of endurance to TLC products.
- Overestimated customer need – customers haven’t needed the level of I/O (especially write I/O) expected from flash and so can cope with lower endurance products.
- Cost – customers want storage to be cheap and using TLC allows TCO levels that are in many instances cheaper than disk.
Samsung (the only current 3D TLC NAND vendor) indicated in a presentation I attended last week that their current 3D products (using 48 layers) will scale from 256Gb to 1Tb chips within 2 years – a 4x increase in capacity. This means up to 192 layers and who knows what the limit of the technology will be. Samsung indicated that planar SSD will pretty much die out and all future products will be based on 3D technology, with well over 75% of the market being TLC by 2019. Ignoring the micro-details of the technology for a moment, we can predict that flash will continue to grow in capacity and will continue to fall in price, in line with the way storage prices have declined for decades, making it more cost effective than hard drives.
To stay competitive, the all-flash vendors will need to adopt 3D and TLC technology, as within 12 months, the benchmark effective cost for flash arrays will be below $1/GB. How long will it be before we see fully tiered flash products with multiple levels of NAND capability? How many architectures will cope with multiple hardware tiers? This could be an interesting next step in the future evolution of all-flash systems.
Here’s another thought; where does this leave the hybrid guys? Tintri have delivered an all-flash product, as have Tegile, however Nimble Storage hasn’t (and coincidentally has see a dive in their share price over the last 2 days, following poor results). All-flash using 3D-TLC NAND could quickly eliminate the cost benefit hybrid solutions were offering, making it very difficult for the hybrid-only providers to compete. It’s a fast-moving tough market for the array vendors, which promises a rollercoaster 2016.