It is 2015 and flash storage continues to deliver on its promise to transform the performance and economic capabilities of the enterprise storage market. The need for ever increasing capacity in the same amount of physical space has been constant. The flash manufactures have dealt with this in two main ways: cramming more data into a single NAND flash cell or making smaller cells, called geometry shrink, and packing them closer together. Both methods have similar outcomes of increased capacities and lower costs, but come with reduced performance and reliability.
The attempts to store more data on a cell has led us from the single level cell to the multi-level and triple level cells, with each subsequent cell architecture increasing the amount of data being stored. As they store more capacity they become more prone to errors and write endurance problems, which have slowed the adoption of TLC in the enterprise.
The use of multi-level cells in the enterprise space is common practice, but has started to run into the flash capacity wall. Flash manufactures are turning to triple level cells to help overcome that wall. Since TLC stores even more data on a cell it comes with an increase in wear leveling error correction requirements, and power consumption. Up until this point triple level cell architectures have been seen as only suitable for consumer applications because of its low write cycle levels. It is simply not ready for the demanding write workload of the enterprise; however that is changing as technology advances.
Today’s triple level cells will not replace multi level cells in the modern enterprise storage ecosystem anytime soon, and likely not ever. Technology like 3D NAND is far more suitable for ousting MLC as king of the enterprise, however TLC has a place in the hybrid storage array. Until now, the hybrid storage array has been a combination of high performance flash and high capacity spinning media. Flash was used to, essentially, accelerate the performance of the capacity tier. This model has worked very well in enterprise storage arrays and server SAN software offerings to allow for dramatic performance improvements while balancing costs.
Triple level cell based architectures are far more affordable than MLC based, but are still more expensive than spinning media. The price gap between TLC and traditional hard disk drives is narrowing at a very quick rate. When data reduction techniques like compression and deduplication are factored in the price gap is not that far off. In the future the raw costs between flash and magnetic media will be comparable. The path of TLC in the all flash storage system in the enterprise is clear. All flash configurations will become all flash hybrid arrays with a high performance; write endurant tier and a low cost, low write capacity tier using triple level cell drives.
The biggest unknown is how 3-D NAND technologies will impact triple level cells in the enterprise from a price perspective. Flash cells have been getting smaller and smaller which makes packing more data into them a challenge. This is especially hard with triple level cells at 16nm. At that size is becomes a very big problem in terms of write endurance and cell reliability. This is the scaling limit of flash on the market today. Soon we will be in the year of TLC in the enterprise, but the question is how long will that year last?