LWN: Comments on "Taking control of SSDs with LightNVM"

MTD subsystem has an FTL

mbjorling — Mon, 11 Jan 2016 10:17:20 +0000

Many vendors are already working on implementing support for host-based control. There is many ways to differentiate. For example ECC, flash characterization, and extensions that further offload the host.

MTD subsystem has an FTL

mbjorling — Mon, 11 Jan 2016 09:35:41 +0000

MTD solves the problem of directly driving raw flash. Whereas Open-Channel SSDs have flash controllers embedded in hardware, that takes care of scheduling, driving hundred of flash chips, and provide power capacitors for durability. The host primarily handles data placement, I/O scheduling, and garbage collection and leaves everything else to the SSD controller. Making it more efficient for >1M IOPS devices.

MTD subsystem has an FTL

alison — Mon, 11 Jan 2016 05:35:08 +0000

a0273324@ti.com, I had a similar reaction. Why not use UBI as an LVM for NVM? I'll have to compare LightNVM to see how it differs. At least the news is welcome that bypassing the on-board management engine will soon be possible. I wonder how many manufacturers will support the new interface if doing so undermines their claim of having a differentiated storage controller.

MTD subsystem has an FTL

a0273324@ti.com — Mon, 04 Jan 2016 18:58:02 +0000

Is there any reuse from the MTD subsystem (Memory Technology Devices) which provides a similar abstraction for raw flash?

Taking control of SSDs with LightNVM

aleXXX — Tue, 28 Apr 2015 09:13:09 +0000

This surely is not the NIH syndrome at work ?