©2019 by Circuit Blvd., Inc.

  • Bumsoo Kim

Open-Channel SSD: Easy to build and use?


People easily guess that the Open-Channel SSD will ease the process of building SSDs.


It's less than 30% true.


The Open-Channel SSD initiative, like other software-defined something, was started to solve (1) the problem of mismatch between memory life cycle and SSD deployment process in cloud industry; and (2) difficulties of accommodating new storage requirements in the cloud with standard SSDs.


Traditional SSD firmware is running on a specially designed ASIC, featuring standard functionalities. Its coverage of memory management is contained in a single SSD module which is divisible into just a few separate logical spaces. All firmware components were seamlessly integrated into a single image to provide best-of-breed performance.


Taking out complicated mapping, wear-leveling, garbage collection, scheduling from the SSD, and leaving only lower level NAND access and ECC functions inside the SSD is the simplest way to make an Open-Channel SSD. The problem comes in immediately after this separation. The Open-Channel SSD spec has almost infinite freedom of interpretation and implementation. This means you cannot use an Open-Channel SSD as a standard commodity and have to properly implement both its firmware and host side software.


If you say that storage system development with the Open-Channel SSD is straight-forward and easy, and making and selling an Open-Channel SSD is taking a low-hanging fruit even for new memory manufacturers, you are more than 70% wrong. This is actually opening Pandora's box. It will affect the whole process of business, including logistics, procurement, deployment, operation, replacement, sales process, customer relationship management, partnership, as well as technological artifacts.


#openchannelssd