These days we all look back at our first SSD and marvel at how tiny it was in capacity. Your author remembers reviewing super-tall 32GB drives that, despite their capacity, were shaped like Nintendo Entertainment System (NES) cartridges when they first came out so many years ago. We have all moved on to 512GB and 1TB/2TB SSDs by now, but someday we will look back at those and laugh at how we could ever stomach such lowly capacity. That is, assuming Samsung comes through on a 256TB SSD it showed off this week at the Flash Memory Summit in Santa Clara, CA.
The company was very light on details about this future drive at the summit and certainly did not say when we could buy it or how much it would cost, which is unfortunate. The company barely mentioned it, stating only that it is designed for enterprise and that it's something it is currently working on. It will be made for applications that require "maximum data storage," obviously, but with a nod to the power requirements of a single server rack. Despite its humungous capacity, Samsung says this quad-level cell (QLC) drive uses seven times less energy than combining eight 32GB SSDs. Sadly, that's all the details Samsung offered on this future storage device.
The company also unveiled a PBSSD architecture, described as a "petabyte-scale ultra-high capacity solution." Since a petabyte is 1,000 terabytes, we can see it going together with the drive described above. It says it's been working with Meta on this, and it increases scalability by varying the capacity of the storage array according to the applications being used. Essentially, it sounds like a cloud storage solution where multiple users are utilizing the same storage device, where it can group those users' data together by the applications they use. It says it's deploying "Traffic Isolation technology," which "eliminates interference in performance and response delays even when multiple users use a single SSD."
One of the keys to making this work is Flexible Data Placement (FDP), which Samsung says has already been ratified in NVMe for future usage scenarios. This technology allows the host controller to specify where data is written for different users, which can reduce write amplification, which is a situation where the data written to a drive is multiplied and not the logical amount required, thereby reducing the drive's lifespan. Samsung says the technology it uses for this is entirely open-source as well.
