Micron has started to ship its new 7450 NVMe SSDs, which bring 176-layer NAND capacity to the datacentre for the first time.
In capacity terms that means drives of 15.36TB, compared with 7.68TB in the preceding 7400 product with 96 layers. The extra capacity afforded allows the use of TLC NAND, which provides faster reads and better durability than the QLC flash usually required for bulk storage SSD use cases.
Use cases could, in fact, range from deployment as a server boot disk to intensively accessed database storage. SSDs of more than 8TB have generally only been used in cold storage arrays, often NAS hardware that holds backups or object storage systems built to share media that will only ever be read rather than written to.
“The reality is that the IT hardware market continues to suffer supply chain problems,” said analyst Patrick Moorehead of Moor Insights and Strategy.
“In this context it is fundamental to be able to get all the capacity you need in the minimum number of units,” he said. “But this should not be the case just for archiving use cases. Hyperscalers need fast and durable SSDs to maintain the levels of service they offer and enterprises need to run databases.”
According to Micron, the 7450 provides latency of less than two milliseconds in routine workloads that comprise 70% reads and 30% writes. That low latency allows it to achieve one million IOPS. The 7450 connects via PCIe 4.0 and uses four channels, with potential throughput of 6.9GBps.
Micron will offer the 7450 SSD in close to 30 versions that correspond with delivered capacity, workload and format. Micron aims to cover all datacentre use cases with formats that include classic 2.5” disk (U.3), the M.2 format typical of extension cards and the E1.S format that is part of the new EDSFF form factor.
Read more about flash storage EDSFF brings built-for-NVMe connectivity to the datacentre. Aimed at enterprise use cases, EDSFF offers incredible flash density in shared storage and hyperscale deployments, but storage products that use it are still thin on the ground. QLC flash storage: What it costs and its best use cases. QLC flash brings high capacities at costs coming down close to spinning disk. So what are the use cases for QLC solid state, which make best use of its relatively low endurance? In U.3, capacities range from 960GB to 15.36TB. In E1.S format they go from 960GB to 7.68TB. In M.2 they start at 480GB and go to 3.84TB. The SSDs vary in size and length according to their capacity. That applies to versions called “Pro”, which support complete daily writes of their capacity for five years.
Micron also offers the “Max” version, which supports three total re-writes per day. Here, capacity is a little reduced because cells serve as emergency resources when principle cells are at capacity. For example, in a Max drive, an SSD of 15.36TB has a useable capacity of 12.8TB, while the 960GB version offers 800GB.
TLC vs QLC flash NAND flash technologies depend on switching using small electrical charges in very small cells in the chip, and so the various flash generations have been distinguished by the number of voltage switches they have packed into a NAND cell, hence single-level cell (SLC) was the earliest of them, with just a single zero or one possible. The latest is quad-level cell (QLC), with four bits per cell and 16 switching states possible.
But packing more and more into small spaces comes at a cost. The small voltages and extremely small sizes at which they must operate mean a consequent decline in durability and make them more susceptible to interference, both physical and electrical.
And so, while QLC is an advance that brings more capacity in a small size, it also means the format must be more read-heavy, to avoid wear, and have lower endurance.
That’s why the ability to pack more capacity into an older flash generation – i.e. TLC – is seen as a good thing.