Some choices are easy, some choices can be tough. We can nudge the tough choices nearer to the easy ones when we have clear information.
For example, multiple factors influence NVM Express (NVMe) SSD selection. Physical characteristics — like form factor, height and power draw — and basic capabilities — like endurance, capacity and performance price — are all important. The characteristics that are most important for you to meet your specific needs will determine which SSD is right for you.
Where to start? Which NVMe SSD is the right one to solve your challenges? Getting started can be difficult.
Micron nudges that tough choice closer to the easy choices by building an NVMe SSD portfolio that’s both broad and tuned. Broad enough to help with nearly any workload and any application and tuned to ensure a great fit. Micron’s broad NVMe SSD portfolio enables precise application and workload matching, helping you take advantage of the benefits of NVMe flash storage, accelerating the enterprise and the cloud.
What’s in a form factor? Turns out, quite a bit.
It wasn’t so long ago that, when we were designing an enterprise platform, boot could be a challenge. Typical enterprise platforms had to give up a storage slot for their system boot device. Back then (the “good old days”?), our platforms usually couldn’t build in a dedicated boot device, so we used a front-load drive slot. Sure, we would have liked to use that slot to increase how much data we could store per platform, but it wasn’t really a choice. Because we only had storage slots, we used them for boot-up.
Then we started building storage systems with internal, board-mounted sockets for system startup devices. We could have nearly any popular interface we wanted — whatever we were ready for.
Here’s how the Micron® new 7300 NVMe SSD fits: Startup devices need good read performance (GB/s), and they may need excellent write performance as well. They should be available in relatively small capacities (say, 400GB or so) since we’ll be storing our OS there but not much else.
Our 7300 is available in a small M.2 form factor with capacities starting at 400GB. It also comes in PRO (read-intensive) and MAX (mixed-use) versions (with the PRO versions being a bit higher-capacity) and U.2 form factor for mail storage.
Our 9300 is available in the 15mm U.2 form factor. This is because it offers powerful performance and extremely large capacities (up to 15.36TB per SSD).
Sharing powerful, high-capacity NVMe SSDs using namespaces
Hang on a second — NVMe namespaces and shared NVMe SSDs? Yes! Let’s take a brief detour into NVMe namespaces. (If you’ve already been down this road, you can skip ahead.)
What are NVMe namespaces? In this post from nvmexpress.org, you learn that NVMe namespaces are made up of nonvolatile memory that can be exposed to host systems in sections or blocks. These sections or blocks are shareable (or not, your choice). Pretty cool, right? But wait — that sounds like partitioning, something we’ve known about and used with hard-disk drives for decades. What’s so different about NVMe namespaces? Glad you asked.
For this blog, I’ll focus on the biggest difference — resources and queues.
When we partition storage on a disk drive, we divide that storage up into logically separate groups of LBAs (logical block addresses). Hosts target input/output to specific LBA ranges (via the associated partitions), and all that I/O goes through the same queue.
Read that again — the same queue — meaning that all I/O may be fighting for resources, depending on which partition it is directed toward. If multiple requests come through, there’s a lot of fighting, a lot of competition and a lot of opportunity for data to get intermixed.
See how mixed up the partition I/O is at the top left? If we’re sharing high-performance storage, we want to do it efficiently, but that competition and complexity can’t be good for I/O performance.
NVMe namespaces are different than partitions. With namespaces, NVMe SSDs offer dedicated submission and completion queues, helping NVMe SSDs keep the I/O for each namespace separated.
Performance-focused SSDs like Micron’s 9300 can easily handle multiple I/O requests and multitenancy deployments in hyperscale data center and cloud infrastructures. Namespaces can improve elasticity for and use of a broad range of applications that will thrive on cohesive I/O.
NVMe SSDs are designed for broad, multiple workload adoption; form factor flexibility; and compact, low power.
A few more factors to consider
When you’re looking for an NVMe SSD for your next project, ask yourself this: Are you planning on sharing it with multiple applications, or will you be using it as a building block at the platform level? If you’re planning on sharing, the 9300 is the way to go. (Here’s an interesting post on using NVMe namespaces to improve application performance from @Micron Storage and Wes Vaske.) If not, then other factors will steer your choice.
And that’s not all! Micron has easy-to-use guidance to start you down the right NVMe path — with information ranging from the basics of NVMe SSD design to port options, power use, endurance and performance (IOPS and GB/s). There are multiple factors that influence NVMe SSD selection; your specific needs will determine which factors are most important for you, helping you select the right Micron NVMe SSD.
As I said before, Micron’s broad NVMe SSD portfolio enables precise application and workload matching, helping you take advantage of the benefits of NVMe flash storage, accelerating the enterprise and the cloud.
Learn more about Micron’s complete NVMe SSD lineup at www.micron.com.