Over the past year, we’ve seen more Virtual Systems clients move their full stack to the cloud than any previous year. They’re storing data and backups, creating DR sites, using cloud instances to power their applications, and trading physical computers for cloud workspaces. As more customers deployed cloud technologies in 2021, Virtual Systems explored the idea of upgrading all customer production workloads to NVMe storage as an avenue to maximize customer performance and experience in the cloud. This summer we finally did it and we now offer NVMe upgrades at no additional cost to our customers. Let’s explore why Non-Volatile Memory Express aka NVM Express™ (NVMe) is important to our operation — and yours. (You might want to buckle up for this one because we’re gonna get a little nerdy.)
What is NVMe SSD?
There is some confusion about what NVMe actually is; for example, we often hear people ask if it’s “better than solid-state drive (SSD) storage”. NVMe isn’t better or worse than SSD – it is SSD. The best way to think of NVMe with respect to SSD is to understand NVMe as a “type” of SSD, or an enhancement of SSD rather than a competitive technology. NVMe is a communications interface and driver that enhances the performance and compatibility of PCIe SSD with flash technology. Whew. Confused yet? Stick with us, it gets nerdier but it’ll all make sense soon…
NVMe replaced Advanced Host Controller Interface (AHCI) communications and drivers, which use a high number of CPU cycles and aren’t compatible with every system. And, since it is an older technology, it wasn’t designed to support today’s high-performance computing.
- When you compare NVMe and AHCI:
- NVMe has 64,000 command queues; AHCI has 1
- NVMe can send 64,000 commands per queue; AHCI has 32
- Latency (the time it takes for data to travel or one component waits for the availability of another) is 2.8 microseconds with NVMe; It’s 6 microseconds with AHCI
- NVMe supports 1,000,000 input/output operations per second (IOPs); AHCI supports 100,000
Additionally, NVMe communicates directly with CPUs. AHCI, by contrast, requires an extra step of communication with a SATA controller that, in turn, communicates with the CPU. Maybe some of the terminology is confusing but I hope you see where this is going. NVMe takes on some of the CPU/RAM demand from a workload and maximizes the actual CPU/RAM assigned to that workload by freeing those resources from having to carry tasks that could be assigned to the storage. It’s like adding additional CPU/RAM resources to a workload but doing it directly on the storage.
What NVMe Means to Your Business
NVMe’s stats translate to some important benefits:
- Speed: NVMe is designed for speed, using a direct path to processors, decreasing CPU cycles, and decreasing latency, aka wasted time. It enhances the performance of the applications you use now, and it also gives you a runway for real-time data applications, Big Data analysis, advanced technologies, and more.
- Efficiency: This technology also enables data compression at an average of 8 times more than legacy technologies. Highly efficient storage can mean reductions in other resources like CPU & RAM. Efficiency also means your cloud footprint can do more with less, helping you achieve better performance with less cost and problem management.
- A drop in infrastructure-related software and OS issues. Aligning your infrastructure to your software stack and database workloads is both an art AND a science. An application guide might tell you your software will be happy on cloud instances offering 2Ghz/core but your app won’t act happy until it has 2.8Ghz/core underneath it. What appears to be a “buggy application” may be an application that’s simply coded to want more resources. You may never see that problem in your infrastructure health reporting but it’s there just the same. Putting NVMe storage under your OS and applications creates the optimal environment for all workloads, reducing issues related to resource starvation that don’t show up on infrastructure intelligence reports.
Wasted time and inefficient storage add up to unnecessary time, effort, and costs. Overcoming those challenges will reduce the total cost of ownership (TCO) of storage and, in turn, can enable business growth with greater digital capabilities and more cost-effective operations.
Shared Accelerated Storage, Thanks to NVMe
Gartner has pointed out another benefit of NVMe – bringing together the benefits of network-based storage systems and direct-attached systems, which the analysts at Gartner call “shared accelerated storage” (SAS).
Pure Storage explains that NVMe over Fabric – i.e., over Ethernet of InfiBand – reduces latency that existed when storage was outside the server, now making it comparable to attached storage. Also, NVMe allows processors to focus on storage rather than I/O processing. Pure Storage refers to this as “data-centric architecture,” designed to focus on data rather than IT systems. This architecture is also more scalable and flexible than other options, and it can enhance data backup and security.
We Have Your Back
This blog is definitely more technical than most on our site and even though we didn’t go into all of the details about NVMe, we know most of our readers are less interested in speeds and feeds and more interested in what technology does for your business. So, let’s distill this down to a short elevator pitch: NVMe is faster, more cost-effective, and more generous to your applications than last-gen technology, which makes your cloud experience cleaner, faster, better. And, if you’re running your applications with Virtual Systems, now you’re using it.
Contact us to discuss the whole picture and to learn how Virtual Systems can help make your digital transformation better.