Accelerate Critical Applications
Accelerate Critical Applications
Should enterprise data centers look to NVMe storage to accelerate critical applications?
Join the DZone community and get the full member experience.Join For Free
Self-hosted vs Managed Service? Learn how managed enterprise graph databases reduce project costs and increase time-to-delivery.
The need to accelerate business critical applications is driving enterprises to search for ever faster storage and network connectivity. Indeed, 2018 seems to be the year in which every sector of the economy is asking if it’s possible to gain quicker insight from business analytics software — even while data sets to be analyzed are continuing to grow in size.
For this reason, the real price of storage is no longer the cost of data at rest, but the velocity at which it travels over a network and the time it takes to draw meaningful, actionable conclusions from it. This is why enterprises are starting to explore NVMe-oF (Non-Volatile Memory Express over Fabric), particularly financial services businesses, online retailers, cloud service providers, and healthcare companies.
Although we are still in the early adoption stages, businesses are starting to realize that SAS/SATA connections won’t let them exploit the full speed and power of their solid state media. In terms of performance, IDC suggests that Network-attached AFAs using NVMe over Fabric can deliver more than 50 percent lower latencies than the best SAS/SATA-based AFAs, and bandwidth and throughput can be as much as 10 times higher. This is particularly important now that the SSDs are getting even faster, with access times approaching 10 microseconds or less. It is key to many enterprises because it can unlock access to more real-time workloads, particularly those critical data analytics applications.
Getting to Grips With Timely Data Analytics
Data analytics helps build business insight and enables companies to jump on trends that can grow their profits and extract the full value from the information they hold. This is why Gartner and IDC both predict that a significant part of growth and profit for organizations will be built on their ability to extract, process and action information.
In some respects, it’s also a self-fulfilling prophecy. As businesses start to see the benefits of NVMe-oF at a rack level, they are going to expand the size of their databases and data centers. Improved processing speeds will make it possible to work with even larger data universes which, in turn, opens up possibilities for advanced supercomputing applications and much more insightful and actionable data analytics.
For a real-world example, consider how this might work for a financial company. Financial firms need to read and process data fast in order to make the right and the most profitable business decisions. However, data is only relevant for a limited time and there is a tight window for making the right decision. This poses a major problem in the world of high-speed analytics and big data management. Not enough data, and/or not enough time to process it can result in a decision that is less accurate and that might even cost money — particularly if running investment or fund risk assessments, for example. Replacing local SSDs with shared accelerated NVMe-oF storage has the potential to yield a 24-to-1 storage consolidation, provide increased capacity at 1/3 of the cost of local SSDs and deliver on par performance.
But it's not just the financial sector that can benefit, it could also have a transformative effect on healthcare, especially in genomics. Each initial genome analysis produces about 300GB-1TB of data per person. A secondary analysis generates a further 1-4GB of data called the “variants.” On a regular network with SSDs and a SCSI connection, the secondary analysis servers could take anything up to 5 days to crunch the 300GB-1TB and produce the 1-4GB of “variants.” That's because of the bottleneck on the network while the data is in motion. Using NMVe-oF instead has the potential to bring processing speeds down to one day. This would be particularly helpful when running multiple tests on an individual or processing lots of patients at the same time in a busy hospital.
In terms of costs, the IDC suggests that NVMe-oF is currently priced at a premium as the technology is still so new, however, that premium is quickly going away as larger storage vendors begin to ship NVMe-based systems. Additionally, there are a number of start-ups around the world that are working to find lower cost ways to exploit the benefits of NVMe-oF and bring the technology to a much wider user base. It’s therefore unsurprising IDC predicts that by 2021 50% of primary storage vendor revenue will be generated by NVMe based storage platforms and NVMe over Fabric will be in use on most of those systems.
With that in mind, it makes sense for enterprises to start exploring the technology right now and gaining an understanding of what it can do for them. This will enable them to integrate it into their own environments in the most cost-effective way and to take advantage of the benefits before their competitors.
Opinions expressed by DZone contributors are their own.