TREND: Virtualization for the win.
Enterprise IT will continue the server-based storage trend that started several years ago, with more focus than ever on software-defined storage (SDS) approaches like shared high performance NVMe. NVMe storage and persistent memory from Micron are critical components of next-gen IT infrastructure and developed in-synch with the cloud and virtualization leaders.
From our customer and partner strategic planning sessions, we can see big changes just on the horizon. Actual VM & container implementations are saturating data centers, yet there is further opportunities for gains in performance and density. With around 10,000 customers currently for VMware® vSAN, the market leader, growth rates are 150% year over year.
Everybody is researching how virtualized storage can dramatically simplify storage management and reduce costs of infrastructure, how enterprises can dynamically create and provision storage services easily per virtual machine, and control them by policy.
PREDICTION: Flash memory and storage will deliver improved VM performance.
Back in 2014, Micron introduced all-flash storage for vSAN, and proved 20X performance (IOPS) increases over HDDs. VMware saw the future like we did, and provided vSAN 6.0 all-flash hardware guidance.
Things have gotten even better since. Already, Micron has all-flash vSAN 6.2 solutions comprised of Ready Nodes with advanced DRAM and enterprise solid state drives. Micron’s Enterprise Solution team collaborated with SuperMicro to develop a vSAN AF-4 Reference Architecture that delivers 2.5x to 22x more 4k IOPS and 60% to 95% lower average latency than a similarly equipped hybrid design with flash for the cache memory and hard disk drives for capacity storage.
Testing shows all-flash vSAN Ready Nodes also respond more quickly; this lower 4k latency with large datasets is already delivering performance benefits in hyperscale data centers, big data and other workloads requiring lightning fast analytics.
What’s coming next will accelerate use of hyper-converged infrastructure (HCI) bringing a new foundation for VMs. Combined with flexible persistent memory, such as Micron NVDIMM-N™, this will enable low overhead, faster scalability and improved quality of service for cloud-native applications and other latency-sensitive workloads. We expect new opportunities and use cases that go beyond anything we have seen with flash media use today, and not just for VM.
TREND: Have IT your way with hybrid cloud.
Public cloud? Private cloud? On premise or off? Yes, to all of it.
Cloud-based services have replaced traditional enterprise on-premise deployments to become the way IT does business. The hybrid cloud model lets enterprises place their data wherever it works best and migrate when things change. IDC predicts that 80% of IT organizations will be committed to hybrid architectures by 2018. It’s hard to beat the speed of deployment, cost savings and access to modern applications and platforms of cloud computing.
Our conversations with cloud and infrastructure leaders indicate investments in continuing to make cloud computing easier, safer and more agile. The key to a successful hybrid IT environment is to focus on workload and application optimization with ease of management as deployments scale. Luckily, storage and memory infrastructure has also evolved to support this focus.
PREDICTION: Cloud consistency will streamline management.
However, the hybrid cloud creates challenges in management. Data architects demand to move workloads and data gracefully between private and public clouds — greatly enabled by persistent memory and fast flash storage — looking to deliver on-premise, optimized IT services with the speed and operational efficiency of the public cloud.
“We want all our clouds to look and perform the same,” we heard from hundreds at AWS Re:Invent 2017. HCI with ultra-fast, low-latency NVMe drives that can be accessed locally within their host server will form the responsive architecture for private as well as public cloud, helping to provide consistency for easier management.
Providing remote direct memory access (RDMA) that can be shared across the cluster, ensuring high-bandwidth connections like NVMe-oF™ (NVMe over Fabric) will help build consistent performing networks. VMware allows partners like Micron to develop drive software that enables remote management of our SSDs in the vSphere™ dashboard. Amazon is also expanding its AWS™ software with IT management tools. These should help storage architects create data policies that work for multiple workloads across clouds.
TREND: Storage and compute will travel further to the edge.
Edge computing is a new paradigm enabled by advances in cloud computing, storage, mobile, smart devices, and wired and wireless networks.
It’s mostly driven by applications where high capacity is needed to efficiently store and quickly retrieve ever increasing amounts of structured and unstructured data including artificial intelligence, machine learning and IoT.
Do you need a data center at the edge to maximize the benefits? No, but we’re getting there with innovation in edge processing and high-density fast storage, like Micron’s 64-layer 3D NAND technology on a microSD card that packs 32GB of storage on a tiny 59mm silicon die. That’s 5-6 times the density of equivalent performance planar NAND technology. Extending compute to the network edge enables the analytics and other processing, typically near-real-time, when the data is the most valuable – when it’s new, and saving costs on data transmission.
Analytics is becoming layered or phased, with the immediate analysis and response at the edge, and more trend analysis and predictive modeling performed in the cloud. An example is the innovation in video surveillance. It’s now possible to build highly reliable video surveillance ecosystems with high density local microSD cards with 256GB+ storage, which edge based deep-learning or artificial intelligent frameworks, and algorithms for advanced video analytics.
PREDICTION: Software optimized for flash memory and storage will change the industry.
A fast, low latency device won’t increase performance of the whole IT platform by itself. This struggle with bottlenecks, unused storage and other network and application limitations is one we’ve worked on for years.
In fact, R&D work we’re investigating with our ecosystem partners and customers at Micron Storage Solutions Center in Austin, Texas, has us very optimistic. We see new versions of applications and operating systems that will soon interact with innovations in fast storage-class memory and persistent memory devices in a way that maximizes benefits across the enterprise.
Adding enhanced security throughout IT systems will have little impact on performance, while improving power profiles at a system level, increasing densities for both compute and storage platforms – further reducing operational costs. Placing analytics at different layers in the infrastructure will optimize results and put insights closer to users. These and more are coming. HCI has contributed, NVMe-oF has contributed, our announcement of 32GB DDR4 NVDIMM-N is a product with great potential to exploit in these environments.
This will start (as it always does) with the industries that care about the lowest of low latencies, and having the ability to monetize screaming fast throughput for the benefit of their users, applications and infrastructure needs. Look for how these developments will deliver better experiences, performance, power, and more as they spread to more applications.
The key theme of 2018 will be combining the hardware advancements companies like Micron have made in the last few years with software and data center design advancements that will yield great synergies for industries from automotive to finance. What a great time to be a part of the ongoing digital transformation of business!