If 2019 was any example, 2020 will be filled with technology innovation and IT data center transformation. Let’s take a look at our predictions for 2020:
The new private cloud will be created with elastic bare-metal infrastructure
While VM-deployed applications will continue to move to the public cloud, more bare-metal private cloud data centers will be created in 2020 to run performance-sensitive bare-metal and Kubernetes-native applications. This is being driven by the growth and investment in data-intensive apps and the need for scale-out data center architectures.
In 2020, Smart NICs finally come of age
Smart NICs have been used for many years by cloud providers to offload network switching and encapsulation, but in 2020 we’ll see SmartNICs for storage applications and in private data centers. Amazon’s NitroSmartNIC was the first with NVMe storage capabilities, but now both Mellanox and Broadcom have announced similar capabilities in their chips, while high-flying startups Pensando and Fungible are building systems revolving around SmartNIC capabilities. Why SmartNICs now? Aren’t these just “front-end processors”? The most compelling feature of SmartNICs is that they provide a separate security domain that cleanly separates user code from provider code. In these days of CPU vulnerabilities like Spectre and Meltdown, it may not be safe to assume the correctness of any security boundaries within a CPU.
2020 will be the Year of Kubernetes for stateful applications
While people realize they have a lot more work to do to deploy Kubernetes for their data-intensive applications, the benefits of adopting containerization for persistent storage will continue to drive Kubernetes evolution. And while the introduction of the CSI plugin for Kubernetes makes it easier for storage providers to integrate into K8s, user requirements for scale-out application deployment on top of inexpensive, commodity hardware remains a challenge which is just now being solved.
The end of VMs in your data center
Have you heard? The cloud is eating the world. Well, some of the world that is. For virtual machine deployed applications, it’s clear the public cloud is a simple, cost-effective shared infrastructure that works well. And most VM-deployed applications are low in scale and fairly steady-state. We predict that all of these apps move out of the data center to be replaced by data-intensive applications and bare-metal applications where the public cloud is outlandishly expensive.
Composable infrastructure drives growth of NVMe over fabrics
AI/ML, Big Data Analytics and high-speed transaction processing are driving demand for high-performance NVMe flash. For these scale-out, data-intensive applications, NVMe over fabrics is the dominant choice. And the trend is moving toward using standard 100G Ethernet along wIth NVMe/RDMA or NVMe/TCP. Ethernet is well-understood and ubiquitous making it highly cost-effective, and with speeds at 100G today and soon 400G in the data center, proprietary network technologies (aka, fibre channel or infiniband) are rarely required. According to IDC, NVMe flash has already replaced SAS in the data centers. Its advantages are many: low $/IOPS, low power and cooling and significantly reduced data center footprint. We believe composable infrastructure will drive further NVMe over fabrics adoption in 2020 by automating configurations and resources that can be adapted on the fly.
Nothing but Net for NVMe
As NVMe Over Fabrics takes off in the data center, we’ll see the advent of fabric-native storage devices. These are not traditional controller-heavy SAN systems that speak NVMe-oF, but rather storage devices that use leaner, cheaper, and faster chips to create a direct connection from SSD to network. Western Digital, Toshiba, Mellanox, Broadcom, Marvell, and many other companies are all developing products in this area.
It’s Raining Workloads!
In 2020, many more organizations that are building Deep Learning models will discover that cloud economics don’t work for the continuous and voracious requirements that DL training has for computing and storage. Cloud repatriation will be the norm as DL workloads get serious.
Composable infrastructure hits its stride
And will be largely driven by tier-2 public and sizeable private clouds. As Kubernetes ripples through the public cloud providers and enterprises look at the platform for providing a seamless application deployment approach from laptop to cloud, cost efficiencies are being investigated to take advantage of this common application deployment platform. Composable infrastructure integrated seamlessly into Kubernetes provides a clear win by handling the physical platform orchestration for persistent data applications under the logical orchestration provided by Kubernetes all on top of commodity hardware.
Privacy laws make data encryption everywhere a reality
By now, every company has come up with a first response to the EU’s General Data Protection Regulation (GDPR). Implementing it is a journey. While data segregation is part of the strategy, some companies are likely finding that separating data to be encrypted from “regular” data in order to deliver to GDPR requirements is causing more headaches than they planned on. Encryption of all data, both in-flight and at rest, will increasingly become the norm in every data center.
VMware (Dell EMC), Red Hat (IBM) and the Cloud Kings duke it out for Kubernetes domination
At some point, people will look back with fondness at the notion that everything will simply move to the cloud. With Google Cloud in fourth place in adoption (behind Amazon AWS, Microsoft Azure and Alibaba Cloud) things still seem to be in flux. It’s not so much moving to the cloud as moving to which cloud. Of all the infrastructure players under threat, surely VMware is the most exposed, But they are pivoting with recent acquisitions to attempt to take the leadership position with a spate of recent acquisitions, and establish themselves as the leader in multi-cloud and hybrid cloud via the Kubernetes platform. Just to keep us all tuned to the show, IBM closed its acquisition of Red Hat, and is taking on the move to Kubernetes without the baggage of a legacy data center virtualization strategy. The coming year should provide increasing clarity as to the success of each player’s strategy to dominate IT moving forward.