2017 Trends Reflect Faster, Better, Cheaper, Safer, Mantra Of Hyperscale
As clients make a shift from on-premises traditional data centers to the cloud, mega vendors will rush to build out their hyperscale data centers. Hyperscale data centers use nodes to flex up and flex down on compute power, storage, networking, and memory as demand increase or decreases. While only 6 to 7% of the world’s workloads are in the cloud today, Constellation estimates that by 2020, 67% of the world’s workloads will be processed by cloud data centers. Moreover, the growth in public cloud continues to exceed the growth in private cloud due to operational efficiency, improved security, and improvements in hyperscale. Constellation predicts by 2020 305 million (66.8%) of the cloud workloads will shift to the public cloud with 151 million (33.2%) in the private cloud data centers.
Constellation sees the following trends in 2017 for the data center:
- Expect higher power density for the power plants for the new economy. With AI in the cloud driving workloads and demand, clients will buy compute power by the Kw/H. These hyperscale data centers have got to ramp up 10 to 100X in compute power and efficiency. The goal is to improve server side usage from 15 to 17% to 40 to 50%. Network functions virtualization (NFV) will improve this. The goal is to drive down PUE sizes and quickly flatten data center architectures. Workload density will move beyond the aspirational 3.0 workloads per server benchmark by 2018.
- Push towards more green, more cold. The quest for efficiency continues with data center construction in colder locations and zones. Where possible, new construction requires more green power, more underground design, and higher efficiency of systems. A lot of engineering is going into the design of more efficient connections, better use of power, and more importantly distribution efficiency. The biggest opportunity will be in the smaller data centers, not the hyper scale ones. Constellation estimates almost 40 billion kw/hrs to shave off from the legacy on-premises data centers alone.
- Increase security at the chip level and in the physical perimeter. Given the criticallity of the data center, organizations must up both their physical security as well as their network security. Expect an increase in more physical security systems. Importantly, security at the chip level is helping to improve the scale of defense required in sophisticated hacking and reducing the impact of inside jobs.
- Greater use of DCIM to benchmark performance. The result is anywhere from two to three times less down time. Resiliency and high availability take center stage in improving up time. The convergence of logical and physical layers through SDDC should improve in maturity.
- Continued adoption of OCP. With Facebook leading the way in the open compute project (OCP) and industry standardization, Constellation expects better inter-operability over time across all areas of the data center. More importantly, how organizations bridge the gap in hybrid data centers will exponentially improve efficiency. This industry shift enables on-premises and cloud environments to work well together.
Understand the cloud and AI requirements for your data center strategy? Ready to ditch yours and jump to Amazon, Google, IBM, Microsoft, or Oracle? Let us know what your experiences have been and feel free to reach out. Add your comments to the blog or reach me via email: R (at) ConstellationR (dot) com or R (at) SoftwareInsider (dot) org.
(Cross-posted @ A Software Insider's Point of View)