The ability to run a lights-out data center is practical today. The advantages are that people are rarely in the data center, allowing higher operating temperatures of the data center and reduced needs for cooling, thus improving energy efficiency.
Remote management capabilities can be deployed throughout the systems of the facility, from power to HVAC, network to firewall, and physical access to logical access.
Labor costs are minimized as well with a lights-out data center.
Expert Blog by Marc Cram, Director of New Market Development, Server Technology.
Tools are available for automating the software maintenance, the network security, the data deduplication, the operation optimization for energy efficiency, and the workload routing and scheduling. All that is lacking seems to be a robotic system for replacement of failing hardware infrastructure, but that is coming.
Planning for a lights-out data center begins with an understanding of operating objectives, such as:
- Uptime goals versus cost of downtime
- Number of users of functionality of the data center – CDN, AI, cloud
- Efficiency goals for water, energy, et cetera
- Operating philosophy – predictive/preventive maintenance versus reactive, hyperconverged versus disaggregated
- Distance of data center from support personnel – choosing a response time
- Physical security of the infrastructure
Designing a new, cost-effective lights out data center today involves implementing resiliency through software, choosing a design that easily works with Kubernetes or Docker for containerization of workloads. This allows work to move easily around from machine to machine or between data centers. It also minimizes the need for UPS run time when utility power drops.
Intelligent PDUs, DCIM
Consideration should be given to efficiency goals for the data center. Is active cooling needed, or is free air cooling sufficient? Can the data center run without water? Hardware must be designed to support the environment in which it is running. And, low power devices that run without needing fans or liquid cooling pumps can reduce the number of potential points of failure, improving reliability and uptime.
Plan for remote power management solutions in the form of Intelligent PDUs, which provide the capability to securely monitor and control device power through a network connection. It allows for reboot of locked-up devices and power-up sequencing for example. In addition, environmental sensors should be used throughout the data center to continuously measure temperature, humidity, air pressure, smoke, fire, and floor moisture.
Use a DCIM tool to collect the data and prepare that data for analysis by machine learning or AI tools to optimize the operating efficiencies. Software that looks for trends and intelligently alarms based on those trends enables predictive/preventive maintenance to be performed, minimizing unplanned downtime. Automation of basic system maintenance functions such as BIOS and operating system updates, antivirus and anti-malware updates is standard operating procedure in today’s data centers.
Finally, as an alternative, you may want to have a look at many of the larger colocation facilities. They provide many of the remote management tools that enable enterprises to operate colocation-deployed infrastructure to be successfully run without daily physical access.