In today’s environment, IT and infrastructure real-time operating decisions often are made within different areas of an organization without regard for the potential impacts on each other as various control levers are pulled. These actions may even be in direct conflict with the service level agreement (SLA), efficiency efforts, and power usage effectiveness (PUE) requirements, as meaningful data and overall system visibility are difficult to obtain.
The data center is directly dependent on the critical power infrastructure, making it important to balance IT utilization rates to optimize PUE. A poorly designed system can result in unplanned downtime, excessive energy consumption, and constrained growth.
What is needed is a basic framework of approaches and considerations for data center operational best-practices to optimize data center PUE, without conceding availability, in response to IT utilization rates.
Data center design
Data center design and build-out are usually established early in a construction process. For example, a 2 megawatt data center with a power system of 3 x 800 kilowatt UPS as the primary backup and a redundant 800 kW UPS is installed and run from the start, but the IT loads start at near zero and may only be 500 kW to 1 MW at the end of the first year.
Cooling is also sized from the initial setup for the full build-out. Again, a 2 MW IT load design with a conservative 1.6 PUE suggests a cooling plant sized at 3.2 MW at full capacity, likely with a 10% to 20% reserve for redundancy and/or peak demands and changing climate conditions.
Deployment assumptions involve starting in one point of the building (corner, wall/side, center point) and installing IT assets in rows from that point, keeping to the smallest footprint applicable for the types of devices installed, power availability, and cooling.
Best-practice: Try to deploy IT assets in "pods" of contained space with dedicated hot air return plenums to maximize cooling unit performance and overall system efficiency.
Increase utilization rate to improve efficiency
A concern often raised in discussions about redundancy is utilization rate. A 2N UPS system that traditionally has the highest availability will exhibit lower utilization rates and a higher PUE if it’s deployed without intelligent system control. Each bus of a 2N system can only be loaded to 50% maximum so that one bus can provide full load in the event the other bus is not available.
Many business critical data centers use 40% as the peak loading factor on each bus in this configuration to allow for variations in IT power draw and provide a cushion for immediate expansion capability. This level of loading causes the UPS to operate at less than optimal efficiency. Moreover, customers have expressed concern that they don’t trust all UPS suppliers to be able to support 100% load.
Best-practice: Find a UPS supplier you can trust, one that can support an intelligent UPS capable of a broad, flat energy efficiency profile and provide 100% of rated load across the range of high- and low-line conditions. The UPS should also be able to withstand temperatures to 40°C, blocked air filters, fan failures, and various altitudes as robustness insures availability.
Utilize the "intelligence" features to design and deploy a scalable system architecture where individual UPS modules are brought on-line in concert with load growth and rotated on a periodic basis for lifecycle enhancement. Such a system will maximize overall power system efficiency and enhance system reliability and availability while also optimizing the data center PUE.
Manage heat efficiently
The approach data center managers are taking with cooling also is evolving. The current approach of using precision cooling systems to maintain temperature and humidity within a specific range is being replaced by managing heat in the most efficient and effective way possible.
Best-practice: Use of containment to control and increase the supply air temperature for cooling equipment, increasing the efficiency of that equipment and maximizing opportunities for economization.
IT systems being deployed today don’t require the same level of temperature control as previous-generation systems. They can withstand wider temperature ranges, including higher maximum operating temperatures. The American Society of Heating, Refrigeration, and Air Conditioning Engineers (ASHRAE) has recognized this development and raised its limit for the safe operating temperature of computer electronics to 80.5° F, creating new opportunities for heat management.
Data center managers are being challenged to maintain or improve availability in increasingly dense computing environments while reducing costs and augmenting efficiency. Doing your due diligence on the latest UPS technology and efficiency optimization modes will help you choose improved critical power systems with the highest availability and new levels of utilization and efficiency.
Bob Friday, Chief AI Officer for Juniper Networks, explains how the advanced technology is transforming operations.
Contact center leaders from 8x8, Awaken Intelligence, and 360insight discuss the importance of agent experience.
AI may force enterprises to rewire parts of their data centers so they are fully optimized to run such workloads. The question is do you use Ethernet or InfiniBand?
