There’s increasing pressure to manage our data centers efficiently from an energy perspective. Using electrical equipment generates heat requiring more cooling and so the cycle flows. Moreover, the threat of global warming requires we act responsibly. We share three practical steps for improving data center efficiency to keep moving in the right direction.
The accountant will certainly be pleased if we do so. Maggie Shillington, senior research analyst at IHS Markit believes average data center power usage could reach 3,670 kilowatts by end 2019. The average cost of electricity fluctuates widely. However, no matter where you are there’s money to be saved by using it wisely.
First Principles for Improving Data Center Efficiency
You don’t always need to keep growing. It may be a false economy to keep adding racks and cooling to accommodate new systems. Intelligent data center management suggests three things we can do within existing infrastructure. Briefly, these things are:
- Monitor power usage in the system
- Identify and eliminate hot spots
- Consolidate stranded power capacity
Step One: Monitoring Equipment at Device Level
The quest for improving data center efficiency begins with collecting data concerning power usage at device, rack, and cabinet level. This information will enable us to improve energy efficiency by refining capacity and supply.
We can take this a stage further by introducing power distribution strips (PDU’s) enabling us to do power metering in critical racks. These also help improve power quality and execute intelligent load balancing.
These and other data infrastructure tools provide support at several levels:
- They supply useful inputs for reducing energy consumption
- They help us decide optimum locations for additional servers
- We can monitor overall power utilization and break this down
Some IT teams avoid installing these intelligent data infrastructure tools because of IPv4 address prices touching $18 per IP address, and as much as $500 per port total. However, intelligent PDU’s allow for address consolidation. This allows “up to 32 PDUs sharing one or two IP addresses with failover capability” according to Accu-Tech.
Step 2: Improving Data Center Efficiency by Avoiding Hotspot’s
With our power supply system optimized we now turn attention to their troublesome byproducts, heat and humidity. The solution often lies in prioritizing airflow to hotspots, as opposed to increasing overall cooling capacity.
IT teams should therefore focus additional cool air where heat and humidity limit rack capacity and density. Containment strategies such as filler panels etc. assist in maintaining cooling effectiveness. However, Search Networking adds we also need to separate cooling intake and hot exhaust air.
This does not have to be beyond the capabilities of an IT team supported by a competent contractor. The American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE) published a data center standard aimed at encouraging innovation while offering ways to help ensure energy savings.
Moreover, intelligent sensors atop cabinets can monitor zone temperatures against (ASHRAE) recommendations. However there is still one more step to take along the road to improving data center efficiency and that’s dealing with stranded power capacity.
Step 3: Dealing with Stranded Capacity
Stranded capacity is potential we have acquired but are unable to use. A power utility may have to phase out a working coal power station because it is causing urban pollution in a new suburb. A data center may have under-utilized servers because the software never rolled out completely.
An IT team needs to track down these servers because they are consuming power while not delivering to capacity. Their first step should be locating these ‘ghosts’ by analyzing their CPU usage reports. Then they can start to rationalize what they have.
Intelligent monitoring devices including PDU’s add value by helping IT take control over power consumption at this stage. They could, for example:
- Provide intelligent reports about available capacity
- Add value to decisions whether to increase capacity
- Provide valuable inputs regarding server consolidation
- Help balance the workload across the data center
A data center manager can calculate their total center capacity by applying assumed peak power consumption per server, and by rack. However, this method perpetuates stranded capacity because it does not account for variability of load on servers.
Towards a Better Solution for Predicting Power and Cooling Requirements
Neil Rasmussen, writing for Schneider Electric White Papers defines data center physical infrastructure capacity management as “the action or process for ensuring power, cooling, and space are provided efficiently at the right time, and in the right amount to support IT loads and processes”.
He stresses the need for an inclusive approach to avoid power and cooling infra-structure problems, including overheating, overloads, and loss of redundancy. It therefore follows we need to predict power and cooling requirements at enclosed rack level.
Neil Rasmussen’s systems approach holds that stranded capacity is the consequence of one or more of the following factors:
- Floor and rack space
- Power supply
- Power distribution
- Cooling capacity
- Cooling distribution
A particular device needs enough of all these factors to function optimally. However, by the law of averages there will always be unused areas with idle power, and racks with insufficient cooling. In practice, stranded capacity is usually down to one of these situations:
- Sufficing cooling, but with inadequate distribution
- Unoccupied floor space without spare power
- Air conditioners in the wrong positions
- Some areas too hot, while some are too cold
- Sub optimal use of power distribution strips
It’s often possible to reallocate stranded capacity as opposed to investing in new resources. At other times, small investments may be necessary by way of improving data center efficiency. There’s also a cost associated with stranded capacity. Simply doing nothing fritters away our capital return.
Conclusion: 3 Steps Along the Road to Improving Data Center Efficiency
Optimizing data center efficiency is a journey of continuous improvement. Monitoring power usage in the system, identifying and eliminating hot spots, and consolidating stranded power capacity all provide a useful framework for moving ahead.