Five Ways to Save Energy in the Data Center

Posted by James Ball on Oct 18, 2018 9:00:00 AM

It’s no secret that data centers can consume a lot of energy. In fact, they will be consuming one-fifth of the global power by 2025 and they will represent 40 percent of the carbon emissions by 2040. When framed in this context, saving energy in data centers isn’t only about saving money, it’s also about saving the planet.

AdobeStock_102020317

Adobe

Ditch High Power Bills with These Tips         

Data center managers can make a huge difference. Here are some energy-saving options to consider:

 

Use Well-Established Cooling Processes

Heat is a big culprit. It makes servers run less efficiently, which ends up wasting energy and generating more heat. It's a downward spiral, meaning data center managers need to take precautions to keep their facilities cool.

There are well-known techniques such as hot/cold aisle containment, evaporative cooling (EC), and liquid immersion cooling. The use of these techniques will depend on your facilities structure and dynamics. However, the right techniques can keep your energy bill low and improve the performance of the servers. Check out the American Society of Heating, Refrigerating, and Air-Conditioning Engineers (ASHRAE) guidelines for data centers for a detailed explanation of this.

 

Optimize Equipment Power Usage

Data centers grow organically with new servers being added as demands grow. Sometimes servers are put in for ad hoc purposes. The problem with this organic growth is that units are not optimized.

As a result, servers are often underutilized or even idle. The problem is that an idle server doesn’t mean there is no power usage. An idle x86 server, for instance, can consume 30 to 40 percent of the maximum power capability, even though it isn’t doing anything productive (see also: zombie servers). Finding the idle and underutilized servers and simply switching them off can significantly decrease data center power usage.

Another simple trick that can help is using the CPU power-management feature. Almost 50 percent or more power on a server is used by the central processing unit (CPU). Today most servers have CPU power-management capability, which might slightly affect performance, but only to a negligible degree. 

  

Use High-Efficiency Power Supplies and UPS

Inefficient Power Supply Units (PSUs) and Uninterrupted Power Supply units (UPS) both contribute to energy loss in their own way.

Typical PSUs that convert alternating current (AC) to direct current (DC) operate at 80 percent efficiency. But the voltage regulator can further decrease the efficiency of the PSU. Look for ENERGY STAR® PSUs and voltage regulators to get better performance.

UPS units store power to guard against power interruptions. A UPS system would typically reach its greatest efficiency at approximately 80 percent utilization, with its worst efficiency at utilization less than 50 percent. Common UPS architecture is 2N redundant where the utilization is less than 40 percent under normal operating modes. Thus, it is important to review efficiency ratings across the entire utilization band.

Most UPS units provide efficiency information for full loads. But they never operate in full load capacity. So, track UPS power usage for partial loads and see the kind of efficiencies you expect to actually get in real world operating scenarios.  

 

Run Data Center Energy Audits Regularly

Data center managers need to know the current power use patterns to make smarter decisions. Because data centers are constantly in flux, the best way to keep track of infrastructure efficiency is to run audits.

Audits will help you track any anomalies in the infrastructure or electrical usage. If you maintain a regular audit history of your equipment, it’s easier to figure out if the new power usage spike is coming from the cooling system changes or the server rack reorganization. It also provides tracking granularity.

Overall, data center managers can significantly increase energy savings through process improvements, optimizations and regular audits. And, if you're unsure where to start, contact Instor's team of data center infrastructure experts to help get you back on track. 

 

Seek Out Low- and No-Cost Efficiency Gains

­­Often, data centers are built to operate up to a specific maximum load. Stranded capacity can result from differences in load characteristics (i.e. higher or lower kW/rack density than designed). This can lead to wasted energy and increased energy costs.

In a multi-tenant, pre-fabricated space, your decisions about UPS, PDU and lighting infrastructure have often already been made for you. Still, there are opportunities to maximize performance with the following practices, which we dub “low-cost / no-cost” techniques: 

  1. Balance power distribution loads at the rPDU (rack PDU), RPP (remote power panel) and PDU level. By using rPDU’s with color coded circuits to facilitate balanced load distribution within the rack, the electrical infrastructure will operate more efficiently while improving reliability. 
  1. Balance airflow with power consumption at the row/pod level. Airflow generally can be reduced from 160+ cfm to 125 cfm with proper containment and airflow management techniques, reducing fan power requirements and increasing efficiency of the cooling infrastructure.

Note: not all “high density” racks need a high airflow grate in front of them nor does every rack need an airflow tile. In fact, over -provisioning of airflow tiles (high or conventional 25 percent perforated) leads to lower static floor pressures and inefficient operation to the point where there is insufficient flow where it is truly needed.

Often, initial deployments are 25 percent of projected full capacity. Not all CRAC/CRAH units are necessary to provide sufficient cooling. For this reason, idle units should be put in “Standby Mode.” Meanwhile, over-provisioned airflow tiles can be masked off or covered with rollout rubber floor mats. Be diligent in keeping containment doors closed and replacing open U space with blanking panels. Remove airflow tiles from hot aisles and other areas where they are unnecessary. Observe CRAC/CRAH units for demand fighting (i.e. units in “heating mode” while others are cooling.)

Finally, scan rack inlet temperatures (if an environmental monitoring system isn’t present) on a regular basis to ensure that rack inlet temperatures are within ranges recommended by ASHRAE and equipment manufacturers.

Follow these best practices and you’ll be well on your way to lower power bills.

References: 

 

 

 

Topics: Data Center Power