TECHNOLOGY IN SYSTEMS
You Want to Sleeeep . . . Managing Power Consumption
Intelligent Power Consumption and the Data Center
When it comes to data center (DC) technology, gaining visibility into power consumption is a critical part of the business case for any provider, given that power is often the biggest expense. With proper planning and device selection, the integration of the thousands of
BY JON TROUT, TRENDPOINT SYSTEMS
Page 1 of 1
According to the U.S. Department of Energy, data centers can consume 100 times more energy than a standard office building, and approximately three percent of all electricity in the U.S. is used to power them. It is estimated that DC power consumption will only continue to rise.
Providing energy, cooling and maintaining the right temperature for rack equipment to run optimally are all parts of the equation; and misjudging capacity planning for energy usage also results in wasted money, wasted energy and lost opportunities for operational optimization. Additionally, a lack of accuracy in the metrics provided can result in downtime that can cost organizations millions in hard dollars as well as brand equity.
The solution we often hear from data center managers is their vision of a “Single Pane of Glass” to manage their facility. The “Single Pane” concept has fueled the explosion of the Data Center infrastructure Management (DCIM) market with the need for one system to have the ability to manage and control the entire facility. Typically, the power measurement and monitoring aspect of the data center consumes the majority of the data points that build the DCIM system. The power infrastructure is of utmost importance to the reliability and efficiency of the data center. Power monitoring combined with a DCIM system can provide insight into the various aspects of the facility. The DCIM software is the bridge that collects the raw data from the devices in the facility and translates the data into “actionable intelligence.”
Capacity planning is one example of how power monitoring, combined with a software system, can assist in data center operations. The software can alert and alarm on capacity issues that may impact the redundancy of the facility. Most facilities use A and B sources of power distributed to the server racks. If the A source and B source are both running at 70% and the B source is lost due to an outage or failure, the A source will not have the ability to independently carry the entire server load. The software system can also identify areas in the facility that have excess capacity, enabling server loads to be redeployed, physically or virtually, to these locations. The DCIM system is the watchdog, constantly monitoring these types of conditions and delivering alerts when abnormalities are observed.
In addition to capacity management and reliability, data centers are also concerned about efficiency. We hear debate around the Power Usage Effectiveness (PUE) metric and whether the PUE calculation is the best way to measure efficiency. While the PUE metric may have its critics, it still provides a benchmark for facilities to compare their efficiency with others in the industry. However, the PUE metric should be derived from a granular point in the power distribution chain. Losses in the Universal Power Supply (UPS), transformers and distribution chain should be taken into account when measuring the efficiency of the facility. The workload is performed at the server level, so losses and wasted energy upstream should be monitored and considered when looking for areas to improve efficiency. Power monitoring through the power distribution chain can be analyzed in the DCIM or software system to calculate the true efficiency of a facility (Figure 1). Areas that are lacking the expected efficiencies can be seen easily through the software system and action can be taken to make improvements.
Display showing data center infrastructure efficiency along with power usage efficiency.
Avoiding the Pitfalls
Typically power is monitored at every stage of the distribution chain inside the facility to properly manage the reliability and efficiency of the data center. Utility feeds, generators, switchgear, distribution panels, power distribution units (PDUs) and remote power panels (RPPs) are all monitored so facility managers and operators can manage their infrastructure at every stage of the system (Figure 2). A comprehensive power monitoring solution will consist of many different types of meters. The integration of these devices into the DCIM system can be a costly and time-consuming project. Multiple communication protocols, onboard configuration, logging and alarming are essential. These features need to be applicable to the business requirements of optimization, speed of information and speed of deployment.
A data center requires coordinated monitoring of interacting power systems on a variety of different meters and devices.
It is important to assess the ease of integration of the solution as well as the scalability. This includes both the ability for a system to be flexible enough to be integrated in the correct manner the first time as well as the adaptability of the system to work with new software.
Think Flexibility: With all of the power distribution types (e.g., PDUs, panel boards, busway, etc.) that may be in your one or multiple data centers, and the variety of vendors who supply these products (e.g., Schneider Electric, Siemens, Eaton, GE) that might be present, the power monitoring products you install need to be able to interact effectively with all of them. Choosing a “platform” that covers the spectrum of power distribution types and vendors, as well as various amperage sizes and circuit configurations, will simplify deployment and streamline integration into software systems.
Stay Adaptable: We’re seeing a rise in the number of data centers that use busway power because of the adaptable power distribution system it provides. PDUs and panel boards are also frequently changed and modified to support the dynamic data center environment. Make sure that your power monitoring products can adapt when the power shifts—that way you’ll save money on rebuilding or replacing the existing meters.
Achieve Utility Grade Accuracy: Most data center power meters claim to offer accuracy that is within five percent of the actual power utilization. Although five percent is not insignificant, the best practice in power monitoring is to attain utility-grade accuracy, which is within one percent of the actual amount of power consumed. The reason? A utility-grade level of accuracy enables co-located and other data centers to fairly rebill clients for the cost of energy.
Additionally, the importance of accurately measuring power consumption grows in the aggregate. A 400,000 square-foot data center being off in measurement accuracy for all of the installed equipment significantly impacts the bottom line.
The right data collected from the right points within the facility at a granular level makes it possible to gain an end-to-end view of what’s happening. This enables better efficiency assessments and energy management in general, making it possible to identify specific problem areas and boost optimization.
Knowing exact amperage eliminates tripped breakers; tracking voltage gives visibility to variations in power to prevent trouble; tracking the power factor can prevent inflated power bills; and tracking wattage allows visibility into the heat generated by each circuit (Figure 3). Operators can identify kilowatt hours to track energy by end user and groups.
Measuring the heat generated by each unit in the systems enables quick visibility into anomalous conditions.
Measuring wattage allows users to identify when an anomalous amount of power is going to a group of racks. If usage is off compared to the average, then it’s likely that something needs remediation. This enables root cause analysis of issues, minimizing maintenance and troubleshooting costs.
Avoid Non-standard Protocols: Does your data center use SNMP to communicate between systems? Modbus TCP? BacnetIP? If the power monitoring meters don’t utilize standard communication protocols, it will be more time-consuming and costly to integrate the meters with a DCIM or Building Management System (BMS). The best way to avoid this problem is to make sure that any metering system you purchase has the right communications protocols for the software system you plan to utilize. Ideally, your metering platform should be able to support all your power distribution products, communicate easily with the software, and interact seamlessly with all of the other components of the data center.
Look for Greater Device Functionality: Typical monitoring solutions require a complex and costly network of protocol conversions, middleware and data interpretations to provide the operations and engineering teams with a comprehensive picture of power consumption in the facility. Features such as onboard Ethernet, onboard data logging, onboard alarming and an accessible Web interface can reduce the failure points and cost associated with a monitoring deployment.
Data center power monitoring is not just about reducing energy costs; it is part of a larger conversation that affects the overall operations of your data center’s various systems. Data centers are dynamic environments, where server stacks and power distribution sources might need to be reconfigured based on the center’s changing needs over time. Ideally, your data center monitoring solutions need to connect, interact and integrate with all of the various “moving parts” within your data center—without incurring additional costs or complexities along the way.
Knowing your power consumption at a granular level can be a powerful way to help achieve the overall initiatives in the facility. Selecting the right power monitoring system for your facility’s current and future needs is an important part of the selection process. Implementation and integration of the power monitoring system are steps that can be easily overlooked, but they play a big role in the effectiveness of the facility and the DCIM or BMS systems they complement. Flexibility, adaptability, accuracy, communications and device functionality are important characteristics of a successful power monitoring system.