Best Practices for Data Center Power Monitoring

Best Practices for Data Center Power Monitoring

datacenterpower112013-612x300 One of the more challenging problems facing today’s data center managers is serving the opposing demands of ensuring 7×24 uptime and reducing operating costs. As a result, a growing number of firms are deploying monitoring solutions designed to provide a discrete view of a data center’s power consumption as a means of achieving both objectives. But to be truly useful, any monitoring device needs to enhance rather than hinder the flexibility and adaptability of the facility’s power distribution infrastructure. Fortunately, there are a variety of best practices that data center managers can consider when selecting monitoring components to reduce deployment costs, maximize uptime and maintain the flexibility to distribute power as needed.

Maintain Flexibility Across Your Global Footprint

Despite the increase in processing power at the rack level, the physical space dedicated to data center operations has increased exponentially over the past several years, and data center managers must consider the effects of physical growth when deploying any system. Data center providers are building networked facilities spanning the globe in an effort to reach new customers and create a redundant architecture. As a result, individual operators often deploy various methods of power distribution across their facilities. A data center design in an existing commercial office park may use a slab floor and busway system to reduce build schedules and offer flexibility in the data center, whereas the same provider may operate another facility using a raised floor with freestanding PDUs and RPPs. When deciding on a power monitoring system, it is important to consider all the facilities potentially affected by the implementation. Thus, power monitoring systems should be flexible enough to interface with equipment from any vendor, regardless of the method of power distribution. Limiting the options to monitoring infrastructure provided by a specific vendor and/or a specific type of distribution equipment can result in increased costs and a lot of frustration during the implementation and integration phases of a monitoring deployment. The selection of a monitoring system should be decoupled from the selection of power distribution equipment. The monitoring should be chosen as an overall initiative of your organization, similar to a DCIM or BMS system, and not case by case as equipment is added.

Stay Adaptable Over Time

Metering devices deployed as a part of an overall monitoring solution should be able to keep up with any adds/changes made to the data center’s power distribution infrastructure. Data centers are a dynamic environment where the needs of the facility are driven by constantly changing requirements and new technology from the IT field. As an example, you should not lose the ability to monitor circuits sub-feeding power to another part of your facility just because you increase the size of a few breakers in a panel. Branch circuit monitors (BCMs) that include the capability to support CTs with different amperage thresholds and variable circuit-strappings (one, two and three pole) on the same board give a data center manager the flexibility to distribute power as needed and still maintain the ability to monitor every circuit in the facility.

Reach for a Higher Standard of Accuracy

With the huge increase in colocation facilities incorporating bill-back into their business model, BCMs providing utility-grade, kWh metrics are becoming the standard for data centers looking to reduce power consumption and maximize uptime. Enterprise facilities are also relying heavily on kW and kWh metrics in an effort to manage capacity and control internal cost centers. But utility-grade accuracy (<1% deviation) is about more than looking for ANSI and IEC compliance, as some meters don’t include CT accuracy in their compliance documents. Also, some manufacturers will claim 1% accuracy, but fail to mention the narrow operating conditions required to realize the utility-grade accuracy. A common example is current transformers rated at 10% to 90% of the scale only providing accurate readings of 5–45A on a 50A model. If your environment has circuits below the 5A threshold, you may generate unreliable data. Current transformer linearity is a major culprit of unexpected accuracy variances. It is also common to see power monitoring devices optimized to read a circuit with only a high power factor (>.950). Although this may seem acceptable, you may miss potential problems in your power distribution system owing to inaccurate power measurement across all variables. Inaccurate power monitoring can lead to gaps in capacity, causing an unreliable failover plan. Colocation facilities using a bill-back model can lose revenue or credibility with their customers. You should consider your operating environment and how the accuracy specifications for the monitoring devices fit both your current and long-term needs.

Avoid Proprietary Communication Standards

One of the more consistent trends in the network computing space over the past couple of decades is the continued drive toward communication standards across application, presentation and transport infrastructure (think of the OSI model). And the benefits of this open-standard philosophy certainly apply to the facilities side as well. Because switch gear, PDU’s, busway and power panels tend to be deployed at various times in the lifecycle of the facility, it is critical to build a monitoring solution around a metering platform that is based on industry-standard protocols such as Modbus TCP, SNMP and BACnetIP. Proprietary communication standards associated with metering devices integrated into electrical distribution equipment can create dark points in a power monitoring solution and drastically reduce the ROI of a BMS, DCIM or power monitoring application.

Look for Capability at the Device Level

From a hardware perspective, a highly functioning BCM does more than deliver a robust data set of power metrics. Devices requiring gateways for protocol conversion and middleware to reduce network traffic increase complexity and create unnecessary single points of failure in your monitoring infrastructure. Power meters that support on-board Ethernet can simplify software integration, and those supporting on-board data logging can drastically reduce traffic across your network. A few devices on the market allow power metrics to be stored on the device, reducing the burden of the network layer to feed the data to one or more required servers in the facility. Network outages are also less costly, as the power data trends can be recovered from the data stored at the device level. More capability at the device level can greatly reduce the cost and increase the ROI of a power monitoring system. Power monitoring solutions can significantly help data center managers as they strive to maintain 7×24 availability and drive operational efficiency. And with the cost of power and density of power in critical facilities both on the rise, the demand for granular power metrics has never been more acute. Adding monitoring hardware and software should not complicate the deployment of future power distributio
n infrastructure. By choosing flexible, adaptable, accurate and highly functional devices supporting a wide range of industry-standard protocols and power distribution equipment, data center managers can deploy a scalable monitoring solution and maintain the ability respond to ever changing power distribution requirements of a critical facility.

%d bloggers like this: