One of the keys to a properly functioning data center with high availability is ensuring that clean, steady power reaches the IT equipment. Part of this critical task falls to uninterruptible power supplies and backup generators, but this power must be distributed effectively and safely. Breakers protect branch circuits from overloading, but this safety precaution can become a tremendous hassle is the circuit isn’t loaded properly. Furthermore, increased energy efficiency can be difficult to achieve without knowledge of how circuits perform over time. To address these and other concerns, branch circuit monitoring systems (BCMSs) can give data center managers the information they need to identify potential problems, avoid branch circuit overloading and identify candidates for efficiency improvements.
Why Monitor at the Branch Circuit Level?
Tracking overall data center power consumption along with overall IT power consumption is sufficient for calculating the power usage effectiveness (PUE) metric, but it lacks the granularity required to know what’s happening on individual branch circuits. Tracking power consumption down to the level of individual components or even power strips may be impractical, however, but monitoring power at the branch circuit level, although slightly less granular, can yield information that enables the data center manager to identify problems and avoid downtime.
Consider, for instance, the circuit breaker on a branch circuit. These devices are designed to ensure that a circuit is not overloaded, creating a potential hazard, by breaking the circuit when a certain maximum current is reached. When this happens, equipment connected to the circuit loses power. The problem with circuit breakers is they are unforgiving: they don’t care if the heavy power load is at a time of day when downtime could really hurt business.
Part of the difficulty with circuit breakers is that IT power loads can vary significantly over time. Relying exclusively on the manufacturer’s power ratings for equipment may leave a margin of safety when loading a circuit (i.e., the circuit load is the sum of the equipment ratings, and that sum does not exceed the maximum power of the circuit), but it can also leave the circuit’s capacity vastly underutilized. If a circuit is loaded with servers that, for instance, never reach half their rated power levels, that circuit will never exceed half its capacity. Data centers, thus, may often “overload” the circuit (in the sense that if every piece of equipment on the circuit reached maximum power consumption, the circuit capacity would be exceeded) to avoid wasting costly infrastructure. But as computational loads vary, this overloading can lead to a “real” overload, tripping the circuit breaker. Such an occurrence can easily correspond with heavy usage of the data center—a terrible time for an outage!
Furthermore, changes in the configuration of a data center, even if tracked carefully, can still increase the chances of overloading—or result in greater underutilization of circuit capacity. In a facility that is growing or making regular upgrades or adjustments, the number of changes can easily cause problems in this regard.
The problem is thus a lack of information: without measurements of power usage on branch circuits, data center managers can’t be sure that the load on the circuit isn’t too much. When connecting equipment, data center managers must choose an appropriate estimate of the power consumption of that equipment—usually some kind of average consumption. But an average doesn’t take into account the effect of power spikes, particularly if several pieces of equipment all exceed their average consumptions all at once. An Eaton white paper (“Who tripped the circuit?”) clearly summarizes this situation: “Under normal operating conditions, data center managers could mistakenly assume there’s plenty of capacity on a branch breaker or panel board. After all, day-to-day power consumption seems well within limits. But there’s always the potential for one or more servers to increase computational load, draw more power to match the workload, and cause a branch breaker to trip. With the increased use of denser, more space-saving equipment, the risk of overloading a branch breaker is higher than ever.”
Benefits of BCMS for the Data Center
This situation makes clear some of the benefits of a branch circuit monitoring. But simply taking periodic measurements of a circuit (e.g., with a handheld current meter) cannot detect the short-lived surges in power consumption that can still trip a breaker. A branch circuit monitoring system (BCMS) takes (nearly) continuous readings of current in each connected branch circuit, collecting the data and presenting it to the data center manager for analysis and, if necessary, action. By looking at power consumption trends over the course of a day or many days, the data center manager can identify loads that are in danger of tripping a breaker (or that significantly underutilize the circuit’s capacity).
In discussing the benefits of its Powerware Energy Management System for the data center, Eaton notes some of the benefits of tracking power consumption at the branch circuit level: “With this information, you always know how close a circuit is to exceeding its overall rating—and whether or not a device can be added to a branch circuit or panelboard. This insight enables you to operate at maximum efficiency, using data center assets, energy and space wisely. Without this insight, you wouldn’t know if you’re pushing the limits of a breaker or any safety factor.”
A good BCMS, like many other types of data center infrastructure management (DCIM) products, provides both local and remote access to collected data and information. A centralized interface enables the data center manager to monitor and review branch circuit data from anywhere—inside the data center or beyond. BCMSs often include data analysis tools as well, providing trending and information to aid the data center manager in getting a comprehensive power picture.
In addition to helping achieve optimal loading of branch circuits, a BCMS can also provide information that the data center manager can use to increase power efficiency. Examining power consumption at the branch circuit level—rather than just the facility level—enables better isolation of equipment that is underutilized or inefficient, aiding in equipment consolidation and other efficiency improvement projects.
Conclusions
Many data center best practices are justified by the low capital and operational costs of added infrastructure relative to the massive business costs of data center downtime. As data center managers seek to do more with less, underutilized and inefficient branch circuits are a drag on an optimally running facility, but overloading can result in downtime when usage is at its highest—a dreadful proposition for the business. Branch circuit monitoring through a centralized BCMS can enable better utilization of capacity while avoiding too close an approach to maximum circuit capacities, which, if exceeded, result in tripped breakers and thus downtime. Furthermore, continuous power monitoring at the branch circuit level allows the data center staff to identify potential problems and inefficiencies. Although some operational and maintenance costs are associated with a well-deployed BCMS, the benefits—particularly relative to the costs of downtime—can quickly outweigh these costs for many data centers.
Photo courtesy of Tom Raftery
RAGHU KONDAPALLI LSI Corp.
Click to enlarge chart.
