Home > Data Center Efficiency > DCEP: Data Center Energy Productivity

DCEP: Data Center Energy Productivity


Since the 2007 EPA Report to Congress on Server and Data Center Energy Efficiency was published, much has been written on improving energy efficiency in critical facilities. Many of the conversations, until now, have addressed efficiency without productivity. While PUE and DCiE focus primarily on infrastructure- the mechanical and electrical systems that support the data center-are they enough to brand an entire data center efficient?

A comprehensive efficiency grade must include the productivity of the data center-the “useful work” performed by IT equipment. A new metric called Data Center energy Productivity (DCEP) incorporates both the infrastructure and the IT equipment when evaluating data center energy efficiency.

What is DCEP?

DCEP is a sophisticated metric that quantifies useful work compared to the energy it requires. It can be calculated for an individual IT device or a cluster of computing equipment.

DCEP = Useful Work Produced / Total Data Center Energy Consumed over time

Like PUE and DCiE, time is important to this measurement. DCEP has an “assessment window” where useful work and energy are compared relative to a user-defined time limit.

Calculating DCEP

The work of a data center is unique to its operator. For this reason, it’s difficult to prescribe a universal productivity metric. DCEP itself is a work in progress, and the Green Grid is currently evaluating several alternatives to measure useful work in different ways. Both DCEP and these proxies may require functionality or instrumentation not included in off-the-shelf IT hardware.

Useful Work

Useful work depends on two readings: the tasks performed by the hardware and the assessment window. Tasks should be as specific as possible, while the assessment window “should be no shorter than about 20 times the mean run time of any of the tasks initiated in the assessment window” (The Green Grid, 2008). A data center should define both figures according to their workload and business model.

The Green Grid proposes the following equation for useful work:

(Sum of all tasks * Value of the task) * Time Based Utility Function * Absolute Time of Completion

This equation acknowledges that some tasks are more important than others. Some may be mission critical (task value) while others may involve a response time that’s an integral part of an SLA (time-based utility function).

To simplify the calculation, the task value and utility function can be assigned as (1), meaning that all tasks are weighted the same. From there, useful work boils down to the number of tasks (jobs, transactions) carried out by the hardware during the assessment window.

Total Data Center Energy Consumed

This calculation involves two figures: the kWh of the hardware in question and the PUE of the facility. If the hardware and infrastructure are efficient, this reading will improve.

Total Energy Consumed = kWh of energy consumed by hardware * data center PUE

To arrive at this answer, a facility must have an efficiency benchmarking program for PUE/ DCiE and the ability to measure power at the device level. While not all data centers have this capability, new software, the latest rack power products, and power meters can all track power consumption at the device level. This consumption in kWh is multiplied by the PUE figure to arrive at the total energy consumed during the assessment window

DCEP Going Forward

DCEP represents a necessary progression in metrics, for the efficiency conversation must include productivity.

By measuring PUE/DCiE consistently, users can identify and remediate infrastructure deficiencies through sealing cable cutouts, tuning hot/cold aisles, and adjusting temperatures. Small improvements in physical infrastructure go a long way, and users should begin the process with the initial PUE/DCiE benchmark.

With these improvements in place, users can then focus on the output of their data center and the productivity of their compute systems. This process may open doors to consolidation, virtualization, and the decommissioning of idle, older compute platforms.

The mantra of the moment is right-sizing. Calculating DCEP allows users to right-size virtual and physical infrastructures to support business needs.

Measure your PUE