Blurred Boundaries: Hidden Data Center Savings

Every disruptive technology in the data center forces IT teams to rethink the related practices and approaches. Virtualization, for example, led to new resource provisioning practices and service delivery models.

Cloud technologies and services are driving similar change. Data center managers have many choices for service delivery, and workloads can be more easily shifted between the available compute resources distributed across both private and public data centers.

Among the benefits stemming from this agility, new approaches for lowering data center energy costs have many organizations considering cloud alternatives.

Shifting Workloads to Lower Energy Costs

Every data center service and resource has an associated power and cooling cost. Energy, therefore, should be a factor in capacity planning and service deployment decisions. But many companies do not leverage all of the energy-related data available to them – and without this knowledge, it’s challenging to make sense of information being generated by servers, power distribution, airflow and cooling units and other smart equipment.

That’s why holistic energy management is essential to optimizing power usage across the data center. IT and facilities can rely more on user-friendly consoles to gain a complete picture of the patterns that correlate workloads and activity levels to power consumption and dissipated heat like graphical thermal and power maps of the data center. Specific services and workloads can also be profiled, and logged data helps build a historical database to establish and analyze temperature patterns. Having one cohesive view of energy consumption also reduces the need to rely on less accurate theoretical models, manufacturer specifications or manual measurements that are time consuming and quickly out of date.

A Case for Cloud Computing

This makes the case for cloud computing as a means to manage energy costs. Knowing how workload shifting will decrease the energy requirements for one site and increase them for another makes it possible to factor in the different utility rates and implement the most energy-efficient scheduling. Within a private cloud, workloads can be mapped to available resources at the location with the lowest energy rates at the time of the service request. Public cloud services can be considered, with the cost comparison taking into account the change to the in-house energy costs.

From a technology standpoint, any company can achieve this level of visibility and use it to take advantage of the cheapest energy rates for the various data center sites. Almost every data center is tied to at least one other site for disaster recovery, and distributed data centers are common for a variety of reasons. Add to this scenario all of the domestic and offshore regions where Infrastructure-as-a-Service is booming, and businesses have the opportunity to tap into global compute resources that leverage lower-cost power and in areas where infrastructure providers can pass through cost savings from government subsidies.

Other Benefits of Fine-Grained Visibility

For the workloads that remain in the company’s data centers, increased visibility also arms data center managers with knowledge that can drive down the associated energy costs. Energy management solutions, especially those that include at-a-glance dashboards, make it easy to identify idle servers. Since these servers still draw approximately 60 percent of their maximum power requirements, identifying them can help adjust server provisioning and workload balancing to drive up utilization.

Hot spots can also be identified. Knowing which servers or racks are consistently running hot can allow adjustments to the airflow handlers, cooling systems, or workloads to bring the temperature down before any equipment is damaged or services disrupted.

Visibility of the thermal patterns can be put to use for adjusting the ambient temperature in a data center. Every degree that temperature is raised equates to a significant reduction in cooling costs. Therefore, many data centers operate at higher ambient temperatures today, especially since modern data center equipment providers warrant equipment for operation at the higher temperatures.

Some of the same energy management solutions that boost visibility also provide a range of control features. Thresholds can be set to trigger notification and corrective actions in the event of power spikes, and can even help identify the systems that will be at greatest risk in the event of a spike. Those servers operating near their power and temperature limits can be proactively adjusted, and configured with built-in protection such as power capping.

Power capping can also provide a foundation for priority-based energy allocations. The capability protects mission-critical services, and can also extend battery life during outages. Based on knowledge extracted from historical power data, capping can be implemented in tandem with dynamic adjustments to server performance. Lowering clock speeds can be an effective way to lower energy consumption, and can yield measurable energy savings while minimizing or eliminating any discernable degradation of service levels.

Documented use cases for real-time feedback and control features such as thresholds and power capping prove that fine-grained energy management can yield significant cost reductions. Typical savings of 15 to 20 percent of the utility budget have been measured in numerous data centers that have introduced energy and temperature monitoring and control.

Understand and Utilize Energy Profiles

As the next step in the journey that began with virtualization, cloud computing is delivering on the promises for more data center agility, centralized management that lowers operating expenses, and cost-effectively meeting the needs for very fast-changing businesses.

With an intelligent energy management platform, the cloud also positions data center managers to more cost-effectively assign workloads to leverage lower utility rates in various locations. As energy prices remain at historically high levels, with no relief in sight, this provides a very compelling incentive for building out internal clouds or starting to move some services out to public clouds.

Every increase in data center agility, whether from earlier advances such as virtualization or the latest cloud innovations, emphasizes the need to understand and utilize energy profiles within the data center. Ignoring the energy component of the overall cost can hide a significant operating expense from the decision-making process.