When your data center is new, it’s designed to be as efficient as possible when it comes to cooling, capacity, and performance. As you add equipment over time, however, things can become a whole lot messier.
“Every data center,” says Andy Lawrence, vice president of research, data center technologies, at 451 Group, “is initially designed to be as efficient as possible in cooling the IT equipment for a defined amount of energy, to work to 100% of its design capacity, and to work to a level of risk that was agreed at the outset according to business criteria.”
That ideal doesn’t last. Most data centers start out empty and fill, he says, and may never come close to their design efficiency in terms of energy efficiency or other elements. “Meanwhile, decisions about power distribution and cooling, perhaps to reduce energy waste, may increase the risk of downtime.”
Maintain Efficiency and Capacity
To maintain ideal levels of capacity and efficiency over time, you need to understand how your plans and objectives impact the data center. There are metrics to help you do that. PUE, for example, is useful if you’re trying to work out the energy efficiency of your cooling overhead or data center infrastructure overhead, Lawrence says. Coefficient of power, or COP, relates to the efficiency of using electricity to provide cooling, he says.
Metrics such as PUE and COP don’t provide a big-picture view of the data center’s health when it comes to availability, capacity, and efficiency. “Other metrics quantify one aspect of data center performance in isolation of other aspects,” says Sherman Ikemoto, director at Future Facilities. “System-level performance blindness degrades the system.”
Future Facilities worked to overcome those issues in designing the ACE Data Center Performance metric. “Data centers must do three things: protect the IT hardware, support a specified amount of IT hardware (usually expressed in kW or MW of IT power draw), and meet operations budget constraints,” Ikemoto says.
Most organizations distribute responsibility for these objectives across individuals or teams, he says, which makes it difficult for organizations to understand the impact of one team on another or on the data center as a system. “The ACE Score quantifies the impact of facility, real estate, and IT plans on the objectives of others and on the operational intent of the data center system.”
With a view of the entire data center, you can strike a balance between operational and system performance, Ikemoto says. For example, many companies are installing containment to improve PUE. But PUE overlooks the impact of containment on the health of the IT equipment or on the ability to support the intended IT capacity. The ACE Score, however, reveals the impact of these changes on PUE, IT, and data center health simultaneously, enabling operators to greatly improve the cost/benefit assessment of a change like containment, he says.
The metric can help with other situations, such as determining the additional IT load your data center can accommodate, quantifying server availability by predictively modeling power and cooling failure, or determining cooling efficiency by visualizing airflow and temperature.
The Best Data Center
The ACE metric, Lawrence says, “is quite good for a particular situation that a lot of data center managers are in: How do we get the best out of the data center we’ve designed? And how do we do our best to meet those design criteria? As the metric points out, that involves tradeoffs in availability, capacity, or efficiency.”
Lawrence says the ACE metric is helpful for organizations with availability, capacity, or efficiency constraints and for organizations with more than one data center where they’re considering moving or consolidating workloads or considering building another data center.
The ACE metric is a simple way of saying, against these models, we’ll run out of availability, capacity, or efficiency if we make this change, Lawrence says. “You won’t know unless you can model for it and see where you reach the limits.”
Availability: % of IT configuration (in kW or MW) that is connected to sources of redundant power and cooling.
Capacity: % of full IT configuration (in kW or MW), projected from installed IT configuration, that is connected to sources of redundant power and cooling.
Efficiency: DCiE (inverse of PUE).