The growth of processing-intense applications in distributed environments makes power availability a top priority.
By: Dave Dimlich
President of SD3IT
The availability of power has always been an issue in deployed environments, but never more than now. As organizations place more artificial intelligence, sensors and processing capabilities in remote locations, they need to manage an increasing amount of energy use without sacrificing performance or resilience. And before they get to processing power and integrated connections, they first need to ensure the availability of electricity, mirroring a concern that is taking shape throughout the industry.
The growth of AI and resource-intensive inference is driving an overall increase in power usage regardless of location. The Department of Energy predicts that energy demand in the United States will grow by 15% to 20% over the next decade, citing AI and data center expansion, which is being pursued to accommodate AI, as major contributors to that growth. When broken down to data centers, the projected growth is fairly astounding.
The Electric Power Research Institute (EPRI) estimates that data centers will consume 9% to 17% of U.S. electricity by 2030, up from 4% to 5% today. The apparent reason for such a wide range in the projections is that the growth is happening at such a fast rate that it’s hard to predict precisely—EPRI notes that its current estimates are 60% higher than they were a mere two years ago. (In Virginia, the only place where data centers currently consume more than 20% of the state’s electricity, data centers could account for as much as 57% of electricity use by 2030, according to EPRI.)
Although data centers themselves will consume the lion’s share of that energy, power is also increasingly in demand at the edge, particularly when AI is involved. Edge deployments by military, industrial or other enterprises may traditionally have involved mostly low-powered devices performing relatively simple tasks. But that has changed dramatically with the growth of cloud capabilities, the need for on-site processing and the use of AI.
At SD3IT, I see energy management becoming an increasingly important part of how organizations design and deploy distributed edge infrastructure. The systems that support AI, sensors, communications and real-time decision-making must be engineered to operate within the physical constraints of their environments. That requires an integrated approach to computing, power, cooling, networking and lifecycle management.
The Growing Demand for Power at the Edge
Local processing can reduce dependence on distant data centers, limit the amount of information that must travel across a network and support faster decisions when connectivity is constrained. As I discussed in an earlier blog, resilient networking and localized computing are essential when organizations must continue operating without reliable access to the cloud. Organizations distributing computing across more locations must account for the availability and quality of power at each site. A remote facility, industrial installation, transportation hub or tactical environment may not have access to the electrical capacity, cooling infrastructure or on-site technical support available in a conventional data center.
This has given rise to edge data centers, which put computing resources closer to where data is generated, processed and used. In the case of AI, which typically demands a substantial amount of processing, it enables the delivery of services at the speed users have come to expect.
The challenge for enterprises that are making AI a primary component of operations is ensuring the availability of the energy they need at distributed locations that lack the electrical infrastructure, cooling systems and dedicated operational teams that centralized data centers provide. Edge deployments often involve small enclosures, remote facilities, vehicles or locations exposed to dust, moisture, temperature extremes and unreliable electrical service. Physical space is limited, and sending technicians to troubleshoot or replace equipment can be expensive and time-consuming.
AI adds another layer of complexity. Processing workloads can fluctuate, creating changes in power consumption and heat generation. Performance could decline if a system can’t dissipate that heat effectively, and inadequate power distribution could cause a shutdown. Those failures can interrupt industrial processes, disrupt communications, delay decisions or prevent mission-critical applications from functioning when they are needed most.
Gain Visibility Into Power Consumption
A good place to start is by understanding how and where energy is being used. Distributed infrastructure requires consistent monitoring of power consumption, temperature, cooling performance and backup power systems. Data Center Infrastructure Management (DCIM) tools can provide visibility into these conditions, helping operators identify inefficiencies, anticipate capacity constraints and make informed decisions about equipment and workloads.
This is particularly important when an organization manages dozens or hundreds of remote sites. A small inefficiency at one location may seem insignificant. Across a distributed footprint, however, unnecessary power consumption and cooling requirements can add up quickly.
Monitoring also supports better capacity planning. Operators can identify underused resources, determine whether power and cooling systems are appropriately sized and recognize potential problems before they affect operations.
Match Computing Resources to Operational Needs
Organizations can improve efficiency by allocating computing resources according to operational priorities. Workload orchestration can help determine which processes must run locally, which can be handled elsewhere and which can be delayed when power or cooling capacity is constrained. Noncritical background activities, for example, may be able to wait when systems detect that critical resources are needed for time-sensitive inference or communications.
Architecture is also an essential consideration. Hybrid architectures that combine local processing with centralized analytics and management can allow organizations to balance performance, connectivity and energy requirements. The right approach depends on the workload, the environment and the consequences of an interruption.
Of course, more efficient operations aren’t the only goal, particularly with edge systems that must be able to withstand power disruptions and inhospitable environmental conditions. Depending on the type of deployment, a resilient edge design may include uninterruptible power supplies, intelligent power distribution, battery storage, backup generation or other local energy resources. These capabilities can help systems ride through interruptions, protect equipment and provide time for an orderly transition when primary power becomes unavailable.
Cooling likewise deserves careful attention. More efficient components and appropriately designed enclosures can reduce heat generation and cooling requirements. In environments where conventional cooling is impractical, equipment selection and thermal design become especially important.
Another part of the equation is ensuring communications within the components of the architecture. Remote monitoring and out-of-band management can allow operators to diagnose problems, adjust configurations and recover equipment without dispatching personnel to every site. Redundant communications paths may be necessary when the primary network is unavailable or degraded.
The Difference Between Deployment and Operational Readiness
The key to ensuring that all these capabilities work together is integration, a primary focus of SD3IT. We help organizations bring the technologies supporting distributed operations together, including rugged computing platforms, power and cooling components, resilient communications and full monitoring capabilities along with ongoing maintenance and lifecycle support.
As a value-added reseller and integrator for federal, military, infrastructure and commercial companies, we work with industry-leading partners to implement solutions that meet operational needs. Schneider Electric, for example, has extensive experience with implementing DCIM tools, as well as with modular power architectures, scalable cooling strategies and deployable facility designs. Another partner, edgeTI, enables disparate edge systems to operate under centralized control with platforms such as edgeCore.
Supply chain planning, secure provisioning, configuration and deployment logistics are also part of the equation. Systems must arrive ready to integrate into the larger operational environment, and organizations need a practical plan for monitoring, maintaining and updating them throughout their service lives.
And, increasingly, energy management should be part of planning from the beginning. The most effective solutions will balance computing performance with energy efficiency and resilience, and give operators the visibility to understand what is happening across distributed sites with adaptable workloads. Because at the edge, power is now an architectural decision.
About SD3IT
At SD3IT, we help federal agencies, defense organizations and commercial enterprises build secure, resilient technology environments that accelerate mission success. From AI-ready infrastructure and zero trust architectures to edge computing, data center modernization and systems integration, we design, drive and deliver solutions that help customers operate with confidence across today’s increasingly complex mission environments.