Data centers worldwide are experiencing great upheaval. They’re growing by leaps and bounds — thanks to big data, IoT, cloud, mobile computing, and automation — but at the same time, there’s tremendous pressure on data center architects, operators, and providers to improve efficiencies and keep costs in check while delivering new services and scaling quickly.

While noteworthy upgrades have been made to servers, storage, and networks, the most limited commodity in today’s data centers is power. The advent of software-defined power (SDP) — following the model of what virtualization did for servers — can provide the aforementioned stakeholders with newfound strategies for offering customers heightened levels of flexibility and availability while lowering costs.

Amid all the technological advancements in today’s data centers, power remains the one must-have element. After all, data centers do not have the option of running out of power. So instead they must over-design, over-provision, and over-spend on the power infrastructure to make sure there’s always sufficient capacity to support the IT infrastructure.

But what if there was a new solution, a simpler and easier way to meet the power needs of today’s bustling compute, storage, and network resources? The time is now for SDP. As cloud, hyper-scale computing, 5G networks, IoT, virtual reality, and artificial intelligence gain momentum and become more widespread, SDP is the missing piece that can alleviate power-capacity pressures and optimize energy utilization.

The Data Center Challenge

While the compute infrastructure in today’s data centers has become more efficient, the costs of power and cooling infrastructure have remained high, at between $6 million to $20 million per megawatt. The result typically constitutes more than 40% of the cost of running a data center. The infrastructure that brings power from the substation to the servers in the various racks represents a dominant portion of this — typically more than 70%. Across the board, whether for enterprises or cloud providers running data centers, or for colocation providers that increasingly host their infrastructure, the pressures to reduce costs are intense and relentless.

Meanwhile, enterprises are coping with large increases in the scale and scope of IT operations within constrained budgets. These environments increasingly feature power-intensive hypercompute and machine learning applications, both on the edge and core of an enterprise’s data center. New form factors, such as micro data centers and edge data centers, are only increasing the cost of the power components. This reality exacerbates the urgency to reduce power costs — without compromising reliability and/or availability.

Meanwhile, cloud providers, under pressure to drive efficiencies, are forced to extend beyond the modest increases they derive on the IT side. At the same time, colocation providers, which provide most of the data center infrastructure coming online, are facing the combined hurdles of high capital costs, longer time to full occupancy, and low business margins. Though the industry has been consolidating to help optimize these factors, the high cost of power infrastructure remains a formidable barrier.

Further complicating the picture, the megatrends of IoT, cloud, mobile, and automation are converging simultaneously. To fully realize the potential of these technologies, data centers need to become more affordable to operate while achieving gains in scalability, flexibility, programmability, and intelligence. While there is clear progress on the IT side toward these goals, power technology remains trapped in archaic hardware-centric solutions and mindsets.

Enter Software-Defined Power

Power is ripe for disruption, and SDP is poised to bring power into the 21st century. Similar to software defined approaches most people are familiar with in the compute, storage, and network spaces, SDP provides a common control plane for power components. The results let data centers efficiently allocate available capacity along with reserves of redundant and back-up power. In the process, SDP improves IT density and paves the way for highly scalable, flexible, and intelligent power distribution. It constitutes the final pillar to realize the vision of the software-defined data center (SDDC).

Typical power infrastructure today is expensive, overprovisioned, and underutilized in order to ensure power availability. The unused power infrastructure, which can amount to 40% to 80%, is recognized as “stranded power” (Figure 1). Considered wasteful but necessary, this scenario arises from two factors:

• IT users need to provide for peaks and buffers to cover the uncertainty of estimated and expected power use. Thus, power must be overprovisioned to cover this rarely utilized “resource.”

• Power infrastructure providers build redundancy in circuits and back-up power systems to ensure power availability in the event of any circuit failure or outage. While the level of redundancy and backup can vary widely, it is captured today by a four-tiered rating popularized by the Uptime Institute. Data center power topologies are static designs, where fixed power budgets and resiliency are allocated to users in racks, rows, or rooms.

SDP safely taps into the unused peaks and buffers as well as redundant and back-up power to repatriate this unused capacity. The result makes it available for added workloads and increased power utilization. In a brownfield operation, this enables more IT workloads to be added in an existing footprint. In greenfield opportunities involving a new data center, module, or new room in an empty shell, SDP can increase power utilization to safely support a given workload.

With SDP’s abstracted control plane, disparate power systems can coordinate actions based on decisions made by central software. The SDP software, which is installed in a host rack, collects data across connected power components and then provides a real-time view of power usage patterns from rack-outlet to rack-to-row to data center level. Machine learning and big data analytics are employed to predict power usage and control run-time actions on power distribution to/in/by/through each power component.

These decisions are dictated by software policy to connected power components in the data center. SDP-compliant power components act on these software policies and execute on SDP’s optimized power distribution strategy. Thus, SDP acts as an “open power hypervisor” that not only stitches together the disparate power components in a data center, but also includes the control planes of compute, storage, network, cooling, and security. The result: One cooperative SDDC control plane (Figure 2).

SDP’s control plane, in combination with two types of hardware agents, transforms a static and single-tiered infrastructure into a dynamic, multi-tiered environment. Intelligent batteries peak shave, like a hybrid car’s battery, and intelligent, fast-acting power switches tap into redundant circuits and/or mix multiple power sources to maximize power availability (Figure 3).

SDP also exposes its coordinated control plane in the form of APIs for easy integration into a single control plane with third-party interfaces, such as those relating to power, cooling, network, security, IT workload orchestration and placement, server power management, and data center infrastructure management (DCIM).

Enabling Tomorrow's Data Centers

SDP raises the tide for everyone by helping to fulfill the vision of a true SDDC. Data center operators, cloud providers, enterprises of all sizes and shapes, data center architects as well as engineering firms, power, and IT component vendors, all can benefit in unique ways that increase efficiency and performance while reducing costs and risks.

For data center providers, current data center designs, based on a specific tier rating and/or static power budgets, deliver poor infrastructure utilization levels and attract a narrower customer base. This often leads to longer times to achieve full occupancy and break-even revenues while decreasing margin potential.

With SDP, however, CFOs and CEOs have the potential to boost revenues while achieving up to four times increases in margin. In addition to realizing returns of capital employed and elevating company value, there are a multitude of benefits from increasing utilization levels from dynamic infrastructure multi-tiering enabling providers to:

• Attract customers with lower price points without compromising gross margin levels.

• Accelerate occupancy in any given footprint as customers rightsize their tier ratings to their needs. For example, a hyper-compute oil exploration customer that needs only four dedicated hours of high-density power per day needs only the hold-up time in the event of power failure for safe shut down and recovery. Similarly, VM workloads can opt for a lower tier footprint provided they are guaranteed the hold-up time for successful workload migration. Big data workloads, which are high-energy consumers, can buy power when available rather than pay premium tier prices.

• Right size new data center builds with lower capital expenditures and quicker time to revenue.

• Increase customer stickiness by offering higher visibility, flexibility to scale, and optimization of workload placement.

From the point of view of planning, designing, and operating data center infrastructure, SDP simplifies and improves automated operation and accelerates time to build and modify by enabling data center architects and managers to:

• Optimize capacity planning and increase precision
• Upgrade visibility by helping diagnose events and improve response time
• Automate and minimize human intervention and errors
• Automate and manage operations more intelligently.

Advantages For Enterprise-Owned Data Centers

Today, enterprise-owned data centers constitute more than 70% of all data center operations. New IT functions increasingly are housed with cloud and colo environments as their IT operations transform into a hybrid model, spreading across their own and multiple providers (Figure 4). With SDP, however, data centers can increase the elasticity and efficiency of existing infrastructure.

In this scenario, SDP, from the perspective of an enterprise CFO, CIO, or overall infrastructure operations view, promises huge benefits in cost savings, time-to-market acceleration, flexibility, and scalability. Whether the data center footprint is owned or leased, here are some of the key advantages:

• A reduction in the total cost of ownership (TCO) for servers: More than $2,000 per server can be saved from avoided power infrastructure
• Accelerated consolidation and cost savings: Faster end-of-life (EOL) cycles, thus creating headroom in data center footprints
• Increased workload density in racks
• Minimized workload bursting into cloud saves on cloud costs
• Accelerated time-to-market, as critical IT deployments can avoid the wait of consuming power provisioning projects
• Minimized dependence on highly skilled engineers and managers for specialized technical environments

SDP Benefits For Power & IT Vendors

Power and IT vendors are suffering from commoditizing margins in a large albeit low growth market. Virtualization and IoT technologies only exacerbate the pressure on their margins as they drive increased customer efficiencies and commoditization.

SDP helps increase power and IT vendors’ market and wallet share in two fundamental ways. By conveying the TCO savings or revenue and margin increase to customers (enterprise and provider, respectively), power and IT vendors can successfully differentiate their products which in turn helps increase their market share. In addition, by tapping into the savings or increased cash flow enabled by SDP, vendors can offer SDP-powered, value-added products and service offerings. The key for power vendors is the opportunity to move up the value chain with end-to-end managed solutions.

For technology vendors, because virtualization as well as workload orchestration and management systems are not power aware, the critical link to SDDC is missing. By adding SDP to the mix through APIs, IT vendors have the potential to deliver a power-aware optimized environment ready for SDDC while easing entry into new markets with software-appointed data center operations.

Realizing The Potential Of The SDDC

SDP has only begun to broach its vast potential. When the technology’s control abstraction plane integrates with other control and orchestration interfaces (Figure 5), the true power of SDDC can be fully envisioned. The prerequisite to optimization of hyper-complex and dynamic data center environments is the combination of intelligence, automation, and control components driven by machine learning and analytics.

Early SDP proponents will be among the first to realize the next wave of innovations and dramatic changes in data center economics and management. Applications such as “follow the sun,” remote management and control, failure prediction and remediation along with accelerated simulations and planning for autonomous intelligent operation will fuel data centers of the future.

Back To Top