Distributing Power at the Edge
Keeping power flowing and properly distributed to very dense rack-based servers is critical, but not all rack power distribution units (rPDUs) are the same.
In our ever-evolving digital world, enabling compute resources closer to people and applications is driving the exponential growth of edge computing. In fact, according to Markets and Markets research, the global edge computing market is projected to reach USD 15.7 billion by 2025. The analysts estimate that hardware components hold the largest market size, owing to the large-scale deployment of hardware components for decentralizing storage and computing operations, enabling comprehensive edge infrastructure deployment, and reducing network traffic. Moreover, analyst firm Omdia expects server deployments at edge locations to double through 2024, totaling an estimated five million units.
And, it’s not just personal data usage that is growing. Rapid advancements in 5G, AI, autonomous vehicles, IoT/IIoT, telehealth, analytics, and many other processes will drive data centers to be more intelligent and server configurations to be more densely populated. Keeping the power flowing and properly distributed to very dense rack-based servers is a critical piece of the power availability equation. However, it’s important to consider that not all rack power distribution units (rPDUs) are the same.
Small But Mighty
Space is at a premium in edge deployments, requiring infrastructure equipment to have a slim footprint while being robust and flexible to power a variety of edge devices reliably. When deploying power distribution at the server rack, selecting rPDUs with a small footprint as well as the right amount of receptacles and remote management is the first consideration. Some rPDUs on the market are perfectly designed for space-constrained areas and offer many more usable power receptacles. For example, vertically configured rPDUs from Gateview Technologies maximize power density with up to 52 kW in a single 72” high x 2.18” wide footprint. Depending on the rack and cabinet design, even multiple single- or three-phase rPDUs can be installed in the rear of the cabinet to provide power to an entire rack of servers.
Another way operators can deal with limited rack space is to add switching capability to individual outlets that require control. Instead of purchasing bulky switched PDUs, which reduce the receptacle density of the PDU, users can easily install SwitchLok cords where needed and maintain the slim profile of an unswitched PDU – a significant benefit in dense server rack environments. The ability to selectively switch outlets can be a cost-saving, scalable solution for operators who find that they need control and monitoring capability that can be incorporated during the initial installation or as an easy retrofit at the individual receptacle level. Communication and control further simplify deployment.
With the demand for more receptacles on rPDUs, the secure handling of power cords can eliminate frustrations and accidental removal of cords. Many mechanical designs do not make efficient use of space, resulting in larger PDUs providing the same number of accessible receptacles. A smarter and patented receptacle design, called CORDLOK™, accommodates typical pronged plugs with a two-position gripping latch, replacing other cumbersome or ineffective mechanisms preventing micro-gaps that can fail to secure the electrical plug to the receptacle. With this type of design, users can single-handedly insert and remove cords with a positive click when fully inserted. To optimize installation into the server rack, users can select units that incorporate a standard angled whip (cable) entry that allows the input cord to immediately exit the rack without interfering with equipment in the rack.
The very nature of edge computing is that locations are typically unmanned and can be sited inside modular data centers, industrial parks, cell towers, or smart cities, to name a few dispersed locations. This requires power distribution units to provide intelligent remote management capabilities, including an interface dashboard, to provide visual historical graphing of power status. By monitoring the rPDU’s operations via SNMP or through the PDU web server, operators can quickly view all vital parameters, including alarms, amps, kilowatts, volts, and the internal temperature of the PDU to ensure smooth operations and facilitate troubleshooting remotely (Figure 1). In addition, knowing the temperature and humidity in high-heat racks is an added measure of safety and reliability. Small battery-free environmental sensors (1.4in x 1.4in x .81in size) can be added to each rPDU to provide a complete view of the rack environment. Users can deploy them with the original PDU installation or add additional sensors as needed in previously configured racks.
Figure 1. Gateview Technologies' PowerLok with CordLok and remote monitoring.
Not all PDUs are the Same
When depending on power components, reliability is a key concern. Every system in the power chain must be able to deliver or manage power 24/7. While PDUs are often taken for granted due to their ubiquity, selecting a poorly constructed PDU can wreak havoc with racked servers. Mass-produced power distribution units are built with traditional high-fault wiring methods that rely on the labor-intensive use of insulation displacement and fast-on connectors. These hand-worked connections are inconsistently assembled, resulting in high failure rates. A few manufacturers utilize a state-of-the-art manufacturing process that reduces the number of internal contacts. Robotically soldering connections from the line cord to each receptacle eliminates the potential for human error, increasing the reliability of the PDU by as much as 300 percent.
The demand for 24/7 connectivity and data closer to users and applications will increase rapidly, putting more pressure on edge computing infrastructure. Ensuring the intelligent distribution and monitoring of power at the server rack will help enable edge computing to be configurable, resilient, and scalable to meet our insatiable appetite for continuous content delivery.