RSystems

Power · Hardware

Electrical & Rack Power

Clean, redundant power is the foundation everything else runs on. A practical guide to circuits, voltage, plugs, and UPS sizing — and the hardware we reach for, from a 1U CyberPower to an APC Symmetra.

Every switch, server, access point, and camera in your environment depends on one thing that rarely gets discussed until it fails: clean, uninterrupted power. Get it right and nothing thinks about it for a decade. Get it wrong and you discover the gap at the worst possible moment — during an outage, when the equipment that was supposed to keep running simply doesn't.

This is a practical guide to how rack power works, how to size it, and the hardware we reach for to protect it.

A Little Electricity

Three quantities describe any electrical circuit: voltage (V, the electrical "pressure"), current (I, measured in amps, the rate of flow), and resistance (R, how much the circuit opposes that flow). They relate through Ohm's law:

V = I × R

Power — the work actually being done, measured in watts — is voltage times current:

P = V × I

You don't need to do the math on site, but it's worth keeping that second equation in mind, because it explains the one place voltage choices come up in a rack.

Voltage and Phases

Standard North American power is single-phaseThe standard electrical supply for most residential and small commercial environments — 120V or 240V delivered on a single alternating current waveform. 120V, and that's what the overwhelming majority of equipment we install runs on. A 20-amp 120V circuit delivers roughly 1,920 watts — and you plan to about 80% of that, around 1,500 usable watts. For a typical rack, that's plenty.

Higher-power applications occasionally need more. Because power is voltage times current, you can push more watts over the same wiring by raising the voltage — which is why very large or dense loads sometimes move to 208V or 240V, and why the biggest deployments use two- or three-phase powerDelivers three alternating current waveforms 120° apart on three conductors — more efficient power delivery for high-density equipment, and the standard for data center electrical distribution.. But that's the exception. Unless we're feeding something like a Symmetra, we spec 120V single-phase circuits: almost everything in a normal rack is happiest there, and it keeps the whole installation simpler.

Plugs Have to Match — All the Way Down the Chain

This is where installations quietly go wrong. A power connection only works if the plug, the receptacle, and the equipment all agree on voltage and amperage — and the NEMA connectorStandardized power plugs and receptacles used in North America. The NEMA numbering system encodes voltage, amperage, and locking vs straight-blade type. type encodes exactly that.

Say you want a 20-amp circuit. That alone isn't enough information. A 20A 120V receptacle could be a straight-blade 5-20, or it could be a locking L5-20 — a twist-lock connector that physically rotates to latch so a cord can't be pulled or shaken loose. The two are not interchangeable. In a rack, locking connectors are the norm, precisely because nobody wants a power cord backing out of a socket behind a running server.

The rule that keeps this straight is to match each link in the chain:

  • The UPSProvides battery backup that keeps equipment running through a brief power outage and allows for a controlled shutdown if power isn't restored. input must match the wall receptacle it plugs into.
  • The PDUDistributes power from a single feed to multiple devices in a rack. The power strip for server infrastructure — built for the density, reliability, and monitoring requirements of production environments. must match the UPS output it draws from.

Get either one wrong and you've got equipment you can't plug in, or an adapter quietly introducing a new point of failure into your power chain. Specifying the plugs correctly — wall to UPS to PDU to equipment — is a small detail that prevents a frustrating day on site.

The Hardware We Reach For

For most rack UPS and PDU needs, APC is our first choice — reliable, well-supported, with a battery-replacement and management ecosystem that holds up over the long life these units actually see. It isn't the only hardware we use: when all we're protecting is a small router and a single switch, a 1U CyberPower does the job for far less, and we reach for one often. But for anything that matters, APC is where we start. The models we use most:

  • Smart-UPS SMX1500 (1500 VA / 1200 W) — a single rack or a small closet, ~$1,000
  • Smart-UPS SMX2000 (2000 VA / 1600 W) — a well-populated rack, ~$1,750
  • Smart-UPS SMX3000 (3000 VA / 2700 W) — a dense rack, or headroom to grow, ~$2,500
  • Symmetra / Smart-UPS Modular Ultra (modular, scales into many kW) — whole-room, redundant, mission-critical, from ~$10,000+

(Prices are approximate street prices to show relative cost, not quotes — confirm current pricing at time of purchase.)

Notice the pattern: capacity and runtime climb steeply in price. Doubling the watts and adding battery runtime doesn't add a little to the cost — it multiplies it. That's why sizing matters. You want enough headroom to carry the load and ride through an outage, without buying a small power plant you'll never fully use.

The Symmetra line — recently succeeded by the Smart-UPS Modular Ultra — is a different class of device, modular and redundant by design. Power and battery modules can be added, and an internal module can fail without dropping the load. This is what you deploy when power simply isn't allowed to fail.

Protecting Whole Circuits from the Panel

For a small office, a UPS sits in the rack and protects the rack. At scale, the smarter move is to back entire circuits at the electrical panel — so everything downstream is protected, not just the gear in one cabinet.

Little Island is the clearest example we've built. A pair of APC Symmetra units sit on dedicated 100-amp circuits at the electrical panel in the park's server room, feeding every rack across the park. Every switch, every wireless access pointThe device that creates a Wi-Fi network — bridges wireless clients to the wired Ethernet infrastructure and is managed individually or via a controller., every camera, and every life-safety system draws from them. The result is redundant powerUsing multiple independent power paths to ensure equipment keeps running when a single power source, circuit, or UPS fails. on everything, with more than six hours of runtime — enough to keep the entire infrastructure and life-safety systems running through an evacuation or an extended emergency. Power there isn't a per-rack afterthought; it's an architecture.

Don't Forget the Heat

One last thing that's easy to overlook: every watt of power you deliver to a rack comes back out as heat. Electrical sizing and cooling are two halves of the same problem. If you spec the power for a dense rack but not the cooling to carry away the heat that power generates, you've simply moved the failure from "no power" to "too hot to run."

A good buildout specs electrical capacity and HVAC capacity to the same load, together — so the room can both power the equipment and keep it within its operating temperature. We design both as part of any server room buildout, and it's a core consideration in our network infrastructure design work. Power and cooling are where good infrastructure quietly earns its reliability.