Rack PDUs and Power Planning: Why Circuits Are the Limit

Rack power gets planned around how many sockets are available, which is the wrong constraint. The binding limit is almost always the circuit feeding the rack, and running into it is how a half-empty rack stops accepting equipment.

The circuit is the limit, not the rack

A rack has plenty of vertical space. What it may not have is enough electrical capacity to run equipment filling that space.

Circuits are rated in amps, and there is a working limit below the rated figure that continuous loads should not exceed. Equipment running around the clock is a continuous load, so the usable capacity of a circuit is meaningfully less than its rating.

The practical result is that a rack frequently reaches its power limit well before its space limit, particularly with dense servers or anything carrying accelerators.

Plan power before filling space. Discovering the constraint after equipment is racked means either moving hardware or arranging new circuits, and the second is slow and expensive.

PDU types

Basic. Sockets and nothing else. Cheapest, and provides no information about what is being drawn.

Metered. Reports total draw, either on a local display or over the network. This is the minimum worth buying for any rack you are managing, because it tells you how close to the limit you are.

Switched. Allows individual outlets to be powered on and off remotely. Useful for rebooting equipment that has hung and does not respond to its management controller, particularly in unstaffed or remote sites.

Outlet-metered. Reports draw per outlet rather than in total, which is what lets you attribute consumption to specific equipment and identify what is actually using the capacity.

For a remote site, switched is frequently worth the difference on its own — one avoided site visit can cover it.

Redundant supplies need redundant paths

The most common power design mistake, and it makes redundancy decorative.

A server with dual power supplies is protected against a supply failing. If both supplies plug into the same PDU, on the same circuit, then the PDU or the circuit is a single point of failure and the second supply achieves nothing.

Proper redundancy means two PDUs, fed from separate circuits, with one supply of each server on each. A circuit failure, a PDU failure or a supply failure then leaves everything running.

This changes capacity planning in a way people miss. Each circuit must be able to carry the whole load alone, because when one fails the other takes everything. Loading both circuits to their working limit means that losing one overloads the other and takes down the rack — the opposite of what the design was for.

Practically, that means each path carries roughly half its usable capacity in normal operation.

Working out actual draw

Do not add up power supply ratings.

A supply rating is the maximum it can deliver, not what the server consumes. Actual draw is typically well below it, and summing nameplate figures produces a number two or three times too high — which leads to over-provisioned circuits and PDUs.

Most enterprise servers report actual power draw through the management controller, and metered PDUs report it directly. Use measured figures.

What genuinely drives consumption: processor TDP, memory module count, drive count, and expansion cards — with accelerators dominating where present. Our power supply guide covers this.

Then add headroom for growth, and remember that under-provisioning a circuit produces breaker trips, which are considerably worse than a slow server.

Power and heat are the same problem

Every watt drawn becomes heat inside the rack, so a rack near its power limit is also near its cooling limit.

That connection is why efficiency matters beyond the electricity bill: less efficient equipment draws more for the same work and produces more heat, which the fans then remove using more power still.

If a rack is approaching its power limit, check its thermal situation too — blanking panels, airflow paths and inlet temperature. Our airflow guide covers what to look at.

Where the UPS fits

PDUs distribute power; a UPS conditions and backs it up. They are separate purchases doing separate jobs.

The usual arrangement is UPS feeding PDUs feeding equipment. Two points worth checking.

UPS capacity covers what is actually on it. Growth in the rack changes the UPS load, and a UPS sized at installation may not still be adequate.

Both power paths need protection. Protecting one circuit and not the other means a supply failure on the unprotected path still causes an outage on single-supply equipment.

Our UPS guide covers sizing and shutdown integration.

Practical details

Plug types. Rack equipment uses several connector standards, and PDU outlets must match what you are plugging in. Check before ordering, particularly with mixed-age equipment.

Vertical or horizontal. Vertical PDUs mount in the rack’s rear channel and consume no rack units, which matters when space is tight. Horizontal units occupy rack space but are easier to reach.

Cable length and routing. Power cables at the rear compete with data cabling for space, and a dense mass of cable blocks exhaust airflow.

Labelling. Label both ends of every power cable with what it feeds and which PDU it comes from. During an incident, tracing cables by hand is time nobody has.

Common questions

Planning a rack build? Tell us the equipment list and we will help work out the power requirement before you fill the space.