What Is a Power Feed? Panels, Breakers, and Rack Power Explained

The word feed gets used for almost every part of the power path. A colocation contract calls the circuit a feed. A rack diagram calls the PDU input a feed. An electrician calls the cable from the panel a feed. By the time the term reaches a change ticket, it can mean the breaker, the cable, the outlet, or the whole path, and everyone in the room assumes they mean the same thing.

The useful definition is narrower. A power feed is a named electrical path from a panel breaker to a rack or device. It is the record that ties the protective device in the electrical room to the equipment in the cabinet, and it carries the details that decide whether a new server fits: voltage, amperage, phase, utilization limit, role, and connected endpoint. Once that definition is fixed, the rest of the power vocabulary becomes easier to place.

At a Glance: Power Terms

The path from utility to server

The power path is a chain, and every link changes the name of the thing you are looking at. Utility power arrives at the building, passes through switchgear and an automatic transfer switch, and reaches a UPS. The UPS output goes to a distribution panel. A breaker in that panel protects a feed. The feed lands on a PDU inlet or directly on a device inlet, and the device power supply converts it to the low voltage the server needs.

Source, panel, breaker, and feed

A source is the supply side of the chain. In a documented model, sources include utility entrances, UPS units, generators, automatic transfer switches, busways, and anything else that can deliver power to a panel. A source can feed another source or a panel, and the link between them records the direction of flow and the role of that input.

A panel is the distribution board. It has a voltage, a main amperage, a phase, and a maximum utilization that tells the model how much of the main rating is usable. The panel is not the same as a patch panel, and the name collision causes real confusion in DCIM projects. If you mean the electrical distribution board, say electrical panel or power panel.

A breaker is the protective switch inside the panel. It has a physical position, a rating in amps, a pole count, and a status. The position is the part a technician reads in the electrical room, which is why it must be stored exactly as printed rather than converted to a sequence number. The rating is the part the capacity calculation cares about, because a feed cannot safely exceed the breaker that protects it.

A feed is the named path from the breaker to the load. It records the panel, the breaker, the rack or connected endpoint, the supply type, the phase, the voltage, the amperage, and the utilization limit. It also records the role, which is where A and B feeds become distinguishable, and the redundancy group that pairs them.

What a power feed record contains

The fields are not arbitrary. Each one answers a question that comes up during a deployment, a capacity review, or an outage.

The capacity fields deserve their own explanation because they are the ones most often copied incorrectly. A feed record usually carries a voltage and an amperage, then applies a utilization limit. That limit is the continuous load allowance, not a comfort margin. The model calculates capacity as voltage multiplied by amperage, multiplied by the utilization percentage, and for three phase feeds multiplied again by 1.732. The result is the usable wattage that the feed can carry continuously.

That number is not a substitute for the device draw. A server with a 1,200 watt power supply does not draw 1,200 watts in normal operation, and adding nameplate ratings across a rack will overstate the load, sometimes by more than a factor of two. Use the measured or documented draw per device for the load side, and use the feed capacity for the ceiling. A power feed exists to tell you where the ceiling is, not to pretend the equipment always touches it.

How to read a feed label and a panel schedule

A useful feed label has the panel, the breaker position, and the rack or role. A label such as PNL-A3-42 tells a technician that the feed comes from panel A3, breaker 42. Adding R07-A tells them it serves rack R07 on the A side. The exact convention is less important than making it readable at the panel and at the rack, because the label is what connects the physical world to the record.

The panel schedule is the other half. It should list the breaker position, rating, poles, and a description of what the breaker feeds. If the schedule is a photo taped inside the panel door, transcribe the active rows before the next change window, then mark the photo as a historical reference rather than the source of truth.

You can also compare the feed against what the host reports about its own power supplies. On many servers, DMI type 39 exposes the power supply nameplate and status. The output is vendor dependent, so treat it as a cross check rather than a complete inventory.

sudo dmidecode --type 39 \
  | grep -E "Name:|Max Power Capacity:|Status:"

        Name: PWR SPLY,750W,RDNT
        Max Power Capacity: 750 W
        Status: Present, OK

The signal is the difference between the supply rating and the feed capacity. Two 750 watt supplies do not require 1,500 watts of feed capacity, because the server does not draw the sum of its nameplates. They do require two independent paths if the server is expected to survive a single feed failure, and that independence is a property of the feed records, not of the power supplies.

Feed, circuit, and PDU are not interchangeable

A circuit is the physical electrical path. A feed is the record of that path, with the panel, breaker, rack, role, and capacity attached. The circuit exists even if nobody documented it. The feed exists only when someone wrote it down, which is why the term should point at the record and not at the cable.

A PDU is a device that takes one or more inputs and distributes power to outlets inside the rack. A feed can land on a PDU inlet or directly on a server inlet. When a PDU sits between the feed and the servers, the feed capacity is the upstream constraint and the PDU outlet rating is the downstream constraint. Both matter, and neither replaces the other.

The same distinction applies to outlets and inlets. An inlet is where power enters a device. An outlet is where a device provides power to something else. A PDU has inlets and outlets. A server usually has inlets and no outlets. A feed connects to an inlet, and an outlet connects to another inlet through a cable. If the model collapses all of these into the word feed, the physical path becomes impossible to trace.

A/B feeds and the two cord mistake

A/B power is two independent paths into the same rack, each capable of carrying the full rack load on its own. The A feed and the B feed should start at separate upstream sources, pass through separate panels and breakers, and land on separate PDUs or inlets. The role field records which side a feed represents, and the redundancy group pairs the two records without relying on the feed name.

Two cords do not prove redundancy. If both cords land on the same PDU, the same panel, or two panels fed by the same UPS, the server has redundant power supplies but one failure domain. The feed record is the place where that distinction becomes visible, because it can show the panel, the breaker, the connected inlet, and the upstream source for each side. A pair of feeds with the same panel or the same source is a pair on paper, not a redundant design.

The sizing rule is equally unforgiving. Each feed must be able to carry the entire rack load during the failure of the other. If each feed can deliver 5 kW, the rack budget is 5 kW, not 10 kW. A feed record that stores capacity per side makes that rule checkable without redoing the electrical math every time a deployment is proposed.

Trace one feed from the rack

The fastest way to test whether the vocabulary matches the physical world is to trace one feed from the device back to the panel. Do it on a rack you can physically reach, and write down each stop rather than relying on memory.

1. Start at the device inlet and read the cord label. If there is no label, the trace starts with a labeling task.
2. Follow the cord to the PDU outlet and record the outlet number. If the cord lands directly on a wall outlet or a busway tap, record that instead.
3. Follow the PDU inlet back to the feed label. The label should name the panel and breaker, not only the rack.
4. Open the panel schedule and find the breaker position. Confirm the rating and the description match the feed record.
5. Compare the upstream source link for that panel with the source you expected. If the panel has no source link, the trace stops at the panel and the record is incomplete.
6. Mark the feed as verified in the record, with the date and the person who checked it.

The trace is deliberately physical. A record that cannot survive this walk is a record that will fail during an outage, when the person following it is under time pressure and cannot ask the original installer what they meant. The same walk also catches the most common A/B mistake: two cords from the same server landing on outlets that belong to the same PDU or the same feed.

The feed checklist

Before you call a power feed documented, confirm these facts:

• The panel exists, is in the correct site, and has a documented upstream source.
• The breaker position and rating match the physical panel schedule.
• The feed has a rack or a connected endpoint, not just a name.
• The role is primary or redundant, and the redundancy group matches the other side of the pair.
• The supply, phase, voltage, and amperage match the design, not the defaults.
• The utilization limit reflects the continuous load rule used by the facility.
• The calculated capacity is greater than the load you intend to place on it.
• The connected endpoint is a real inlet, not a placeholder.

If a feed cannot answer where it comes from, what protects it, what it powers, and how much it can carry, it is a label on a cable, not a record. The next time you open a rack, pick one feed and trace it back to the panel. The gaps you find on that one path are the same gaps that will be waiting during the next outage.

Obelinf models the source, panel, breaker, feed, and endpoint as separate records, so the power feed is a path rather than a free text note. That structure is what makes the capacity and redundancy checks possible, and it is the part worth getting right before the next deployment.