For a rack, power must be calculated not by the number of devices or the nameplate ratings of the power supplies, but by the actual load in watts, the redundancy scheme, the phase configuration, and the required battery runtime. In practice, this means the following: first determine the actual consumption and the failure scenario of one power path, then choose the UPS by watts and runtime, and only after that select the PDU by current, input, outlets, and connection scheme. Eaton explicitly states that proper UPS sizing requires accounting for the relationship between watts, VA, and power factor, and that a UPS rating should not match the load with no margin.
UPSs and PDUs solve different tasks. A UPS protects against power dips and outages, while a PDU distributes power inside the rack and must match the line, current, input connector, and output types. That is why a rack should not be designed on the principle of “first we will buy a UPS, then some kind of PDU with the required number of outlets”: an error in one element quickly breaks the whole scheme. In its rack and PDU guide, Cisco treats separate PDUs as part of a proper rack power architecture, while Vertiv in its rack PDU documentation emphasizes correct connection to the calculated power branches and compliance with the electrical parameters of the line.
Where the calculation starts
Before choosing a UPS and PDU, you need to gather not only a list of devices, but also the real input data for power. You need an inventory of the equipment in the rack, the actual operating power of each unit, the presence of two PSUs and the A/B scheme, the input voltage, single-phase or three-phase power, plug and socket types, the required battery runtime, and an understanding of what exactly is considered a critical load. It is one thing to survive a brief power dip, and another to wait for a generator to start or to shut down virtualization and storage properly. Practical UPS sizing guides recommend starting with load inventory and actual consumption calculations, not only with nameplate values.
What to collect before the calculation:
- actual consumption of servers, storage systems, network equipment, and auxiliary devices;
- peak load, if it differs noticeably from the normal one;
- the redundancy scheme: N, N+1, A/B;
- the type of input line and allowable current;
- input and output connector types;
- the required runtime;
- the rack growth plan for the next upgrade cycle.
The main mistake at this stage is adding up PSU ratings from the labels. An 800 W or 1600 W power supply shows the upper limit, not the actual consumption of the server. For sizing, it is more useful to rely on telemetry from the PDU, BMC, hypervisor, or on actual measurements under typical and peak load. Eaton’s practical UPS sizing materials explicitly recommend calculating the total protected load and leaving room for growth.
Watts, VA, and amps: where the confusion comes from
For a UPS, both watts and VA matter, but in practice for IT loads, what is more critical is not just the VA number, but the available power in watts. Eaton gives the formulas Watts = VA × Power Factor and VA = Watts / PF, and also notes that if power factor is ignored, the UPS may be undersized. For a simplified single-phase calculation, current can be estimated as A = W / V. Eaton also separately provides simplified conversion guidance for single phase and three phase, emphasizing that a three-phase scheme requires separate consideration of voltage and configuration.
| Metric | What it means | Where it is used | Typical mistake |
|---|---|---|---|
| W | Actual active power | The main reference when choosing a UPS for IT load | Looking only at VA and not checking the watt limit |
| VA | Apparent power | UPS rating and overall sizing | Assuming that 3000 VA always equals 3000 W |
| A | Load current | Selecting the line, breaker, PDU, and input connectors | Not checking whether the cable can handle the current in a failure mode |
| PF | The ratio between W and VA | Load conversion and UPS sizing | Ignoring it when choosing a UPS |
| Runtime | Battery runtime under a specific load | Battery selection and operating scenario | Estimating runtime by the nominal rating instead of the actual load level |
From a practical standpoint, this means one simple thing: if a rack consumes, for example, 3 kW, you care not only about which UPS has a “nice-looking” VA rating, but also how many watts it can actually deliver and what runtime it will provide at that load.
How to calculate rack load in practice
A practical calculation usually looks like this:
- Make an equipment list and record the actual consumption, not only the nameplate figures.
- Separate the typical working load from peak modes.
- Separate critical and non-critical load.
- Define the redundancy scheme.
- Check the failure mode of one UPS, one line, or one power path.
- Add a reasonable margin for growth and battery degradation.
The power system should not operate at the limit. Eaton recommends choosing a UPS with a VA margin above the total requirement of the protected equipment and accounting separately for load growth. Their materials explicitly state that the UPS rating should not exactly match the actual load, and that selecting a solution with extra headroom is standard practice.
It is especially important to calculate racks with two PSUs and A/B connections correctly. In normal mode, the load may be split across two paths, but if one UPS or one line fails, the remaining path must withstand the emergency scenario. In its documentation on racks and RP-Series PDUs, Cisco shows separate power organization inside the rack and the use of multiple PDUs for fault tolerance. This is important not only for large data centers, but also for small server racks, where errors in calculating A/B branches often remain unnoticed until a failure occurs.
A simple example of the logic: if a rack consumes about 4 kW in normal operation and the servers have two PSUs, this still does not mean that each path can be designed for exactly 2 kW with no margin. If one path is lost, the remaining one must survive the load redistribution, and the UPS and PDU on that branch must not exceed the allowable current and power limits.
How to choose UPS capacity
A UPS is chosen not only by the power figure, but also by the operating scenario. For a rack where stability and power quality matter, online double-conversion is usually considered rather than only line-interactive, especially if the grid is unstable, the load is critical, and the rack contains servers, storage, and virtualization. In Eaton’s official materials, online UPS systems are treated as a solution for more sensitive and critical loads, where power quality and a controlled transfer to batteries are important.
When selecting a UPS, check the following:
- available power in watts and VA;
- the actual load in normal and failure mode;
- runtime under the actual load;
- the presence of bypass;
- the ability to connect external batteries;
- network monitoring and graceful shutdown;
- headroom for rack growth.
The phrase “a 1000 VA UPS” says almost nothing by itself. If you do not check watts, PF, and runtime under the real load, you may end up with a unit that formally matches the nominal rating but does not support the load the way it needs to in practice. Eaton links proper UPS sizing to power factor, actual load, and power margin.
How much battery runtime is really needed
More runtime is not always better. In one rack, 5–10 minutes is enough to survive a brief power dip and allow the automation to work properly. In another, it is necessary to hold until the generator starts. In a third, you only need enough time to shut down non-critical services properly and keep the network, storage, and hypervisor running until a controlled shutdown. In its sizing materials, Eaton emphasizes choosing a UPS for a specific task and recommends taking into account not only power, but also the required battery runtime.
Typical scenarios are as follows:
- 5–10 minutes — survive a short outage or switching event;
- 10–20 minutes — wait for automation or shut down part of the services properly;
- more — already a special scenario that must be justified separately in terms of budget, heat, and battery size.
Runtime is calculated from the actual load, not from the “theoretical rack power,” and it drops noticeably when the UPS is heavily loaded. In addition, batteries age, so sizing based on a new system should not be the only reference point.
How to choose a PDU for a rack
A PDU does not replace a UPS and should not be treated as “just a power strip.” For a rack, the input current, voltage, input plug type, the set of C13/C19 outputs, the form factor, and whether monitoring and remote management are needed all matter. Eaton/Tripp Lite documentation distinguishes different classes of PDU — basic, metered, monitored, switched, and others — while Cisco and Vertiv in their guides assume that the PDU must be part of a properly calculated rack power scheme.
| PDU type | When it fits | Advantages | Limitations |
|---|---|---|---|
| Basic | A small rack with no telemetry requirements | Simple and inexpensive | No load monitoring |
| Metered | You need to see the total load on the PDU | Helps avoid overloading the branch | Usually no detailed outlet-level analytics |
| Monitored | You need rack oversight and capacity planning | Telemetry and remote parameter monitoring | More expensive and requires proper operational practice |
| Switched | You need remote reboot and outlet control | Convenient for operations and power prioritization | Requires a careful access policy |
A PDU is not chosen by the number of outlets, but by the full set of parameters. An error in the input type, allowable current, or C13/C19 outlet set breaks the design even when the total power “seems to add up.” For a dense rack, a monitored or switched PDU is useful not only in operation, but also for proper load growth tracking.
Where mistakes happen most often
The most common mistakes look like this:
- calculating by PSU nameplate rating instead of actual consumption;
- looking only at VA and ignoring the W limit;
- not checking the failure mode of one power path;
- choosing a UPS with no margin for growth and battery degradation;
- choosing a PDU only by the number of outlets;
- forgetting to check the input plug type and the C13/C19 set;
- not separating critical load from non-critical load;
- wanting the maximum runtime without understanding why it is needed;
- not planning monitoring where the rack will grow and change.
Almost all of these mistakes come down to one thing: trying to simplify the rack power system to a single figure in kilowatts, while in practice you need to calculate power, current, line type, connectors, redundancy, and the failure scenario all at once.
Practical selection algorithm
If you need a short working route, it is this:
- Collect the equipment list and the actual load.
- Mark peak load and growth plan separately.
- Separate critical and non-critical load.
- Define the redundancy scheme and verify the failure mode.
- Select the UPS by watts, VA, and runtime.
- Check monitoring, bypass, and graceful shutdown.
- Select the PDU by current, input, outlets, and form factor.
- Check compatibility by connectors, phase configuration, and rack layout.
- Make sure the system passes the one-path failure scenario.
The final principle is simple: UPS and PDU should be chosen as a single power scheme, not as two separate products. Proper calculation starts with the real power in watts and the failure scenario, and ends with verifying that the rack can handle the operating load, the loss of one path, and the required battery runtime without overloads or unpleasant surprises.
Sources
- Eaton — VA Versus Watts
- Eaton Tripp Lite — UPS Sizing
- Eaton — UPS sizing guide
- Vertiv — Rack Power Distribution Unit Installer/User Guide
- Cisco — R-Series Rack and RP-Series PDU Installation Guide
