Dual-Cord Servers: Map Both UPS Power Paths

Give each cord a complete route

Label the physical PSU and its cord before naming the upstream paths A and B. Trace each cord through the actual outlet, PDU branch, UPS output and source. Record any shared bypass, distribution panel, battery arrangement or other dependency relevant to the claimed failure scenario.

A drawing with two lines ending at two UPS boxes is incomplete if both boxes depend on the same unexamined upstream element. Independence is always relative to a defined failure: two paths may tolerate a PSU loss but still share a distribution fault.

Use this proposed path record for each cord. It is a documentation tool, not a wiring instruction.

For example, a hypothetical documentation map might read as follows. Arrows trace dependencies; they are not connection instructions or evidence of an installed system.

• PSU 1 / cord C1 → outlet A-07 → PDU A / branch A1 → UPS A → panel X → source S.
• PSU 2 / cord C2 → outlet B-04 → PDU B / branch B2 → UPS B → panel X → source S.
• Shared failure domain: panel X and source S. Separate PDUs and UPS hosts do not remove that dependency.
• Claim under review: supported survival of a defined PSU/path loss. Whole-source independence remains unproven and must not be claimed from this map.

Replace every placeholder with verified identifiers and mark any untraced segment unknown. The map should make an excluded failure as visible as a supported one.

Establish what the equipment supports

A second installed PSU may support different modes or capacity arrangements. Obtain the server manufacturer's statement for the actual hardware and workload; do not infer that either PSU can carry everything from its presence alone.

Cisco's N93108TC-FX3 switch documentation provides a useful, explicitly model-scoped example: it describes separate sources for 1+1 redundancy and explains PSU load sharing. It is not a server installation manual. Its value here is to show that supported modes and source requirements need explicit equipment documentation.

Highidea's rack UPS products can be assessed as parts of a proposed supply design. Their rack form factor or online topology does not certify that a particular server's A/B system is independent.

Budget the surviving path, not half the normal reading

A path carrying part of the normal demand may need to accept more after the other path fails. Use the equipment manufacturer's supported behavior and representative load evidence when evaluating this state. Include other equipment already assigned to the surviving PDU or UPS.

For a hypothetical illustration, assume one server draws 800 W in the agreed operating scenario and its PSUs normally share that demand. A normal 400 W reading on each path does not establish that each needs only 400 W of capacity. If the supported loss scenario places the server's 800 W demand on the remaining path, that is the state the design must assess.

The figures are illustrative, not a PSU rating, transfer guarantee or recommended reserve. Losses, startup behavior, input limits and other connected loads require their own evidence. Do not double the server's demand merely because it has two cords, or halve the failure-state requirement merely because it shares normally.

Challenge shared dependencies explicitly

List each claimed failure and the expected surviving route. Identify what the claim excludes. A common upstream supply might be acceptable for a narrowly defined PSU-failure objective but insufficient for an upstream-distribution objective.

Use confirmed, failed and unknown statuses for each assumption. Where the path is not traceable, do not mark it independent. Resolve the drawing or field-identification gap with the responsible electrical and IT teams.

Any failure demonstration must use an authorized, equipment-specific procedure. This article does not recommend pulling live cords, operating breakers or transferring production loads to test a drawing.

Retain the map after changes

PDU replacements, server moves and temporary connections can change the failure domains without changing the server's two-cord appearance. Tie the approved map to an asset/configuration revision and update it through the site's change process.

Send Highidea the per-path requirements and agreed surviving demand when requesting UPS options. Keep the complete independence assessment with the system designer so a quotation for two hosts is not mistaken for an end-to-end resilience result.

Primary references

The references support the specific distinctions stated above. Manufacturer examples apply to their documented products; they do not establish the same capability in a Highidea model. Proposed worksheets and hypothetical examples are editorial decision tools, not approved installation or operating procedures.

Cisco: N93108TC-FX3 Hardware Installation Guide: Plan for power requirements