Data center cabling documentation: what to record and how to trace a cable

1. What to record for every cable

One row per cable, one cable per row. These are the DC Sim template’s Cables columns; from_device, to_device and type are required, the rest make the record useful.

Column	Example	Why it matters
from_device, from_port	k8s-0103, iDRAC	Where the cable starts. Ports are free text: Gi1/0/1, eth0, SFP+1, port 5, PSU1, outlet 3
to_device, to_port	cp-r01, port 5	Where it ends. A patch panel is a device with ports
type	network	power, network, storage or console: decides which questions the cable answers
media	Cat6A	Copper category, OM3/OM4 multimode, OS2 single-mode, DAC. A port check needs it
speed	1G	Catches a 10G optic in a 1G port before anyone plugs it in
length_m	2	Spares, slack in the tray, and a check that the cable reaches
label	M-0017	The string printed on both ends
vlan	40	One or more VLANs, so a port change can be checked against the network plan
route	in rack	in rack, overhead tray T2, under floor: where to look for it
color, phase	blue, L2	Colour as installed; phase on power cables for phase balance

Fibre and copper differ in what you must write down. For fibre, the grade (OM4 or OS2) matters as much as the speed, because an OM4 patch into an OS2 port does not work at any speed. For MPO trunks also record the polarity method (A, B or C in TIA-568) and the cassette at each end. For copper, the category and whether it is a DAC are what a port check needs.

A record is only as good as its last update. Change the row in the same ticket as the cable, and walk a sample of racks on a schedule (section 6). A stale record costs you twice: minutes spent tracing during an outage, and the wrong cable pulled during a change.

2. Labels: one name, printed on both ends

ANSI/TIA-606, the administration standard for telecommunications infrastructure (revision D, 2021), is the reference: every space, rack, panel, port and cable gets a unique identifier, built from its location, that matches the records. For a 10-to-50-rack room the practical rules are simpler:

• Racks by row and position (R05, or A05 for row A), the same name in every sheet.
• Panels by rack and use (the demo hall’s copper panel in R01 is cp-r01). They are devices in the Devices sheet with a U position.
• Cables by type and number: the demo hall uses P- for power, N- for network, M- for management, S- for storage and C- for console, then a sequence (M-0017). A location-encoded label (R01-U42-05 to R11-U40-05) reads well in the room but must be reprinted when anything moves; a sequence never changes and the record holds the ends.
• Both ends, machine-printed, within reach of the connector. A label you cannot read with the cable plugged in will not be read.
• Never reuse a label number. A retired cable’s number stays retired, or old tickets point at the wrong cable.

3. Patch panels are two rows, never one

A server’s management port reaching the out-of-band switch through two copper panels is three cables, so three rows. This is a real path in the DC Sim demo hall:

label	from_device	from_port	to_device	to_port
M-0017	k8s-0103	iDRAC	cp-r01	port 5
M-0018	cp-r01	port 5	cp-r11a	port 5
M-0019	cp-r11a	port 5	oob-01	Gi1/0/5

The middle row is the permanent link between the rack’s panel and the row-end panel, run through the overhead tray. Writing one row from k8s-0103 to oob-01 looks tidier and is wrong: the day someone repatches cp-r11a, the record still claims a path that no longer exists, and nobody knows which panel port to check.

A colocation cross-connect follows the same rule: the provider’s panel in the meet-me room is a device, and the cross-connect is its own row, labelled with the provider’s circuit ID.

4. How to trace a cable end to end

On paper: find the row with the device and port you start from; read to_device and to_port. If that is a panel, find the row that leaves the same panel port on the other side and keep going until you reach a device that is not a panel. A lookup on device plus port does it in a spreadsheet.

In the room: read the label at each hop and compare it with the row. The tools only cover parts of the path:

Method	Sees	Does not see
The record (sheet, DCIM, twin)	Every hop, route, label, VLAN	Anything changed since the last update
LLDP / CDP from the switch	The device at the far end of an active link	Patch panels, unpowered devices, the route
Toner and probe (copper)	One physical cable, even a dead one	Anything beyond the panel without re-toning
Visual fault locator (fibre)	Continuity and the far end of one strand	Which port it should be on

Use LLDP as a cross-check of switch ports against the record. It cannot replace the record, because the panels in the middle are passive.

In DC Sim: in the Cabling view, click a cable to see both ends, ports, media, speed, length, label and VLANs. Trace lights the full path end to end through every patch panel, and cutting the cable shows who loses connectivity. The Network view builds the topology from the same rows with the panels collapsed, so a trace and a “fail this switch” question use one record.

5. Common errors and how to catch them

Error	What it looks like	Catch it in the sheet	DC Sim check
Orphaned cable	A row names a device that is not in Devices, or a cable left behind after a decommission	Lookup of every from/to name in Devices	Unknown names in the Cables sheet are reported and the row is skipped
Device with no cable	A server with no network or power row	Count rows per device	Flagged: no network or power cable
Duplicate label	Two rows with M-0042	Count of each label above 1	Missing labels are flagged; check duplicates in the sheet
Port used twice	Two cables on one switch port	Count of device plus port across both ends	Flagged: port used twice
Speed or media mismatch	A 10G DAC on a 1G management port; OM4 into OS2	Compare speed and media with the port’s model	Flagged per port
Too short	A 1 m cable between racks four metres apart	Length against the route	Flagged: shorter than the distance between its ends
Stale record	The trace says port 5, the room says port 7	Only a process fixes this	Re-upload shows a change list

The demo hall carries one planted cabling mistake, and the import finds it: M-0167: 10G on esx-0705 ilo (port max 1G), plus “DAC cannot plug into copper port”. Power cables get the same treatment: both power supplies on one feed is a cabling error with a power consequence, covered in the UPS failure impact analysis.

6. Keep it true

Documentation is only worth its last update. The rules that work in small rooms:

• Update in the change ticket. A move, add or change is not closed until the row is changed.
• Re-check after every change. Re-importing the sheet into DC Sim lists exactly what changed before anything is applied.
• Audit on a schedule your change rate justifies. A busy room walks a sample of racks every quarter; every room walks the racks it is about to touch before a maintenance window.

What undocumented cabling costs shows up in two places: the minutes spent tracing during an outage, and the wrong cable pulled during a change.

7. Where the cable list comes from

• Excel or CSV: the template’s Cables sheet, or a CSV file with the same columns. Start with power and uplinks; they answer the failure questions. The Excel guide covers the other sheets.
• NetBox: DC Sim connects with a read-only token and turns a site’s cables into the Cables sheet, lengths converted to metres; provider circuits are left out and listed. See NetBox to 3D.
• Another tool: export its cable list as CSV into the template’s columns. Nothing is written back to the source.

Dense GPU rows multiply rows, not columns. In the demo hall every GPU node has 2 × 100G uplinks to a leaf pair, every leaf has 4 × 100G OM4 to the spines through the rack fibre panel and the overhead tray, and every server has a 1G management link through two copper panels. Same rules, more rows: 910 cables for 13 racks.

The Cabling and Network layers page lists every cabling and network check. To see your own file in 3D, download the Excel template and ask for a walkthrough.