Data Center Structured Cabling

Structured cabling is the passive network foundation of a data center: a standardized, documented system that connects servers, storage, and network equipment through a backbone, cross-connect points, and equipment outlets.

It is critical because active equipment is typically replaced every 3–5 years, while cabling typically stays in place for 10–20 years. An undersized backbone makes every later expansion more expensive, and in an undocumented system fault-finding becomes trial and error.

We cover the standards and common mistakes of structured cabling in more detail in a separate article.

Both are data center cabling standards, but their scope differs. ISO/IEC 11801-5 is the data center part of the international cabling standard: it describes the hierarchy, the performance categories, and the testing requirements. TIA-942 is broader — beyond cabling it also covers room structure, pathways, and facility layout.

The practical difference is in the designations. ISO/IEC uses ENI, MD, ID, ZD, LDP, and EO; TIA-942 uses entrance room, MDA, IDA, HDA, ZDA, and EDA. The mapping is not one-to-one: the ISO/IEC ZD corresponds to the TIA HDA, and ZDA corresponds to LDP — so it is worth stating in the design and in the tender which standard is being referenced.

For data center cabling, EN 50173-5 applies; for the facility, the EN 50600 series; and for the reaction-to-fire classification of cables, EN 50575 under the CPR. The level of fire-safety requirement is set by national fire-safety regulations and the related technical guidelines, which differ from country to country.

It is worth recording the edition year of the referenced standard in design and audit documentation: performance categories and testing requirements have changed between editions.

In the backbone, fiber is effectively always the answer. The copper channel length for 25G and 40GBASE-T drops to 30 meters, and above 40G there is no standardized twisted-pair option — so copper remains justified for short, rack-level equipment connections.

A switch becomes necessary when the backbone needs more bandwidth than 10G, or when the data hall layout pushes segment length beyond the limit of the given copper category. In those cases a copper-to-fiber transition typically costs less than expanding the pathways.

MPO/MTP backbones are built around a fiber-count base unit: base-12 works in units of twelve fibers, base-8 in units of eight, and at 400G base-16 (MPO-16) also comes into play. The choice is not a cosmetic one — it has to match the fiber usage of the transmission technology. Parallel optics (40GBASE-SR4, 100GBASE-SR4) use eight fibers, so on a base-12 trunk the middle four fibers stay unused; the same 24 fibers serve three parallel optics with base-8 but only two with base-12. With duplex optics (10G, and 100GBASE-DR/FR) both systems are fully used, but a base-12 connector gives six duplex channels against base-8's four — the same port count with fewer trunks and fewer terminations. The decision therefore comes down to whether the data hall is moving toward parallel or duplex optics in the long term. This has to be settled early in design: active equipment can be swapped, but the trunk system stays in place for 10–20 years.

On fiber, Tier 1 testing at minimum (attenuation, length, polarity, with an OLTS), and on critical backbone runs Tier 2 as well, with an OTDR trace. On copper, permanent link or channel testing — with the report recording which one was used.

The package should be complete: test results for every port and fiber, the test instrument type and calibration date, the loss budget applied, the labeling scheme, and up-to-date rack and port assignments. If these are missing, the system cannot be audited later.

Tier 1 testing is an attenuation, length, and polarity check performed with an optical loss test set (OLTS): it tells you whether the segment meets the loss budget.

Tier 2 adds an OTDR measurement, which shows event by event where the loss arises — at a splice, a connector, or damage. Tier 2 therefore does not replace Tier 1, it includes it. At handover, Tier 1 is the minimum.

Tier 2 is worth requesting on longer backbone runs, on spliced segments, and on any system where later fault-finding would affect uptime. The term comes from ANSI/TIA-568.3; ISO/IEC refers to the same thing as basic and extended testing, and it is not the same as the Uptime Institute Tier classification.

Yes, but it is a question of planning. Migration is phased: a parallel pathway is built, cutovers happen in pre-agreed time windows, and every step has a rollback option.

The precondition is a redundant route. If the A and B paths run on independent pathways, one side can be cut over while the service runs on the other. Where there is no route redundancy, the first step of a downtime-free migration is creating it.

A condition assessment once a year is advisable: pathway fill, the state of the patching frames, whether labeling and documentation are current, and the spare fiber and port capacity. In fast-changing environments, every six months.

Cabling typically does not announce itself through failure but through slowly degrading transparency: every undocumented cutover increases the risk and the time cost of the next intervention.

Five factors set the order of magnitude: the number of ports and fibers, the fiber capacity and type of the backbone, how far the pathways are already built out, the reaction-to-fire class of the cables, and the level of testing and documentation required.

The largest cost difference generally comes not from materials but from the environment: phased migration in a live data hall requires considerably more design and labor time than building out an empty one.