What it is
A data center cross-connect is a physical cable or fiber link inside a colocation building that connects two termination points. One end is usually your cabinet, cage or patch panel. The other end is a carrier, a cloud on-ramp, another customer, or a second cage you own. The building runs it through its trays and risers, often from the meet-me room to your space.
It is not an internet service. A cross-connect gives you a private path to a provider's port. The provider still sells the circuit, the IP transit or the cloud connection that rides on top. You can order a cross-connect from the building even when the carrier does not own the building. That carrier-neutral setup is common in colos.
Cross-connects come as fiber, copper or coax. Fiber is the default for carrier and cloud links. Copper, usually Cat6, handles out-of-band management and short runs. Each cross-connect gets an ID, a monthly recurring charge and an install charge. Treat that ID as an asset, because it is the reference the building uses when a link fails.
How it works
Start with the service. You order a circuit, a cloud on-ramp or a port from a provider that is present in the building. You and the provider complete a letter of authorization, often called an LOA or LOA-CFA. That letter names you, the provider, the building, the port and the requested speed. Without it, the building will not let a technician touch the provider's panel.
Next you submit the building's cross-connect order form. Include the LOA, your cabinet or cage number, the provider's location, the connector type and the fiber type. For fiber, say single-mode or multimode and the connector, usually LC. For copper, say Cat6 and whether it is shielded. Ask for a written install date and the name of the technician.
On install day, the building runs the cable from your demarc to the provider's panel. The technician tests it: a light meter reading and a polarity check for fiber, a cable test for copper. Both ends get labeled with the cross-connect ID and the circuit ID. Then you configure the service, which may mean an IP address, a VLAN, a BGP session or a cloud routing handoff. Test with a ping, a traceroute and a throughput check before you accept the handoff. If the cross-connect links two of your own cages, the same order process applies.
Why it matters for a business
A cross-connect gives your traffic a private path that does not cross the public internet. That matters when you connect to a cloud on-ramp, a payment provider or a partner that requires a direct link. It also keeps backup and replication traffic off your internet edge, which makes capacity planning simpler.
It changes how you buy. The cross-connect is a line item with its own install charge and monthly recurring charge. It also has a lead time. If you order the carrier circuit first and the cross-connect second, the circuit can sit idle while the building schedules a technician. Put both orders on the same timeline.
It affects redundancy. Two cross-connects that share the same tray, conduit or meet-me room panel are not fully diverse. Ask for path diversity in writing and check the route on the building's drawings. That is an operational question, not a paperwork question.
Common mistakes
The most common mistake is ordering the circuit and forgetting the cross-connect. The carrier sees a live port. You see no link. Another is a missing or wrong LOA. The building cannot approve the work, and the install date slips.
Connector and fiber mismatches cause delays. A single-mode order with a multimode patch panel will not work. A polarity flip will show link on one side and nothing on the other. Send photos of the panel and the port labels to the building before the technician arrives.
People also label one cross-connect as redundant. If both links run through the same tray, a single cut takes out both. Ask for separate paths and separate panels where the building can provide them. Do not assume the building will do that by default. Keep an inventory too: the cross-connect ID, circuit ID, provider, panel, port, connector type and install date. When a link fails, that record saves hours.
How to get started
Ask the building for its provider list, its cross-connect order process and its current install lead time. Get the forms before you need them. Then map your current links and note which ones share a path.
Choose the fiber type and connector from the provider's handoff, not from habit. Order the cross-connect and the circuit on the same schedule. Ask for a diverse path in writing if the link is part of a redundant pair. Set a test plan for the handoff: light levels, ping, route and throughput.
Keep a simple register. One row per cross-connect with the ID, circuit, provider, panel, port and owner. Update it when you add or remove a link. That register is the difference between a clean change window and a fire drill.
If you need a second opinion on the design before you order, one option is the Cloudverg Expert architecture session at $450/month. It gives you a 90-minute call with a senior architect, a pre-read of your diagrams and a written summary. Your carrier's design team, the building's engineers and an independent consultant are other options. Pick the one that fits the decision in front of you.
Frequently asked questions
Is a cross-connect the same as internet access?
No. A cross-connect is a physical path inside the building. Internet access is a service that runs over that path. You still need a provider to sell you transit, a cloud connection or a private circuit.
What is a meet-me room?
It is the shared space in a colocation building where carriers and other providers terminate their equipment. Cross-connects often run from your cage to a provider's panel in that room. Access is controlled by the building.
Can I get a cross-connect between two of my own racks?
Yes. The building can run a link between two cages or cabinets you lease. It is still ordered, billed and labeled like any other cross-connect. Ask for a separate path if the link is meant to be redundant.