DataCentersX > Facility Ops > WAN & Comms Ingress

DC WAN & Comms Ingress

WAN and comms ingress is the physical infrastructure through which external connectivity enters the data center. The category covers the buried fiber entrance from public right-of-way, the demarcation rooms where carrier circuits terminate, the meet-me rooms where customers cross-connect to carriers and to each other, the WAN edge routers that terminate operator-owned circuits, and the path diversity engineering that ensures no single physical event can sever all connectivity. WAN ingress is where the data center stops being an internal network and becomes part of the global internet.

The physical path

External connectivity enters the facility along a sequence of physical layers, each with operational and security implications.

Path diversity

Resilient WAN ingress requires path diversity across multiple dimensions. Carrier diversity means circuits from at least two independent carriers, ideally with no shared upstream infrastructure. Path diversity means physical fiber routes that do not share trenches, conduits, or BEFs - so that a single backhoe, water main break, or fiber cut cannot sever all connectivity. Geographic diversity at the regional level handles broader events (storms, earthquakes). High-availability data centers typically engineer for at least two BEFs at opposite sides of the building, served by physically separated carrier conduits, with circuits from a minimum of three independent carriers.

Meet-me rooms

The meet-me room is where the data center's commercial value as an interconnection facility lives. In a carrier-neutral colocation facility, the MMR is the room where any customer can cross-connect to any carrier or to any other customer with a copper or fiber jumper. The MMR's interconnection density is the primary driver of facility value in retail colocation - Equinix's MMRs in IBX facilities like LA1, NY4, DC2, FR5, and TY2 are the most concentrated interconnection points in their regions and command pricing based on that density.

For hyperscaler-operated facilities, the MMR concept is replaced by the operator's own peering infrastructure - the operator typically wants direct relationships with major carriers and exchanges rather than mediating customer cross-connects. The economic logic differs but the physical infrastructure (cabling, patch panels, optical distribution frames) is similar.

Carrier hotels

A carrier hotel is a building (or set of buildings) where many carriers terminate their networks specifically to interconnect with each other and with content providers. Notable carrier hotels include 60 Hudson Street (NYC), 1 Wilshire (Los Angeles), Equinix LA1 (formerly Coresite Onepartner Drive), 350 Cermak (Chicago), 165 Halsey Street (Newark), 350 East Cermak (Chicago), Telehouse North (London), Interxion AMS-5 (Amsterdam), Equinix TY2 (Tokyo). Carrier hotels concentrate optical and IP transit interconnection at a level no single-tenant facility can match, which is why retail colocation in carrier-hotel buildings commands premium pricing despite often having older facility infrastructure than purpose-built modern data centers.

Internet exchanges

Internet exchange points (IXPs) sit physically inside major carrier hotels and meet-me rooms. The exchange is a layer-2 fabric that allows participants to peer with each other directly without paying for transit. Major IXPs and their primary host facilities include:

WAN edge equipment

Subsea cable landings

For data centers near coastlines and submarine cable landing stations, the WAN ingress story extends to the cable landing facility itself. Cable landing stations (CLS) terminate transoceanic fiber and provide the first level of switching between subsea and terrestrial networks. Notable subsea hub regions include Marseille (multiple Mediterranean cables), Lisbon (Atlantic cables), Sydney (trans-Pacific cables), Singapore (intra-Asia and trans-Pacific), Virginia Beach (Atlantic), and the US-Pacific coast (Oregon, California). Hyperscalers have increasingly co-located AI training and content delivery facilities near major cable landings to minimize transit latency to global users. Marea (US-Spain), Dunant (US-France), Grace Hopper (US-UK-Spain), 2Africa (Africa circumnavigation), Echo (Singapore-California), and Bifrost (Singapore-California) are examples of cables built or co-funded by Meta, Microsoft, Google, and Amazon to control intercontinental capacity.

Hyperscale operator-owned WAN

The largest hyperscalers operate their own global backbones rather than relying on carrier transit for inter-data-center traffic. Google's B4 network, Meta's Express Backbone, Microsoft's WAN, and AWS's global backbone are operator-built and operator-operated long-haul networks connecting their data center fleet. The WAN ingress at these operator-owned facilities terminates these private backbones in addition to (or in some configurations instead of) public-internet peering. The operational discipline of running a global software-defined backbone at this scale is its own subdiscipline of network engineering and a core competitive advantage for the hyperscalers that have it.

Where this fits

WAN ingress hardware is operated under FACILITY OPS as a physical infrastructure system. The routing, peering, and traffic engineering are operated under Compute Ops as part of Network Operations. The cross-connect commercial relationships with carriers and customers are operated by the facility's revenue and account management teams. The cybersecurity at the WAN edge (DDoS, BGP hygiene, edge filtering) is operated under Security. Compliance evidence on physical interconnection (carrier contracts, cross-connect inventories) flows to GRC:Compliance.

Related coverage

Facility Ops | Compute Ops | Network Ops | Stack: Networking & Fabrics | Security | Physical Access | Life Safety | Colocation DCs