Colocation cage and cabinet fit-out field guide

What a colocation fit-out is, and whose rules you live by

A colocation fit-out is the work of building out your space inside a provider's data center. The space is a cabinet, a caged-off area of the floor, or a private suite, and the line that defines the whole job is ownership. The colo owns the building, the power plant, the cooling plant, and the security. You own the gear in the rack, the cabling inside your footprint, and the configuration.

That split changes how you think about everything compared to a building you control. You do not buy a UPS and a chiller. You commit to power in kilowatts and pay for it on the provider's terms. You do not design the room's airflow. You fit your cabinets into the containment scheme that already exists and stay under the heat the room can take per cabinet. You do not pull your own fiber to the carrier. You order a cross-connect through the meet-me room. And you do not set the access policy. You operate under the colo's security and the service level agreement you signed.

Get four things right and a colo deployment lands clean and close to budget: the power commit, the cooling and density you can hold, the cross-connects and connectivity, and the access. Get any of them wrong and the cost shows up later, as a stranded power bill, a hot cabinet you cannot cool, a connection you cannot turn up, or a crew that cannot get in. This guide covers the deployment from the contract to the turn-up. For how the room itself is laid out, see the white space and gray space guide. For choosing the cabinets you roll in, see the cabinet and rack types guide.

Retail vs wholesale, and cabinet vs cage vs suite

Colocation comes in two commercial models and three space sizes, and the model you fall into is set mostly by how much power and space you take. Retail colocation sells by the cabinet or the cage and serves many tenants sharing the same halls and the same cooling. Wholesale colocation sells by the suite or the data hall, usually starting in the hundreds of kilowatts, and the tenant takes a dedicated, often physically separated, block of the building.

A rough industry line puts retail under about ten cabinets and wholesale above it, with wholesale deals typically committing 100 kW or more and ranging up into multiple megawatts. The split is not just size. In retail you live inside the provider's shared infrastructure and rules. In wholesale you get more control over your own hall, sometimes down to your own containment and your own power topology, and you take on more of the responsibility that comes with it.

Pick the model that matches your real footprint and growth, not the one that sounds impressive. A four-cabinet deployment forced into a wholesale-style commit strands power and money. A fast-growing platform squeezed into single retail cabinets pays a premium per kW and hits a wall on density. The provider's product names and minimums vary, so confirm what each tier actually includes and what it commits you to before you sign.

What the colo provides and what you bring

The single most important thing to settle before you sign is the line between what the provider delivers and what you own. The colo provides the physical building, the power feed to your space, the cooling for the room, the physical security, and the fire and life-safety systems. You provide the IT gear, the cabinets if they are not included, the cabling and PDUs inside your footprint, the configuration, and the operation of everything you installed.

The line gets fuzzy at the edges, and the edges are where disputes live. Who supplies the cabinet, you or the provider? Where does the colo's power stop and your distribution start, at the cabinet, at a busway tap, at a panel? Is structured cabling inside your cage yours to install or theirs to provide? Does the provider monitor your power and temperature, or only the room's? Get each of these in writing, by hand-off point, not by assumption.

Treat the responsibility split as the master document for the whole fit-out. Every later decision, the power order, the cabling, the access list, the remote-hands scope, references back to it. When something fails at two in the morning, the first question is whose side of the line it failed on, and the answer needs to already be on paper.

The power commit: you pay for kW, not for what you draw

In a colo you commit to a quantity of power in kilowatts, and that commit is the largest recurring line on most bills. You are not buying watts as you use them the way you buy them at home. You are reserving capacity. The provider builds power and cooling to serve the kW you reserved, and you pay for that reservation whether your gear draws it or not.

That reservation is delivered as circuits to your cabinet, each with a breaker rated in amps at a given voltage, and the kW falls out of the volts, amps, and how many circuits you take. A common retail cabinet might commit a few kW across one or two circuits. A dense cabinet commits far more. The number you commit is a real decision with two failure modes, and both cost money.

Over-provision and you strand power: you pay every month for capacity that sits idle, and stranded power is dead budget you cannot get back without renegotiating. Under-provision and you choke: the gear you want to add has no circuit to plug into, and adding power later means an order, a lead time, and sometimes a price that is no longer the one you negotiated up front. Size the commit to the load you can actually stand up in the term, with a deliberate, named margin, and revisit it as the deployment fills. The provider's tariff and contract control how the commit is measured and billed, so model your real ramp against their terms before you sign the number.

Circuits, A/B feeds, breakers, and the 80 percent rule

Power reaches your cabinet as circuits, usually delivered on a whip that lands at your PDU or power strip. Each circuit has a breaker rated in amps, and the usable continuous load is less than the breaker number. The common derate is 80 percent: a 20 A circuit holds about 16 A continuous, a 30 A circuit about 24 A. Size your load to the derated figure, because a circuit that rides at its nameplate trips, and a tripped breaker in a colo means dropped gear, not a reset you can reach.

Redundant deployments take two feeds, called the A and B strings, fed from separate paths in the building so that one can fail without taking your load down. Each feed lands its own circuit and its own PDU in the cabinet, and dual-corded gear plugs one cord into each. The discipline that gets missed: each string has to carry the full load alone. If you run both strings at 70 percent and one drops, the survivor jumps to 140 percent and trips, and now the redundancy you paid for is what took you offline.

Order the circuits that match your gear, the right voltage, the right amperage, the right count, single-corded or dual-corded, and the right plug type. Getting a plug or voltage wrong means a power strip that does not fit the whip on install day. The provider sets what circuit types they offer and how they are delivered, so confirm the available options and the lead time before you commit the design.

Metered power vs committed power, and what each costs you

Colo power gets billed one of two ways, and the difference decides how you should run your gear. With committed, or flat, billing you pay a fixed monthly charge for the breaker capacity you reserved, regardless of draw. There is no overage charge, because the breaker is the ceiling: exceed it and you trip, you do not get a bigger bill. With metered billing you pay for the kilowatt-hours you actually consume, usually read from a PDU or branch-circuit meter, on top of a space and capacity charge.

Each model rewards a different behavior. Committed billing is predictable and favors gear that runs near its provisioned capacity, since you are paying for the breaker either way and you want to use it. Metered billing rewards efficiency, because idle and lightly loaded gear costs you less, and it punishes leaving power-hungry equipment running with nothing to do. Neither is universally cheaper. The honest comparison is your real duty cycle against the provider's rate card.

Watch the metering detail closely. Where is the meter, at the cabinet, the circuit, or the suite? Does the rate pass through the utility cost or add a markup? Is there a power usage effectiveness factor applied so you pay for cooling overhead too? These terms live in the contract and they vary by provider, so read how power is measured and priced before you assume which model wins for your load.

The cage: a private perimeter inside a shared room

A cage is a fenced-off section of the data hall floor that gives you a private, lockable perimeter while you still sit inside the provider's shared room. The standard build is steel mesh panels, commonly a 2 in mesh in heavy-gauge wire, floor to ceiling, with a locking door and sometimes a solid lower kick or full solid walls where privacy or noise isolation matters. The mesh keeps airflow and sightlines while marking a hard boundary that only your access list can cross.

A cage buys you separation that an open or shared cabinet cannot. Your gear sits behind your own door, your own lock, and your own access log, away from other tenants' hands and eyes. That matters for compliance scopes that demand a defined physical boundary, and it matters for any deployment where you do not want a neighbor's technician working inches from your equipment.

The cage is also the part of the fit-out with the longest lead time and the highest one-time cost, because it has to be built. A cabinet you can often turn up in days. A cage takes time to fabricate, install, power, and cool, so it pushes your ready date out. Build to the provider's cage standard, since they set the mesh spec, the door hardware, the grounding, and how the cage ties into the room's containment and fire systems.

The cabinet: yours or theirs, and what comes with it

The cabinet is your smallest unit of space, and the first question is who supplies it. Some providers include a standard cabinet in the deal; others let you roll in your own to a spec they approve. Rolling your own gives you control over depth, rail type, cable management, and door perforation, which is worth it for dense or unusual gear. Taking theirs is faster and one less thing to ship. The choice of cabinet, its airflow, its depth, its weight rating, is its own subject, covered in the cabinet and rack types guide; here the point is to settle ownership and fit before install day.

In a shared hall the cabinet lock matters more than people expect. A cabinet on an open row sits where other tenants and the provider's staff walk past it, so the front and rear doors should lock, and on tighter compliance scopes you want a logged or electronic lock rather than a common key. Confirm whether the provider's master access reaches inside your locked cabinet and under what conditions.

Inside the cabinet you mount your PDUs or power strips to land the A and B circuits, and they have to physically match the whips the provider delivers. A vertical 0U strip per side is the usual layout so the strips stay clear of the gear and the airflow. Plan the strip, the plug, and the cord lengths against the actual circuits ordered, not a generic assumption.

Cooling in someone else's room: do not fight the airflow

You do not cool your space in a colo. The provider cools the room, and your job is to fit into their scheme so the cold air reaches your inlets and the hot air gets carried away. Almost every modern hall runs hot-aisle and cold-aisle, often with containment that seals one aisle from the other. Your cabinets face the right way, cold-aisle front, hot-aisle rear, so your intakes pull from the cold side and your exhaust dumps into the hot side. Spin a cabinet the wrong way and you pull your own hot exhaust back through the gear.

The detail that wrecks cooling is the holes. Every open rack space, every unused U, every gap around the rails and under the cabinet is a short-circuit path that lets hot air bypass the gear and mix with the cold supply. Fill every empty U with a blanking panel, seal the floor cutouts and brush grommets, and close the gaps at the sides. This is the cheapest, highest-return work in the whole fit-out, and it is the thing tenants skip.

Live inside the containment as it was built. Do not prop a contained aisle door, do not pull a ceiling panel to run cable, do not block a perforated tile or a chimney with stored gear. The room is balanced for the whole floor, and one tenant fighting the airflow degrades cooling for the row. The provider sets the scheme and the inlet temperature they hold to, commonly within the ASHRAE TC 9.9 thermal envelope; design your cabinet to that envelope rather than to a number you carried in from a room you used to run.

The density cap: how many kW per cabinet they let you run

Every colo has a ceiling on how much power and heat a single cabinet can take, set by what the room's cooling can carry to one footprint. That cap, in kW per cabinet, is a hard design input, not a suggestion. Pack more heat into a cabinet than the room can pull off it and the gear runs hot no matter how clean your own airflow is, because the limit is the room's, not yours.

Air cooling with good containment commonly handles into the low tens of kW per cabinet, with the workable number depending heavily on the room and the aisle design. Below roughly 8 kW most schemes cope easily. Past the high teens or twenties, the provider usually moves you into a high-density zone, a rear-door heat exchanger, or direct liquid cooling, and liquid in colo is its own arrangement with its own contract terms and its own supply and return plumbing.

AI and GPU deployments are where this bites hardest, because a single cabinet can ask for several times what a general-purpose cabinet draws. If your roadmap includes that density, confirm up front whether the facility can deliver it, in which zones, and at what cooling, because a hall built for 5 kW cabinets cannot absorb a 40 kW one by wishing. The provider's published per-cabinet limit and high-density options govern; design to them before you order the gear, not after it arrives.

Structured cabling inside your space

Inside your cabinet or cage you install your own structured cabling, and the colo expects it done to a standard that keeps the shared room clean. Power and data run overhead in most halls, on ladder rack and in fiber raceway above the cabinets, with drops to each one. Where the provider runs ladder rack to your cage entrance, your job is to extend it cleanly to each cabinet and keep the bundles dressed, separated by power and data, and out of the airflow.

Label everything, both ends, to a scheme you write down. In a facility you do not staff full time, a cable nobody can identify is a cable nobody can safely move, and the remote-hands technician who has to trace it is billing you by the task. Good labeling and a current patching record are what let someone who has never seen your rack work on it correctly over the phone.

The cross-connect to the outside world is a separate animal from your in-cage cabling, covered next. Inside your footprint, follow the provider's containment and pathway rules, since how cable is run overhead and where it can penetrate is set by the room design. For how those pathways and the floor grid are organized at the building level, see the white space and gray space guide.

The meet-me room, where you reach the outside world

The meet-me room, the MMR, is the secure room in the building where carriers, networks, and tenants terminate and interconnect. You almost never get into it. Instead, the colo runs your connections from your space to the MMR and patches them through to whoever you are connecting to. The MMR is how a single cabinet in the corner of a hall reaches a carrier, a cloud, or another tenant without anyone trenching fiber across a parking lot.

Your connection out of your space lands at a demarcation point, the demarc, where your responsibility ends and the provider's interconnection begins. From the demarc, a cross-connect carries the circuit to the carrier or tenant in the MMR. You order the cross-connect from the colo, they run and patch it, and you pay an install charge plus a recurring monthly fee for as long as it lives.

Plan the MMR connections before you deploy, not after. A common, expensive surprise is standing up the gear, then discovering the cross-connect to the carrier takes weeks to provision and carries a recurring fee nobody budgeted. The MMR process, the order steps, the lead time, the demarc location, and the fee schedule are set by the provider, so get the connectivity plan and its costs in hand alongside the power and space.

Cross-connects: the order, the LOA, and the recurring fee

A cross-connect is a physical point-to-point cable, fiber or copper, that links two defined endpoints inside the building, your cage and a carrier's rack, your cage and another tenant, your cage and a cloud on-ramp. It is the unit of interconnection in a colo, and it is how you build connectivity without owning any of the path between the two ends.

Ordering one usually needs a Letter of Authorization, an LOA, from the party you are connecting to, telling the colo which port to patch you into. You submit the order, the provider runs the cable through the MMR, and the connection turns up. Copper handles lower-speed links; fiber carries the higher speeds and the longer reaches. The cost is an install charge plus a monthly recurring charge that scales with speed and type, from modest for a low-speed copper link to several times that for high-speed fiber.

Cross-connect fees are quiet and they compound. A heavily interconnected deployment can carry dozens of them, and the recurring total becomes a real line on the bill that nobody flagged at signing. Inventory every cross-connect, what it connects, why, its speed, and its monthly cost, and prune the ones that stopped earning their keep. The provider sets the LOA process, the lead time, and the fee schedule, so confirm them per connection.

Carrier-neutral facilities and cloud on-ramps

A carrier-neutral colo lets you connect to many carriers and networks rather than locking you to one. Connectivity is not bundled into the space contract; you choose your carriers on their own merits and reach each one with a cross-connect through the MMR. That choice is one of the main reasons enterprises pick colo over a single-carrier facility, because it keeps pricing competitive and gives you a second path when one carrier has a bad day.

The same room often hosts cloud on-ramps, the dedicated private connections into AWS, Azure, Google Cloud, and Oracle, marketed as Direct Connect, ExpressRoute, and the like. A cross-connect to an on-ramp gives you a private, predictable path to the cloud instead of riding the public internet, with steadier latency and a cleaner security boundary. For a hybrid deployment, sitting in a facility that already has the on-ramps you need can be the deciding factor.

The depth of the interconnection ecosystem in a building is a real asset, because the carriers and clouds you can reach with a short cross-connect are the ones you will actually use. Confirm which carriers and which on-ramps are present in the specific facility, not just somewhere in the provider's portfolio, before you assume you can reach them.

Security and access: badges, biometrics, and the access list

Physical security in a colo is the provider's to run, and it is layered: a fenced site, a staffed entrance, badge and often biometric access at each door, cameras through the building, and a recorded log of who went where and when. You inherit that perimeter, and your part is the access list, the named people allowed to reach your cabinet or cage, which the provider enforces at the door and at your lock.

Keep the access list tight and current, because it is the part of the security model you actually control and the part auditors check. Names that should have come off when someone left the team are exactly what a review flags. Decide your escort policy too: whether your own staff need an escort, and whether vendors and delivery crews can enter your space at all or only under provider supervision.

Treat the access log as evidence, not paperwork. When a compliance review or an incident asks who touched the gear, the colo's badge and camera records are the answer, and they only help if your access list was accurate. The provider sets the access control technology and the audit support; confirm what they log, how long they keep it, and how you request and revoke access for your people.

Compliance: what you inherit and where the line falls

Colo providers carry compliance attestations on the facility, commonly SOC 1 and SOC 2, ISO 27001, PCI DSS, and HIPAA support, and you can use those to cover the physical and environmental parts of your own compliance scope. This is the inherited-controls idea: the provider's audited physical security, power, and environmental controls satisfy portions of your framework so you do not have to prove the building yourself.

Know exactly where the inheritance stops. The provider's SOC 2 covers their controls, the facility, the access, the power and cooling. It does not cover your servers, your operating systems, your data, or your application. Most frameworks are a shared-responsibility model, and the colo holds up one end. Get the provider's current attestation reports and the bridge letter, read the scope, and map which of your controls you inherit and which stay yours.

The certifications a facility holds are a real reason to choose it for a regulated workload, since rebuilding that physical evidence yourself is expensive and slow. Confirm the attestation is current, covers the specific site you are deploying into, and matches the framework you answer to, because compliance scopes and report editions change and the provider's audit cycle is the authority on what is in force.

Remote hands and smart hands when you are not on site

Remote hands are the colo's technicians doing physical work in your space when you cannot be there, which in a facility hours from your office is most of the time. The lighter tier, often called remote hands, covers simple physical tasks: power-cycling a server, checking a cable, reading a status light, swapping a labeled drive. The heavier tier, smart hands, covers technical work: configuration, troubleshooting under direction, installing or moving gear.

The two tiers price very differently, and the gap is real. A simple reboot might be a flat per-task charge or a low hourly rate, while the same job handled as smart hands, with a technician working under your direction, can cost several times that. Emergency and after-hours response carries a premium on top, and the fastest response tiers can run into four figures per incident plus a retainer.

Remote hands only work as well as your documentation. A technician who has never seen your rack can power-cycle the right device, reseat the right cable, or swap the right drive only if your labeling and your runbook are accurate. The scope of what each tier includes, the response time, and the rate are set in the contract and the SLA, so confirm them and write your runbooks to match what the colo's people are allowed to do.

The SLA: uptime, power, response, and credits

The service level agreement is the document that defines what the provider owes you and what you get when they miss. Read it as the operating contract for the deployment, because every promise about uptime, power availability, environmental conditions, and response time lives here, along with the remedy when a promise is broken. The remedy is almost always a service credit, a percentage of your monthly fee, and the credit is usually the cap on the provider's liability.

Read the definitions, not the headline number. A facility advertising high power availability defines downtime in a specific way, and the redundant-power clause often pays a credit only when both the A and B feeds fail at once, not when one drops. Availability is measured over a stated period, exclusions carve out maintenance windows and tenant-caused outages, and the credit you can claim has a ceiling and a claim deadline. The marketing number and the enforceable number are rarely the same.

Hold the SLA terms against your real requirement, and do it before you sign. Ask how downtime is defined, what triggers a credit and what the credit caps at, what maintenance windows are excluded, what the response-time commitment actually guarantees, and how you file a claim. The provider's contract and SLA are the authority on every one of these, and the credit you can recover is almost never the cost of an outage to your business. Build your own redundancy and recovery on top of the SLA; do not treat the credit as insurance.

The deployment process, order to turn-up

A colo deployment runs in a known sequence, and the lead times stack, so the date you can run production is set by the longest pole, not the shortest. The order: contract and power commit signed, space provisioned, power and cross-connects ordered, gear shipped and staged, gear installed and cabled, power and connectivity turned up, and access set for your people. Each step gates the next, and several of them carry weeks of lead time you cannot compress.

Sequence the long-lead items first. A cage has to be built before anything racks in it. Cross-connects can take weeks to provision through the MMR. Custom power circuits and high-density cooling carry their own lead times. Order those the day the contract is signed, not the week before the gear ships, or the equipment sits staged and idle while you wait on a circuit or a cross-connect.

The turn-up is the moment it all has to be real at once: power live on both strings, cross-connects patched and tested, the gear reachable, and your access working at the door and the lock. Plan a walkthrough that confirms each one before you call it production. For rolling and securing the gear into the rack itself, the rack-and-stack mechanics, lean on the cabinet and rack types guide.

The colo's install rules and conduct in a shared room

Every colo has install rules, and they are not optional. Work in the room often requires a scheduled work order or permit, an approved crew on the access list, and sometimes provider supervision. Material staging is controlled: there is a place to uncrate, a limit on how long you can stage, and a rule that you carry your trash and packaging out rather than leaving it on the floor. Cardboard and pallets in a data hall are a fire and contamination problem, which is why providers are strict about them.

Conduct in a shared room is part of the deal. You are working feet from other tenants' live gear, so you do not touch what is not yours, you do not prop a contained aisle door, you keep noise and disruption down, and you follow the photography and tooling rules. A crew that bypasses the rules can get the whole tenant's access flagged.

Read the provider's installation guidelines before the crew shows up and brief them on it. The rules on staging, permits, trash, hot work, and conduct are the provider's to set and they vary by facility, so confirm the specifics for the building you are entering rather than assuming they match the last colo you worked in.

Scaling inside the colo

Growth in a colo means more cabinets, a bigger cage, or more power per cabinet, and the constraint is rarely your willingness to pay. It is whether the contiguous space and the power are available next to where you already sit. A cabinet two rows away is not the same as a cabinet adjacent to your cage, because it splits your footprint, complicates your cabling, and may sit outside your secured perimeter.

Plan for growth at signing if you can see it coming. Some providers will reserve adjacent space or hold an option on the next cage, which keeps your deployment contiguous as it grows. Without that, you take what is open when you need it, and what is open may be across the room. Power is the same story: adding kW to an existing cabinet depends on the room's spare capacity and cooling, and a hall that is full cannot give you density it does not have.

Lead time is the recurring tax on growth. More cabinets, a cage expansion, additional circuits, and new cross-connects all carry the same provisioning clocks as the original fit-out. Forecast the ramp and order ahead of the need, because the colo cannot conjure contiguous space or power on the timeline of a surprise.

Exit and migration: getting out cleanly

The end of a colo deployment is a project of its own, and it is easiest if you planned it at the start. The contract has a term and a notice period, and missing the notice window can auto-renew you into another term you did not want. Read the renewal and termination clauses when you sign, not when you are trying to leave.

Migrating out means moving the workload before you move the gear, decommissioning the equipment, removing your cabling and cross-connects, and handing the space back in the condition the contract requires. The order matters: you keep power and connectivity live until the workload has landed at the new location, then you decommission. Cross-connects and circuits have to be formally cancelled, or you keep paying for connections to gear that is gone.

Handle the data and the space hand-back deliberately. Wipe or destroy drives to your own standard before they leave a secured facility, get a certificate where you need one, and confirm what the provider requires for the space to be returned, whether you remove a cage and restore the floor or hand a cabinet back empty. The decommission and hand-back terms are in the contract, so confirm them before the move so the exit does not turn into a holdover bill.

What to record for the deployment

A colo deployment lives in its records, because almost no one is standing in the room to see how it was built. The power commit, the circuits, the cross-connects, the access list, and the contract terms are the facts a remote-hands technician, an auditor, or your own on-call needs months later, and the only place they exist is what you wrote down. A field record tool like FieldOS keeps the power, cabling, cross-connect, and access details with the deployment so the person working the rack over the phone is working from the truth.

Capture the space and model, the committed kW and the circuits with their amperage and plug type, the A/B feed assignment, the per-cabinet density cap you designed to, every cross-connect with its endpoints and monthly fee, the access list, the remote-hands scope, and the SLA and contract dates. When the deployment changes, a circuit added, a cross-connect cancelled, a name removed from the access list, update the record in the same step, because a stale colo record is worse than none.

Common mistakes

• Over-provisioning the power commit and paying every month for stranded capacity, or under-provisioning and hitting a wall when there is no circuit to add gear to.
• Running both A and B feeds above 50 percent so a single feed failure overloads the survivor and trips, defeating the redundancy you paid for.
• Fighting the room's airflow: cabinets faced wrong, missing blanking panels, open floor cutouts, or a propped containment door.
• Exceeding the per-cabinet density cap and running gear hot because the room cannot pull the heat off one footprint.
• Standing up the gear with no cross-connect plan, then waiting weeks to turn up connectivity and eating unbudgeted recurring fees.
• Reading the SLA's headline uptime number instead of the definitions, exclusions, and credit caps that actually apply.
• Ignoring the colo's install rules on permits, staging, trash, and conduct, and getting the crew's access flagged.

Field checklist

Standards and references

The first authority on a colo deployment is not a published standard. It is the provider's rules and your contract. The power commit and how it is billed, the per-cabinet density cap, the cooling scheme you have to fit into, the cross-connect and MMR process, the install rules, and the SLA are all set by the specific provider for the specific facility, and they override any general number you carry in. Hedge every one of those to the contract and confirm the current terms.

Where published standards apply, they are mostly about how the facility is rated and how the room runs. The Uptime Institute Tier standard and the ANSI/TIA-942 rating system both classify facility resilience on a four-level scale, and a provider may certify against one or the other; they measure different things and are not interchangeable, so read what the certification actually covers. The ASHRAE TC 9.9 thermal guidelines define the temperature and humidity envelope most halls hold to, which is the target you design your cabinet inlets against. Structured cabling and pathway practices follow TIA and BICSI conventions, applied within the provider's room design.

Three things decide whether a colo deployment runs clean. Commit the right power and cool within the provider's scheme. Cross-connect through the MMR and stay under the density cap. Live by the SLA and the colo's install rules. Each of those bends to the provider, the contract, and the design, so verify the specifics for your facility rather than trusting a generic figure.

Units and terms

Colo carries its own vocabulary, and the same idea shows up under different names across a contract, a tariff, and a provider's product sheet. The definitions below are the ones that decide money and uptime.

Power is committed and billed in kilowatts (kW) for capacity and kilowatt-hours (kWh) for metered consumption. Circuits are rated in amps (A) at a voltage, and the usable continuous load is about 80 percent of the breaker. Cross-connect and space charges carry a one-time install charge and a monthly recurring charge (MRC).

Colocation (colo)

Renting space, power, and cooling in a provider's data center while you own and operate the gear inside it

Retail vs wholesale colo

Retail sells by the cabinet or cage in shared halls; wholesale sells suites or data halls, usually 100 kW and up, with more tenant control

Cage vs cabinet vs suite

A cabinet is a single enclosure; a cage is a fenced private area on the shared floor; a suite is a dedicated room

Power commit (kW)

The power capacity you reserve and pay for, whether or not your gear draws it

Metered vs committed billing

Metered bills the kWh you actually use; committed bills a flat charge for the reserved breaker capacity

Meet-me room (MMR) / cross-connect

The MMR is where carriers and tenants interconnect; a cross-connect is the physical cable linking two endpoints through it

Remote hands / smart hands

Provider technicians doing physical work (remote) or technical work (smart) in your space when you are not on site

SLA

The service level agreement defining uptime, power, response, and the service credit owed when the provider misses

Related tools

Calculators and readiness checks for this work

Compare your options

FAQ

People also ask

• What is white space in a data center?
• What is a rack unit (U)?
• What is a meet-me room?
• What are the types of data centers?
• Cast-in vs post-installed anchors: which should I use?