So you have to build out a new facility

At some point in an infrastructure career, somebody is going to hand you a floor plan.

It might be an empty warehouse, a new office, a manufacturing facility, a sales center. Maybe it’s new construction and the walls don’t exist yet. Maybe somebody signed a lease on an existing building and construction starts next month. Either way the conversation usually sounds like “we need IT to get the building ready.”

If you’ve never done one before, it’s easy to hear that as a networking project. You need some switches, a firewall, Wi-Fi, Internet circuits and a server room. Done. Except you have just become involved with electrical, HVAC, construction, structured cabling, Internet carriers, architects, general contractors, low-voltage contractors, facilities, security, AV, fire protection, and probably several other groups that haven’t introduced themselves yet.

I’ve been involved in building out hundreds of facilities — offices, healthcare, finance, legal, manufacturing, warehousing, logistics, engineering. Some were a few rooms and a network rack. Others involved large server rooms, multiple IDFs, manufacturing floors, warehouses, redundant carriers, hundreds of drops, and wireless designs that had to support everything from laptops to handheld RF scanners. The technology changes between those buildings. The process really doesn’t.

This is the guide I’d give someone getting thrown into their first one.

1. Get the drawings
2. Design the MDF and server room
3. Figure out whether you need IDFs
4. Plan power and electrical
5. Calculate cooling and HVAC
6. Design the structured cabling plant
7. Order Internet circuits early
8. Design the LAN and fiber backbone
9. Design Wi-Fi with Ekahau
10. Plan for everything else that uses the network
11. Coordinate the trades
12. Walk the site
13. Commission and document everything

Before any of it, though, there’s one idea that reorders how you think about the whole project: you’re not designing a network inside a finished building. You’re helping design the building that the network will eventually live in.

1. Get the drawings

Start here. Before I choose a switch, an access point or a rack, I want the architectural drawings. At minimum dimensioned floor plans, and depending on the project, reflected ceiling plans, electrical drawings, mechanical drawings and furniture layouts become useful very quickly. Get the actual CAD files when you can, not a PDF somebody exported six months ago.

The floor plan becomes the common reference for nearly everything that follows. Where is the MDF? Where could IDFs go? Where are the offices, the conference rooms, the manufacturing lines, the warehouse racks, the loading docks? Where are the printers, kiosks, TVs, cameras and access-control panels going? Where do the carriers enter the building?

You will eventually put an alarming number of colored dots on this drawing. Keep it.

2. Design the MDF and server room

For most facilities I’ve designed, the MDF and the server room are the same room, and it’s the center of the building’s infrastructure. Carrier demarcations terminate there. The core lives there. Fiber to the IDFs originates there. Depending on the facility, so do servers, storage, UPS systems and security infrastructure. Which makes choosing that room one of the most consequential decisions in the whole design.

I want it reasonably central, but central isn’t the only question. Can the carriers reach it? Can sufficient electrical service reach it? Can we cool it? Is there plumbing above it? Can equipment physically get into it — not just the rack, but the pallet it arrives on? Is the room large enough to work behind the racks? Can another rack be added later? Does the door swing into space you needed?

And perhaps most importantly: does everyone understand that this is not a storage closet? Server rooms have an extraordinary ability to accumulate folding chairs, Christmas decorations, marketing materials and six boxes of toner. Treat it as dedicated infrastructure space from the first meeting, because the precedent gets set early and it is much harder to reclaim the room later.

Then lay it out properly. Where are the racks, the ladder rack or overhead tray, the electrical panels and receptacles, the UPS systems? Where does carrier equipment mount? Where does the fiber come in, and the copper bundles? Where does the cooling equipment go?

Don’t design the room around today’s twelve rack units of equipment. Buildings have much longer lifecycles than switches do. Empty rack space is cheap during construction; finding another six square feet five years later is not.

3. Figure out whether you need IDFs

Now start looking at distance, because your horizontal copper plant has limits and the building doesn’t care about them. Once runs from the MDF get too long, you need another telecommunications room — an IDF. The MDF stays the center of the network while one or more IDFs extend the access layer out into the rest of the facility.

This comes up constantly in warehouses and manufacturing plants. On a floor plan a building can look perfectly manageable, and then you scale the drawing and realize the far corner is several hundred feet away. Your cable isn’t flying diagonally across open space either — it follows trays, corridors, structural pathways and vertical transitions, and all of that counts.

Don’t design right up against the maximum distance. Leave margin.

An IDF needs most of the same considerations the MDF did: space, cooling, electrical service, UPS, physical access, room to grow. Then connect it back to the MDF with fiber, and pull more strands than you need today. Fiber is inexpensive while the contractor is already standing there with a spool. It gets considerably more expensive when everyone has to come back.

4. Plan power and electrical

You’re now going to spend some quality time with the electrician, and the single most useful thing you can do is give them actual requirements rather than “we need power for the racks.” What voltage, what amperage, how many circuits, what receptacles, and where? Which equipment needs redundant power? How large are the UPS systems? Are there separate A and B feeds? Does the facility have generator power, and which circuits sit behind it?

You need to calculate this. Servers, storage, firewalls and switches all consume power, and modern PoE switching can consume a lot of it. The switch itself may not be electrically interesting until you hang dozens of access points, phones and cameras off it, at which point the power budget matters quite a bit more than the spec sheet suggested.

One thing worth internalizing early: redundant power supplies don’t create redundant power. If PSU 1 and PSU 2 eventually land on the same breaker, you have two power cords, not two power paths. Sometimes genuine electrical redundancy isn’t practical for the building, and that’s a legitimate outcome — just make sure you know which failure domains actually exist rather than which ones the equipment implies.

5. Calculate cooling and HVAC

Everything you just powered is going to turn most of that electricity into heat, so now you’re talking to HVAC. You can estimate the equipment heat load from the expected electrical load and translate that into the cooling capacity the room needs, but don’t stop at the BTU calculation. Ask how the building’s HVAC actually operates.

Does it shut down at night? Does it change setpoints on weekends? Is the server room on the same zone as occupied office space? Where is the thermostat? What happens on Christmas Day, when nobody is in the building but every switch and server is still running?

A vent in the ceiling does not constitute a server-room cooling design. For rooms with meaningful equipment loads, dedicated cooling usually earns its cost. The requirement is that the room stays inside its environmental limits regardless of whether Karen from accounting is currently at her desk. The servers do not know that the building is closed.

6. Design the structured cabling plant

Now bring in the low-voltage contractor, and start covering those floor plans in dots. Every drop should be planned: desks, printers, conference rooms, access points, cameras, phones, TVs, digital signage, kiosks, manufacturing equipment, warehouse equipment, building-management systems, access control, and anything else with an Ethernet port.

My rule is simple. If it doesn’t move and running a cable is reasonable, run a cable. Wi-Fi isn’t a substitute for structured cabling; it’s another access medium, and it still needs cable to reach every access point.

You also need to define the cable specification, the patch panels, racks, pathways, cable management and the labeling standard. That last one matters more than it sounds like it does. MDF-PP01-24 tells me something. “Jack 73” tells me that somebody, somewhere, probably has a spreadsheet. Label both ends, label the faceplate, label the patch panel, and make the documentation agree with the labels.

Then require certification results — copper certified to the appropriate standard, and the fiber tested too. The person who installed the cable should not be the only person on earth capable of working out where it goes.

7. Order Internet circuits early

Ignore the section number on this one. Do it as soon as you know the address.

Carrier timelines can be ridiculous, and a building can have furniture, computers, racks, switches and perfectly functional Wi-Fi while still not being operational because somebody is waiting on a construction crew to bore underneath a parking lot. None of your work saves you from that; only lead time does.

Figure out which carriers actually service the facility. Get quotes. Find out what construction is required, where the demarcations will be, and how the circuits will physically enter the building. Make sure conduit exists from the carrier entrance to the MDF — that is a cheap thing to specify during construction and a genuinely expensive thing to add afterwards.

If you’re ordering redundant carriers, investigate whether they’re physically redundant. Two ISP invoices do not necessarily mean two independent paths. If both circuits enter through the same conduit, use the same pole line, or ultimately ride the same last-mile infrastructure, your logical redundancy can disappear with one backhoe. Sometimes real path diversity isn’t available at that address, which is useful information in its own right. The point is to know which one you’re buying.

8. Design the LAN and fiber backbone

Now we finally get to buy some switches. Notice how far down the list we are.

By this point you should understand the physical topology of the building well enough that the logical topology mostly falls out of it. The MDF is the center, IDFs connect back over fiber, and access switching lives wherever it has to live to serve the cable plant. From there you’re working out uplink capacity, redundancy, switch counts, port density and PoE requirements.

Count your ports — then don’t buy exactly that many. The facility will change roughly seventeen minutes after opening. Someone will add cameras. Someone will move a printer. Facilities will install a controller nobody mentioned. A department will hire twenty people. Leave capacity, and apply the same logic to fiber: a six-strand cable may satisfy the drawing today, but more strands are cheap insurance against a building that has every intention of changing.

9. Design Wi-Fi with Ekahau

Do not design Wi-Fi by spacing circles evenly across a floor plan, and don’t let the electrician choose access point locations because those spots were convenient for pulling cable.

I use Ekahau for both predictive design and measured surveys, and the predictive design comes first. Import the floor plan, scale it correctly, draw the walls and assign realistic building materials — concrete, drywall, glass, block and metal all attenuate RF differently, and if you’re designing a warehouse, model the racks too. Then define the actual requirements: coverage, capacity, client types, applications, expected density, roaming behavior.

Only then place the APs. Ekahau lets you model the expected RF environment before a single access point is mounted, so you can evaluate signal strength, secondary coverage, channel relationships and capacity, and adjust while adjusting is free. It also gives the low-voltage contractor exact AP locations, which matters, because “somewhere around here” has an amazing tendency to become “directly above the new HVAC duct.”

But predictive design is still predictive, and eventually physics gets a vote. Once the building is substantially complete and the network is running, go back with Ekahau and perform a measured survey. Walk the building — actually walk it. Offices, conference rooms, hallways, production floors, warehouse aisles, loading areas, and the weird corner behind the machine where nobody thinks anyone needs Wi-Fi until someone puts a scanner there.

Now you’re measuring the building that exists rather than the one that was drawn. Maybe the contractor used a different wall material. Maybe the shelving changed. Maybe machinery was added. The predictive survey was the design; the measured survey is the verification. You need both.

A note about warehouses

Warehouses deserve special attention, because an empty warehouse and a full warehouse are completely different RF environments — and so are the top and bottom of the same aisle. Racks full of liquid, metal components or dense product will change propagation in ways the empty building never hinted at.

Think about where the clients actually operate. If employees use rugged handheld scanners three feet off the floor between racks of product, that’s the environment I care about — not what my laptop sees while I’m standing six feet tall in the middle of an empty aisle. Client radios aren’t equal either. The shiny new laptop you tested with may have a significantly better radio than the ten-year-old rugged scanner the business actually depends on. Design for the clients that matter, and measure where those clients work.

10. Plan for everything else that uses the network

The network has quietly become the nervous system of the building, which means everyone eventually wants something from it: security cameras, badge readers, door controllers, HVAC controls, lighting controls, AV and conference rooms, digital signage, time clocks, printers, manufacturing equipment, warehouse scanners, guest Wi-Fi, assorted IoT, and vendor equipment of every description.

Some of those systems belong to IT. Some belong to facilities. Some belong to security. And some belong to an outside vendor who will tell you they need “an Internet connection,” then arrive with six devices, a switch of their own and a requirement for three static IP addresses nobody mentioned. Ask early, and ask specifically.

11. Coordinate the trades

This may be the most important part of the entire project, because your design intersects everyone else’s: the architect, general contractor, electrician, HVAC contractor, low-voltage contractor, carriers, security vendor, AV contractor, facilities, fire protection, furniture installers.

Sometimes all of them are working from slightly different revisions of the same building. That’s how the electrical drawing ends up showing a rack in one location while the low-voltage drawing shows it six feet away and the mechanical drawing puts an air handler above both of them. Nobody necessarily made a mistake. Their drawings just never met.

Make them meet. Review drawings, attend construction meetings when it’s appropriate, answer RFIs, mark up plans, ask questions. Make sure everyone understands where the infrastructure is going before somebody drills, pours, frames or closes something expensive to change. Catching a collision in Bluebeam costs essentially nothing. Catching it after drywall costs considerably more.

12. Walk the site

Then get out of the drawings. I like walking these projects several times.

Early on you’re validating assumptions. Where is the carrier entrance, really? Where is electrical service? What does the MDF actually look like now that it exists? Are the proposed IDF locations reasonable in person? Where will the pathways run?

During construction, walk it again. Look at the cable trays and the conduits. Look at the MDF and every IDF. Look at where AP cables are actually landing. Check rack clearances. Find the things that aren’t being built quite the way the drawing suggested they would be. This is normal — buildings are messy, and that’s exactly why you’re there.

13. Commission and document everything

Eventually somebody declares the building finished, and IT should verify that claim independently.

Check the racks, the power, the UPS systems, the cooling. Check the fiber and review the copper certification results. Verify patch-panel labeling and the carrier demarcations. Test failover by actually causing one. Check every IDF. Verify AP locations, run the measured Ekahau survey, and test roaming with the devices the business genuinely uses rather than the one in your bag. Test the awkward places: the loading dock, the conference rooms, the far end of the warehouse.

Then update the documentation to reflect what was built, not what Revision 7 of the construction drawing said would be built. Keep the floor plans, the rack elevations, the cable schedules, the fiber documentation, the Ekahau project, the circuit information, carrier contacts and account numbers, UPS details, switch locations and serial numbers. Document the things Future You would otherwise spend four hours rediscovering.

You’re building more than the network

The first time somebody asks you to build out a facility, the sheer number of moving pieces is intimidating. That’s normal, and it passes.

You don’t need to be an electrician, HVAC engineer, architect, general contractor and low-voltage technician simultaneously. You need to understand where their work intersects yours. You need to tell the electrician what the infrastructure requires, tell HVAC how much heat you’re expecting, tell the low-voltage contractor where cables need to go, tell the carrier where you need the demarc, and tell the architect how much room the MDF needs. And sometimes you need to get all five of them standing around the same floor plan at the same time. That turns out to be a surprisingly large part of infrastructure architecture.

Eventually construction ends. The racks fill up, the switches boot, the access points come online, the users move in. And if you did all of this correctly, almost nobody will notice. They’ll plug in a laptop and get a network connection. They’ll walk through the building and Wi-Fi will roam. A scanner will work at the far end of the warehouse. The server room will stay cool at two in the morning. A failed ISP will disappear behind the other circuit. And thousands of feet of cable will sit quietly above the ceiling doing exactly what it was designed to do.

That’s the goal. Because the best time to solve an infrastructure problem is while the entire building is still just lines on a drawing.