Home › Topics › IoT, Cameras & Building Systems › Building Systems on IP

HVAC, Access Control, and Elevators on Your Network


It usually starts with a sentence: "The HVAC contractor is here. He says he just needs a port." Or you find it afterwards: a discovery sweep turns up a device you have never seen. Three phone calls later it turns out to be the new elevator controller. It was installed during a modernization project nobody told IT about and has been quietly online for two months. The heating, the doors, the elevators, the replacement for the fire panel's dialer: these systems used to live on their own wiring. They are moving onto IP, one contractor visit at a time. I have been told a box "just needs a port" far more often than I have been told what the box does.


This is not an invasion. It is how buildings work now. Facilities systems are bought by facilities and installed by mechanical and security contractors. They are specified by vendors for whom the network is a hole in the wall their box plugs into. Nobody in that chain is thinking about VLANs (virtual LANs, the separate network segments a switch can carve out), monitoring, or what happens when the box cannot reach its cloud. That part of the thinking arrives with the cable, and lands on you.


Done badly, this ends in the worst kind of outage: the building itself misbehaving, with IT and facilities pointing at each other. Done well, it becomes one of the most productive relationships in the organization. This guide covers the whole arc. It covers the intake conversation to have before (or, realistically, after) a system lands. It explains segmentation that respects what these systems are and availability monitoring for the building's brains. It covers vendor remote access without the permanent back door. And it explains the coordination habits that keep a firmware update from becoming a mystery outage.


In this guide

  • How building systems arrive
  • The intake conversation
  • Segmentation and firewalling
  • Monitoring the building's brain
  • Vendor remote access, honestly
  • Change coordination
  • A worked weekend: the controller that went quiet
  • The payoff: one shared dashboard

How building systems arrive


Understanding the arrival pattern explains most of the problems. A building system purchase is a facilities project. The budget line says "chiller replacement" or "door hardware upgrade." The vendor is a mechanical or security integrator. And the network requirement appears late, small, and vague. Often it is literally the phrase customer to provide network connection buried in a statement of work. The installer's job is to make the equipment run, and the fastest path to that is any live port and any address that works. If DHCP (the service that hands out addresses automatically) offers one, wonderful. If not, the installer types in an address that "looked free." That is how a boiler controller ends up squatting on the IP your print server had reserved.


The result, multiplied across projects and years, is a shadow population: controllers, panels, and gateways doing physically important jobs. They are unknown to your documentation, unmonitored, sharing subnets with your laptops, and reachable by anything on the network. Each one also brings a support model designed for a different century. The vendor assumes they will drive out when something breaks. Or increasingly, they assume they can tunnel in remotely through whatever path the installer left behind.


You will not stop building systems from arriving, and you should not want to — networked HVAC genuinely runs buildings better. What you can change is whether they arrive through a door you operate. That means two things. First comes a discovery habit that finds what appeared without asking (regular sweeps, reconciled against your inventory). Second comes an intake conversation that turns "he just needs a port" into fifteen minutes of questions. The conversation is the cheap part, and it is next.


The intake conversation


Every new system that touches the network deserves a structured intake. The general version of that ritual is covered in the new-system intake conversation. Building systems need a facilities-flavored variant of it. The intake has three headline questions, each unpacking into practical detail:


QuestionWhat you are really askingWhy it matters later
Who owns it?Which person in facilities answers for this system; which vendor services it; where the support contract livesAlerts and call-outs need a named person; at 2 AM with the heat stuck on there is no time to find one
What does it need?How many devices and addresses; what it must reach (cloud services, internal servers); what must reach it; expected traffic; vendor remote access; PoE or local powerThese answers become the firewall rules, the DHCP reservations, and the monitoring checks
What happens when it breaks?The physical failure mode: do doors fail open or locked, does HVAC coast or stop; who gets called, in what order; is there a manual fallbackDetermines alert urgency and routing — and whether a network outage is an inconvenience or a safety event

Ask the installer these questions while they are on site and the answers are free. At two in the morning with the heat stuck on, "whose system is this?" is not a fun research project. The third row deserves special respect: fail-open versus fail-secure is the kind of detail that seems academic. Then the day comes when the access-control network segment goes down. You need to know, immediately, whether the building is currently unlocked or whether people are locked out. The installer knows. Write it down while they are standing there.


One more intake output: addresses. Building systems live for decades and vendors configure them by IP. So give controllers fixed addresses or DHCP reservations from a range set aside for facilities gear. Record every one in the inventory with its owner and vendor. Future-you, staring at a mystery address in a firewall log, will send thanks backward through time.


Segmentation and firewalling


Building systems earn their own network segment for reasons that go beyond general IoT hygiene. Their lifetimes are measured in decades, so whatever software they run will be old for most of its life. Patching is on the vendor's glacial schedule. You must assume the devices are permanently unpatchable and design accordingly. Their protocols are often chatty broadcast-happy building-automation dialects that have no business sharing a segment with office traffic. And their compromise consequences are physical: a laptop infection is a bad week, but a hijacked door controller is a different category of problem. The broad strategy for fencing off and watching untrusted device populations is covered in monitoring guest and IoT networks; building systems apply it with extra firewall discipline.


Concretely: put facilities systems on their own VLAN (or several). Access control and HVAC do not need to see each other. And the elevator vendor certainly does not need a path to the door controllers. Default-deny at the boundary (block everything unless a rule allows it), then allow exactly the flows the intake documented. Those are the HVAC controllers to their vendor's cloud service and the access panels to the badge server. They also include the workstation in the facilities office to the controllers it manages. Everything else — especially paths from the general office network into the building segment, and paths from the building segment into your servers — stays closed until someone documents a reason. When the vendor's list of required flows is vague ("the controller needs internet access"), push back politely for names and ports. You will get them. The difference between "outbound to two named services" and "outbound to everywhere" is most of your security posture.


At Ashgrove Housing a routine sweep of the server VLAN turned up an address nobody could account for, answering on a web port. It was the heating controller for the main office, installed three years earlier. The contractor had been told to use any free port in the comms room, and had. The controller's web interface accepted the default login printed in the manual. It had been doing so, on the same segment as the domain controllers and the finance server, for the whole three years. The review described this as "an exposure". The controller now lives on a facilities VLAN behind two firewall rules, and the sweep runs monthly.


Warning: never insert filtering between a controller and the equipment it commands without the vendor in the room. That includes between an access panel and its door hardware, or a fire panel and anything. Those links can have real-time and life-safety behavior that does not tolerate creative networking. Segment around building systems freely; segment inside them only with the people who understand what the wires do.


Monitoring the building's brain


Every building system has a brain. The HVAC plant has a supervisory controller. Access control has one or more panels and a head-end server (the central server the panels report to). The elevators have a machine-room gateway that links them to the vendor's service. These brains are classic sealed devices — no agent, no login for you, vendor firmware. The full outside-in toolkit for that situation is covered in monitoring the unmanageable. The point here is simpler and firmer: the building's brain deserves an availability check even though you cannot see inside it. A ping or port check every few minutes, an alert when it goes quiet — that is the whole ask, and it is astonishing how often it is nobody's job.


The payoff is time. Building systems fail slowly and physically. When the HVAC controller drops off the network on Saturday, the building coasts on its last instructions. The symptom — a hot server room, a cold office — arrives hours or days later, disguised as a comfort complaint. An availability alert converts that into "the controller went unreachable at 14:07 Saturday," delivered while the coasting is still comfortable. Same for access control: panels typically keep working autonomously when they lose their server, using cached badge lists. In that state, doors still open, but new badges do not get provisioned and events stop being logged. That degraded state is invisible by design, which is exactly the kind of failure only monitoring ever catches.


Route these alerts by the intake sheet: the controller-down alert goes to you and to the facilities owner, day one. In practice the whole monitoring plan for a building segment fits on a napkin, and it should:


building segment — availability checks
hvac-supervisor      ping, every 2 min   alert after 5 min    → IT + facilities
access-headend       port check, 2 min   alert after 5 min    → IT + facilities
door-panel-east      ping, every 2 min   alert after 10 min   → facilities first
door-panel-west      ping, every 2 min   alert after 10 min   → facilities first
elevator-gateway     ping, every 5 min   alert after 15 min   → facilities
                     (vendor's own alarm rides on top; this check watches the link)

And where a brain offers more than a ping — some expose SNMP or a status page — take what is offered. A controller that will report its own comm status with its field devices is handing you a depth of visibility most sealed boxes never will.


Vendor remote access, honestly


Every building-system vendor wants remote access, for a legitimate reason. It is how they diagnose and fix without a truck roll (sending a technician to site). Their service contract may promise response times that assume it. The problem is the default implementation: an always-on tunnel or cellular box, installed with the system, connected to the vendor's network forever, invisible to you. That is a permanent unattended door into the segment that controls your building's doors. When the vendor's own network has a bad day, it becomes other people's door too.


The honest posture is three words: controlled, logged, disabled-by-default. Concretely:


  • One path, yours. Vendor access goes through an access method you operate — a VPN account, a remote-desktop gateway, whatever your standard is — not through hardware they installed. If their box absolutely requires its own connection, it gets documented, firewalled to their named endpoints, and monitored like everything else.
  • Off until needed. The account is disabled between service events. Enabling it is a quick, defined step — a call or ticket away, fast enough that it never blocks an emergency. The account gets disabled again afterward. "Fast to enable" is the negotiating answer to "we need it always on."
  • Scoped. The access reaches the vendor's own system and nothing else. The elevator technician has no conceivable need to see your file server, and a properly scoped path means never having to wonder.
  • Logged. Sessions are recorded somewhere durable: who connected, when, how long. Not for drama — for the day a controller reboots mysteriously at 15:40 and you can see exactly who was logged into it at 15:38.
  • Written down. The whole arrangement lives in the support contract or an attached letter, agreed by the vendor. So the next technician works from the same agreement.

Vendors push back less than you would expect, because the alternative posture — unlogged permanent access to building controls — is increasingly hard for their lawyers to defend, too. The ones who refuse entirely are telling you something worth knowing before renewal time. And once the arrangement is in the contract, the next technician does not get to relitigate it from the parking lot.


Change coordination


Two calendars now share failure modes, and neither side can see the other's by default. Traffic flows both ways. Their changes break your assumptions. A vendor firmware push changes which cloud service the controllers call. And your carefully scoped firewall rule silently blocks the new destination. A system upgrade doubles the controllers' traffic. A technician clears a comm issue by changing a controller's IP address, orphaning your monitoring check. Your changes break theirs. A switch reboot during a maintenance window drops every panel in the building for four minutes. The access system spends the morning re-syncing. A VLAN cleanup renumbers a segment the elevator gateway had hard-coded.


The fix is unglamorous: tell each other. Give facilities and their key vendors a heads-up before network maintenance that touches their segments. And ask for the same before firmware updates or service visits — a shared calendar entry or a two-line email is plenty. Fold their systems into your maintenance-window discipline so planned work does not page anyone. And after either side changes anything, both sides glance at the dashboard. Your availability checks confirm their upgrade left everything reachable. Their system's own status confirms your switch work did not orphan a panel. Half of all "the building is acting weird" mysteries are a change on one calendar explained by a symptom on the other. A shared view collapses them from days to minutes.


A worked weekend: the controller that went quiet


The whole guide fits into one Saturday. At 14:07 the monitoring pages: hvac-supervisor unreachable. The on-call admin checks the corroborating signals from the couch. The building segment's switch is fine. The port feeding the controller shows link down. PoE (Power over Ethernet, power delivered down the network cable) draw is zero. So the network is healthy and the controller itself lost power or cable — two minutes of diagnosis, because the checks were built to answer exactly this question. The intake sheet names the facilities owner; one call reveals an electrician has been working in the mechanical room all afternoon. The owner walks over, finds the controller's circuit tripped, resets it. At 14:40 the check goes green. The building never noticed — it coasted comfortably on its last instructions the whole time.


Now run the counterfactual with no monitoring. The trip happens at 14:07 Saturday and nobody knows. The building coasts, then drifts all day Sunday. Monday at 07:00 the office is cold. The complaints start. Facilities assumes equipment failure and books an emergency vendor call-out. The technician arrives midday, finds a tripped breaker, and bills for the visit. Along the way somebody asks, reasonably, why the expensive networked building system did not tell anyone it was offline. The honest answer would have been that it did not have a single ping check pointed at it. One check, one route-to-owner rule, and one intake sheet turned a two-day, four-party mystery into a thirty-three-minute non-event. That is the entire business case for this guide, and it fits in one story.


The payoff: one shared dashboard


The checks you have just built pay a second time when you share them. Take the availability checks from this guide — every controller, panel, and gateway, green or red — and put them on a dashboard facilities can see. Give them a read-only view on the shop wall or the facilities manager's browser. It costs you nothing; every check already exists. What it buys is out of proportion to the effort. Facilities gains something they have never had: real-time proof of whether their systems' network layer is healthy, without having to call IT and ask. You gain a partner who calls you with the dashboard open — "the east panel went red ten minutes ago" — instead of a rumor two days later. And when something physical goes wrong, the first meeting starts from shared facts instead of dueling suspicions.


The cultural shift is the real product. IT and facilities historically meet only during blame: the building did something weird, and each side's opening position is that the other's stuff caused it. A shared dashboard, a shared intake sheet, and a shared change calendar quietly replace that with a peer relationship — two teams that run infrastructure, comparing notes. Their uptime is now partly your uptime; making that visible to both sides is how it starts being managed like it matters.


Where to go next


The arc, compressed: catch systems at arrival with an intake conversation. Segment them with respect for their age and their physical reach. Give every brain an availability check. Keep vendor access controlled, logged, and off by default. Share the resulting picture with facilities. None of it is hard, and all of it is cheaper than one cold Monday. The next time someone says the box just needs a port, you will know the fifteen minutes of questions that come with it. The same posture extends to every other sealed appliance on your network — vendor boxes and the accountability they owe you continues the thread. The rest of IoT, Cameras & Building Systems covers the neighboring fleets, from cameras to devices that talk too much.


Put this into practice with SysPeep Monitor

Every controller and panel in your building deserves a simple question every few minutes: are you still there? SysPeep Monitor discovers what is on your network. It runs availability checks against devices you cannot log into. And it shows the results on live dashboards you can share read-only with facilities — one console, one version of the truth. The free trial runs on any Windows machine.

Download Free TrialLearn About SysPeep