AllTech
01
Unmanaged switches daisy-chained across the building
No VLANs, no port visibility, no way to tell which device is flooding the network — and a consumer switch under a desk that nobody documented.
Network & Infrastructure · UniFi Specialist
The part of the network nobody thinks about
until everything stops.
Switches and gateways are the floor the rest of your technology stands on. When they're specified correctly you never think about them; when they're not, you get intermittent problems nobody can reproduce and a vendor blaming your network. AllTech designs, installs, and maintains that layer — primarily on Ubiquiti UniFi, and on other platforms when the requirements call for it.
UniFi specialist
On-site engineers
Segmentation by default
Local since 2009
01 · The problem
"It's probably the network."
Most of the switching we inherit was bought a box at a time, over a decade, by whoever needed a port that afternoon. It usually works. It also usually can't be troubleshot, monitored, or extended.
02
One flat network
Cameras, guest Wi-Fi, the POS, printers, the plant floor, and the domain controller all in the same broadcast domain, all able to reach each other.
03
PoE budget quietly exceeded
Cameras and access points that reboot at random because the switch runs out of power under load, not because anything is broken.
04
A single 1G uplink carrying the whole building
Fine until backups, camera retention, and file traffic overlap on the same Tuesday morning.
05
The ISP router is the whole network
No segmentation, no failover, no logging, and a device the ISP can replace remotely without telling anyone.
06
No spare, no config backup, no lifecycle plan
When the gateway dies, the recovery path is a two-day shipping estimate and someone rebuilding rules from memory.
07
Firmware from three years ago
On the device every other device depends on.
02 · What we deploy
One managed fabric, documented.
A designed network has a gateway that knows what it's routing, switches that can be seen and segmented, uplinks sized for what actually crosses them, and PoE budgeted with headroom. We build it as one managed system with one place to look, not a pile of independently configured boxes.
01
Gateways
Routing, inter-VLAN policy, DHCP, DNS forwarding, VPN termination where appropriate, and multi-WAN failover. Sized to your actual throughput and to the features you'll have enabled — not to the marketing number with everything switched off.
02
Access-layer switching
Managed switches with per-port visibility, PoE and PoE+ where devices need it, and a documented port map so a problem can be traced to a port instead of guessed at.
03
Aggregation and uplinks
10G copper or fiber between the core and each closet, so backups, camera retention, and user traffic aren't competing for one gigabit.
04
VLAN segmentation
Separate networks for staff, guests, VoIP, cameras, door access, printers, servers, and OT/plant equipment — with explicit rules about what may talk to what. Guest and IoT isolated by default.
05
PoE design
Power budget calculated against real device draw with headroom, not against the number on the datasheet. This is the single most common cause of "random" camera and AP reboots.
06
Multi-WAN and failover
A second circuit (fiber, cable, or cellular) with tested failover, because failover that has never been tested is an assumption.
07
Controller and management
UniFi Network hosting decided deliberately — on-premise console, self-hosted, or Ubiquiti-hosted — with alerting and configuration backups either way.
08
Lifecycle and documentation
Firmware baselines, a documented topology, config backups, and a defined spares position for the devices that would hurt most.
03 · Vendor position
UniFi first, not UniFi only.
We're UniFi specialists, and for the large majority of the businesses we support that's the right answer. It's also a choice we make deliberately, per site, and we'll tell you when it isn't the right fit.
01
Why UniFi is our default
One controller for switching, Wi-Fi, cameras, and door access. No per-device licence to renew. Predictable hardware cost that leaves budget for the rest of the stack. And we run enough of it that our engineers know its behaviour — including its quirks — rather than learning your deployment on your time.
02
When we specify something else
Requirements drive hardware, not the other way around. We move off the default when a site needs deep next-generation firewall inspection with vendor-backed threat intelligence and a support SLA; when dynamic routing or advanced Layer 3 is genuinely required; when an existing platform, cyber-insurance requirement, or compliance framework dictates a specific vendor; when throughput at the edge exceeds what the platform holds up under with inspection enabled; or when a client needs a hardware replacement guarantee that a consumer-channel RMA process does not provide.
03
How we keep that decision reversible
The design is documented independently of the hardware — addressing plan, VLAN scheme, port map, routing and policy intent. A mixed environment is normal and supportable: we regularly run UniFi switching and Wi-Fi behind a different firewall, or UniFi at branch sites with something heavier at headquarters. Nothing here requires you to be single-vendor forever, and we will not pretend otherwise to keep a build simple for us.
What doesn't change. SASE and Zero Trust functions sit at the Cloudflare layer regardless of what hardware is in the rack — see Cloudflare Zero Trust. Choosing a different switch vendor doesn't rebuild your security stack.
04 · How we size it
Specified against what you'll actually run.
- Single office, under ~25 users
- Gateway: Compact cloud gateway Switching: One or two PoE access switches Uplinks: 1G, 10G if the budget allows
- Office with cameras and door access
- Gateway: Mid-tier gateway with headroom Switching: PoE+ access switches, budget calculated Uplinks: 10G to the closet
- Multi-closet building
- Gateway: Gateway sized for inspection enabled Switching: Aggregation switch + per-closet access Uplinks: Fiber between closets
- Manufacturer / plant floor
- Gateway: Gateway with strict inter-VLAN policy Switching: Hardened or industrial access switches on OT VLANs Uplinks: Fiber, redundant where the run allows
- Multi-site
- Gateway: Per-site gateway, consistent template Switching: Same access model at every site Uplinks: Site-to-site or SASE overlay
- High-throughput or compliance-driven edge
- Gateway: Alternate vendor NGFW — see section 03 Switching: UniFi switching still fine behind it Uplinks: 10G+
Sizing rule we hold to: gateways get specified against throughput with the features you want turned on, and PoE budgets get specified with headroom for the devices you'll add next year. Both are cheap to get right on day one and expensive to fix afterward.
05 · How we engage
Three ways to start.
01
ASSESS · Network review
Physical walk of the closets and cabling, switch and gateway inventory with firmware and lifecycle status, VLAN and addressing review, PoE budget check, and uplink saturation review. You get a written topology, a prioritized findings list, and a replacement plan with rough costs. Stands alone as a deliverable — no obligation to buy hardware from us.
02
DEPLOY · Design and install
Addressing and VLAN design, hardware specification and procurement, staged install with a defined cutover window, labelling, and documentation handed over. Cabling and rack work in-house.
03
OPERATE · Managed network
Monitoring and alerting, firmware lifecycle, config backups, port and VLAN changes, capacity review, and warranty/RMA handling. Billed monthly. When a switch fails, we already know which one it is and what was plugged into it.
Cabling and rack work is in-house — see Network & Infrastructure.
06 · How the rollout goes
Four phases, and nothing gets unplugged before the new path is proven.
- 01
Phase 1 · Design and staging
Addressing plan, VLAN scheme, and port map agreed on paper before anything is ordered. Hardware is configured and updated on the bench, not in your closet at 6 a.m.
- 02
Phase 2 · Core and gateway cutover
Gateway and aggregation switch go in during a scheduled window, usually after hours. Old equipment stays in the rack, powered down and reachable, until the new path is proven.
- 03
Phase 3 · Access layer and segmentation
Closet by closet, floor by floor. VLANs get applied in monitoring mode before enforcement where the platform allows it, because this is the phase that finds the label printer with a hard-coded IP and the machine nobody could identify.
- 04
Phase 4 · Verification and handover
Failover tested by actually pulling the primary circuit. PoE draw measured under load. Topology, port map, addressing, and credentials documented and handed to you — including if you later take it in-house or move to another provider.
07 · Who this is for
Where it makes the most difference.
01
Businesses on consumer or unmanaged gear
That have outgrown it and are seeing problems nobody can pin down.
02
Anyone adding cameras or door access
That's a PoE and segmentation project before it's a camera project.
03
Manufacturers
Needing plant-floor equipment isolated from the office network without losing the visibility they need into it.
04
Clinics and professional services
Where segmentation is a compliance requirement, not a preference.
05
Multi-site organizations
Wanting one standard applied at every location, including the next one.
06
Offices moving or building out
The cheapest time to do this correctly is before the walls close.
07
Anyone whose network is one undocumented device away from a bad week
If that sentence landed, this is the conversation.
08 · Honest limitations
What we'll tell you before you buy anything.
01
UniFi is not a next-generation firewall.
It does routing, segmentation, and basic threat signatures well. If your requirement is deep inspection with a vendor SLA and licensed threat intelligence, we'll specify a different edge device and say so during assessment — not after.
02
Gateway throughput drops when inspection is enabled.
Headline throughput numbers assume features are off. We size against the configuration you'll actually run.
03
Hardware support is a channel, not an SLA.
Ubiquiti's replacement process is serviceable, not same-day. For sites where an outage is measured in lost production, we recommend an on-shelf spare and quote it explicitly. Some clients would rather buy a vendor support contract instead — that's a legitimate reason to choose different hardware.
04
The controller is a dependency.
UniFi switching keeps forwarding traffic if the controller is unavailable, but you lose management, changes, and visibility until it's back. We make the hosting decision deliberately and back up the configuration.
05
Segmentation will break something on day one.
Some device — a printer, a scanner, a licence server, a piece of shop equipment — depends on flat-network reachability nobody documented. Finding those is part of the work, and it's why the access layer is phased.
06
New switching does not fix bad cabling.
If the drops are unterminated, untested, or Cat5e where the application needs more, that's the actual problem. We test rather than assume, and we'll quote the cabling honestly.
07
It doesn't replace endpoint or email security.
A well-segmented network limits how far a compromise spreads. It doesn't stop one from starting.
08
Wi-Fi is a separate design problem.
Good switching is a prerequisite for good Wi-Fi, not a substitute for a site survey.
For the two gaps switching leaves open, see Cybersecurity and Wi-Fi design.
09 · FAQ
Common questions
Do we have to use UniFi?
No. It's our default because it fits most of our clients well and we support it deeply, but requirements decide. Mixed environments — UniFi switching behind a different firewall, for instance — are normal and fully supportable.
We already have UniFi that another vendor installed. Will you take it over?
Yes, and that's common. We start with an assessment: what's there, what firmware it's on, what's out of support, and what needs to change before we put our name on it.
Can you reuse our existing switches?
Sometimes. If they're managed, in support, and have the PoE budget and port count for what you're adding, we'd rather reuse than replace. We'll tell you which units are worth keeping and which are the reason for your intermittent problems.
How much downtime does a switch and gateway replacement take?
Core cutover is typically a scheduled after-hours window. The access layer is done in stages with brief per-closet interruptions. Nobody should lose a business day to this.
Do we need 10G?
Between the core and the closets, usually yes — it's inexpensive now and it's what keeps backups and camera traffic from competing with users. To every desk, almost never.
Will segmenting the network break our line-of-business software?
Occasionally, and we plan for it. Some applications assume flat reachability. Those get identified during the phased rollout and handled with explicit rules rather than by giving up on segmentation.
Who owns the equipment and the configuration?
You do. Configuration, credentials, addressing, and documentation are yours, and you get them whether or not you stay with us.
What happens when a switch fails?
Under managed service we get the alert, identify the unit and what was on it, handle the warranty claim, and — if you've bought a spare — swap it and restore the config. Without a spare, you're on shipping time, which is exactly why we quote the spare.
Can you support this if we're not local?
Most of the work is remote-capable, and our engineers are based in Northern Utah — for the right project, we travel further than that. Tell us where you are and we'll scope it.
10 · Related services
The rest of the stack it plugs into
01
Network & Infrastructure
The parent service: cabling, racks, point-to-point, and the rest of the physical layer.
02
Private Cloud Routing
Reach these networks remotely without opening inbound ports.
03
Cloudflare Zero Trust
The security layer that sits above the hardware, whatever the hardware is.
04
Cybersecurity
Endpoint and email defence — segmentation limits the spread, it does not stop the start.
05
Managed IT
Monitoring, patching, and the help desk that answers when a port goes quiet.
Start with a conversation
Start with a look at what you actually have.
Send us a note about your sites and what's giving you trouble. We'll walk the closets, map what's there, and give you a written plan with priorities and rough costs — including the parts we'd leave alone.
