- Slobodan Krsmanovic |
- |
- 0 Comments
Updated October 29, 2026
“TL;DR:Yes, a running plant can be segmented without unplanned downtime in five phases: inventory and observe, agree zones, build the office-to-plant boundary, move one cell at a time in planned maintenance windows with a rollback, then tighten rules. A cell is a line or machine group, a maintenance window is a stop agreed with production, and a rollback is a written way back. TechProComp, a managed IT services provider, works with your controls engineers on the boundary and the change plan.”
For Texas plant owners weighing managed IT services for manufacturing, this guide gives a dated plan to segment an OT network without downtime, ready to approve before any line is touched. OT (operational technology) means the machines and controllers that run production. Working without downtime here means working without unplanned downtime, because no plan can promise an outage-free project, so every stop is scheduled in advance.
Segmentation projects can stall when nobody wants to be the person who stops a line. For the full zone design, read our OT/IT network segmentation playbook.
TechProComp’s phased sequence lets an owner approve segmentation of a running plant without unplanned downtime, because it watches before it changes anything. It builds the office-to-plant boundary before any cell moves, and it ties every planned maintenance window to a written rollback and a named person who can stop the change.

What does the phased plan look like?
Five phases, each starting with a verb: Inventory, Agree, Build the boundary, Move cells, and Tighten. Risk rises with each phase, from watching only to changing rules that can interrupt a flow. NIST SP 800-82 Rev. 3 lists logical separation of the corporate and OT networks among its defense-in-depth measures.
| Phase | Goal | Who is involved | Outputs | Done when | Planned window needed (yes/no) |
|---|---|---|---|---|---|
| 1. Inventory | Know every device and traffic path | IT lead, controls engineer | Device list and traffic map | Every device has a vendor, a protocol and a criticality | No |
| 2. Agree | Settle the zones and the order of work | Owner, IT lead, controls engineer, production supervisor, TechProComp | Zone list, allowed-flows table, owner per zone | The controls engineer has signed the zone list | No |
| 3. Build the boundary | Put one controlled edge between office and plant | IT lead, controls engineer, TechProComp | Firewall pair in permit-and-log mode (allow and record), logging on | Real flows show up in the logs | Yes |
| 4. Move cells | Place one cell at a time into its zone | Controls engineer, production supervisor, IT lead, TechProComp | Cell moved, rollback tested | The cell has run a full shift in its new zone | Yes |
| 5. Tighten | Switch from permit-and-log to deny-by-default (block what the table doesn’t list) | IT lead, controls engineer, TechProComp | Retired rules, alert owner, review date | The log review shows a rule for every real flow | Yes |
How long does each phase take?
The length of each phase, and of the whole project, varies by plant size, and no source we found gives a standard length. Your controls engineer and production supervisor set the pace, because they know which lines can stop and when.
Phase 1: Inventory what runs on the plant network
Record every device and every traffic path by watching the network before you change anything. Security vendor runZero writes that “regular network and vulnerability scanners can cause devices to act erratically.”
That post also favors active discovery first. Our passive-first rule for legacy controllers is TechProComp’s own approach, because a frozen controller means a stopped line.
What should you record for each device?
Walk the floor with your controls engineers and write down six fields for every device. CISA’s asset inventory guidance calls for a “physical inspection and logical survey” and a longer attribute list than this one.
- Device: the make and model on the label
- Vendor: who supports it
- Protocol: the language it speaks on the network
- Who talks to whom: every partner device
- How often: constant, hourly or once a shift
- Criticality: what stops if it fails
Why does Phase 1 use passive observation only?
Passive observation means listening to traffic without sending probes to any device. Your IT lead sets up a copy of the switch traffic, and the team logs which devices talk and how often.
How long observation runs varies by plant size. Let it cover a full production cycle so rare flows show up, then record the result in the device list so the next phase starts from facts.
Phase 2: Agree the zones and the order of work
Draw the zones with your controls engineers in the room. A zone map drawn from the office alone will break flows nobody wrote down, so this phase ends with three outputs: a zone list, an allowed-flows table and an owner for each zone.
What happens in the 90-minute zone-design workshop?

Run it with the owner, your IT lead, controls engineers, production supervisor and TechProComp, and keep to this agenda:
- 10 minutes: review the Phase 1 device list and traffic map.
- 25 minutes: group cells into candidate zones.
- 25 minutes: mark which flows between zones are needed and why.
- 20 minutes: set the order of work, one cell at a time.
- 10 minutes: name an owner for each zone and record open questions.
How do you group cells into zones?
Group by line, by function or by vendor, and pick the criterion that matches how your plant stops and starts. These are TechProComp’s suggested criteria, not a standard.
Grouping by line keeps one product’s cells together, and grouping by function puts like machines together, such as all packaging. Grouping by vendor helps when one supplier supports a whole set of equipment.
Allowed-flows table template
Copy this table into a spreadsheet and add one row for every flow that must cross a zone edge.
| From zone | To zone | Service | Direction | Reason | Owner | Review date |
|---|---|---|---|---|---|---|
Which reference documents guide zone design?
Three public documents help. NIST and ISA/IEC publish guidance and a standard, and the Rockwell and Cisco work is one of several reference designs.
- NIST SP 800-82 Rev. 3, published September 28, 2023, gives guidance on securing operational technology.
- ISA/IEC 62443-3-2, released in February 2020 according to ISA, makes partitioning a system into zones and conduits a risk assessment step.
- Rockwell Automation and Cisco developed Converged Plantwide Ethernet designs, including a March 2022 guide to the industrial demilitarized zone.
Verified as of October 2026. Editions change, so check the current one before you rely on it.
Phase 3: Build the boundary between office and plant
Build the boundary first, because it gives every later change one place to be watched and controlled. TechProComp documents the office network, firewall and boundary to production, while your controls engineer and the equipment vendor decide what a controller can accept. The CPwE industrial DMZ guide says to “Eliminate direct traffic between the Enterprise and Industrial Zones.”
What goes into the boundary?
The boundary has four parts:
- A firewall pair at the office-to-plant edge
- A brokered path to the historian (the plant’s data archive) or MES (production tracking software)
- A controlled remote-access path for vendors
- Logging from day one
Why does the boundary come before any cell work?
It gives you one protected place to test the rules, and logs that show real flows before any cell moves.
“TechProComp has delivered the necessary equipment on schedule and tested the network before the client’s office opening.” – Clutch summary, Miro review
Log-only mode before enforcement
Run the new rules in permit-and-log mode first. The firewall allows what the rules cover and writes down what it saw.
Skipping this step risks blocking a flow nobody documented. Enforcement starts only after a log review shows every real flow.
What goes in the change-approval and rollback template?
Use a one-page template that names who signs, who can stop the change, how production is told, the rollback trigger and the time limit. Fill it in before every window opens. The 2009 edition of NIST SP 800-41 Rev. 1 says “Firewall rulesets and policies should be managed by a formal change management control process”.
What fields does the template need?
| Field | What to write |
|---|---|
| Change description | What changes, on which device, in plain words |
| Approver (who signs) | The name and role of the person who signs the change |
| Stop authority (who can stop it) | One named person on the plant side who can stop the change at any point |
| How production is told | Who tells the supervisor, by what channel and when |
| Config backup location | Where the saved device configuration lives and who checked it |
| Window start and end | The agreed start and end times with production |
| Rollback trigger | The event that makes the team undo the change |
| Rollback time limit | The longest the change may run before the team undoes it |
Who can stop the change?
Name one person on the plant side who can stop any change at any point. Agree on that name before the window opens and write it into the template, so nobody has to decide in the moment whether they’re allowed to stop the work.
The longer-term question of who owns OT network security is a separate decision.
Phase 4: Move cells one at a time in planned windows
Move each cell in its own planned maintenance window, with a written rollback ready before the first command. A planned stop has a known length, and the downtime cost calculator lets you put a figure on an unplanned one. The CPwE industrial DMZ guide recommends a change management system to archive network, controller and computer assets.
What goes into each maintenance window?
Every window follows the same runbook, which is TechProComp’s own template:
- Pre-checks: confirm the config backup, the named stop authority and that production has been told.
- Steps: apply the change in the written order.
- Stop/go check: compare the result with the criteria agreed in advance, and go on only if every one is met.
- Rollback: if a criterion fails or the time limit passes, restore the saved configuration.
TechProComp’s support is hybrid: 24/7 remote monitoring plus on-site visits, not a permanent on-site technician, so windows are scheduled in advance.
What does an illustrative calendar look like?
This calendar is illustrative. It shows one possible pace for six cells over twelve weeks and isn’t a promise of dates.
| Week | Cell | Window activity | Rollback checkpoint |
|---|---|---|---|
| Weeks 1 to 2 | Cell A | Move the cell in one window, then watch the logs | Saved config restored if a stop/go check fails |
| Weeks 3 to 4 | Cell B | Move the cell in one window, then watch the logs | Saved config restored if the time limit passes |
| Weeks 5 to 6 | Cell C | Move the cell in one window, then watch the logs | Saved config restored if a stop/go check fails |
| Weeks 7 to 8 | Cell D | Move the cell in one window, then watch the logs | Saved config restored if the time limit passes |
| Weeks 9 to 10 | Cell E | Move the cell in one window, then watch the logs | Saved config restored if a stop/go check fails |
| Weeks 11 to 12 | Cell F | Move the cell in one window, then watch the logs | Saved config restored if the time limit passes |
“The team is responsive, communicative, and willing to work after hours to fix issues.” – Clutch summary, Salient Systems review
Phase 5: Tighten the rules after a log review
Move from permit-and-log to deny-by-default only after the logs show every real flow. Deny-by-default means the firewall blocks anything the allowed-flows table doesn’t list. The 2009 NIST SP 800-41 Rev. 1 pairs firewall policy with “ruleset reviews or tests performed periodically”.
Each tightening step is itself a change, so it uses the same approval template and a rollback.
What does AI track after the switch to deny-by-default?

Switching to deny-by-default is where segmenting a running plant gets tested, because any flow the logs missed won’t get through anymore. When an operator reports that a screen or a production report stopped updating, TPC Ops Hub, the platform we built in-house, gives that report a fast path. The AI reads it, ranks it by business impact and creates a ClickUp ticket immediately, routed to the right technician with the details already attached. If the same complaint comes from two zones, it flags the pattern so the missing rule gets fixed once and doesn’t get re-ticketed. Your IT lead, your controls engineer and our engineers work from one dashboard. The named stop authority still decides on any rollback, because the AI tracks the issue and people make that call. Our AI ticket handling page shows how it fits a phased project like this one.
Explore AI IT Services
See how our AI turns each blocked-flow report after a deny-by-default switch into a tracked ticket, so you can fix a missed rule before the next zone switches over.
How long should the log review last?
We suggest 30 days, and that’s a suggestion, not a standard. Adjust it so the period covers a full production cycle, including any weekly or monthly jobs. The permit-and-log rules keep running during the review, so nothing is blocked while the evidence builds.
What happens on the day you switch to deny-by-default?
Treat the switch as a planned window, with the template filled in, a config backup and a rollback time limit. Start with the zone that showed the cleanest logs, and keep the stop authority on call.
When do you retire rules and who owns alerts?
Retire any rule that saw no traffic through the review period. Name one alert owner who reads the firewall alerts, and set a quarterly rule review. Each retired rule is a change, so record it in the template with its own rollback.
Those logs and reviews also become the evidence an auditor asks for, and the audit evidence checklist lists what to keep.
How do you handle legacy PLCs and older equipment?
A legacy PLC (programmable logic controller, the small industrial computer that runs a machine) doesn’t have to be swapped out before you segment around it. Treat each case as a set of options and send the specific question to the equipment vendor. A 2016 Rockwell and Cisco migration guide says it’s up to the reader to decide how its content applies to a specific migration.
A device that cannot take a new IP or gateway
Three options exist: address translation at the cell edge, an industrial firewall in front of the cell, or leaving the device in place and isolating it. The same guide describes Layer 2 address translation as hardware-based 1:1 translation on a Layer 2 switch, applied only to devices with private IP addresses such as I/O (input and output) devices. Ask the equipment vendor which option it supports before anyone changes the device.
Layer 2-only or multicast-dependent equipment
Layer 2 means devices that talk on the same local network without routing. Multicast means one sender reaching many receivers on that network.
Ask the vendor whether the equipment can sit behind a routed boundary or should stay inside one zone. An industrial firewall in front of the cell, or leaving the device in place and isolating it, are the options to raise.
A device with no vendor support
Two options remain: isolate the device behind an industrial firewall, or leave it in place and isolate it with tight allowed flows. Start from the manufacturer’s last documentation, and have your controls engineer confirm what the device must reach.
How many hours does the project ask of each person?
These hour ranges are illustrative, vary by plant size and aren’t a TechProComp metric. TechProComp Co-Managed IT works alongside your IT lead and adds capacity without taking ownership away. Clients also get 365-day, 24-hour access to complete IT documentation, as published on techprocomp.com, so the network map and the rules stay with you.
| Role | Phase 1 | Phase 2 | Phase 3 | Phase 4 | Phase 5 |
|---|---|---|---|---|---|
| Owner | 1 to 2 | 2 to 3 | 1 to 2 | 2 to 4 | 1 to 2 |
| IT lead | 8 to 16 | 6 to 10 | 10 to 20 | 12 to 24 | 6 to 12 |
| Controls engineer | 6 to 12 | 6 to 10 | 2 to 6 | 12 to 24 | 4 to 8 |
| Production supervisor | 2 to 4 | 3 to 5 | 1 to 2 | 6 to 12 | 1 to 2 |
| Provider | Not involved | 4 to 8 | 16 to 30 | 12 to 24 | 8 to 16 |
Illustrative hours per role and phase. Actual hours vary by plant size.
What stays with your own team?

The owner’s approval, the controls engineer’s authority over controllers and the production supervisor’s say on windows all stay internal. TechProComp’s founder, Slobodan Krsmanovic, has 25+ years in IT and stays involved in client relationships. TechProComp adds capacity on the office network, firewall and boundary work, and your IT lead keeps the planning and the vendor relationships.
“Anytime a decision needed to be made, Slobo would present recommendations and give me a pressure-free experience while also providing great advice.” – Lucas Christianson on CloudTango
Which pitfalls stop segmentation projects?
Seven pitfalls are worth planning around. Four come from moving too fast and three from leaving the right people out. Each has a fix you can write into the plan, and each reflects TechProComp’s planning guidance, not a study.
Which pitfalls come from moving too fast?
These four pitfalls share one cause: a step skipped to save time, or a change made before the evidence was in. Each has a direct fix:
- Active scanning first: runZero notes scanners can make devices act erratically but still favors active discovery, so this plan observes passively first.
- Skipping log-only mode: run permit-and-log first so surprises appear in logs, not as outages.
- Tightening too early: wait until the log review ends and every real flow has a rule.
- Treating the calendar or hours as a promise: keep both labeled illustrative and re-plan with your controls engineer.
Which pitfalls come from missing people?
These three pitfalls share one cause: someone with authority or knowledge wasn’t in the room. Each fix is a name or a seat at the table:
- Zones drawn without controls staff: hold the workshop with your controls engineers in the room.
- No named stop authority: write the name into the template before the window opens.
- Choosing a controller fix without the vendor: ask the equipment vendor first.
Frequently Asked Questions
How do you segment a plant network without stopping production?
Segment in five phases and schedule every stop in advance. Inventory the network by watching it, agree zones with your controls engineers, build the office-to-plant boundary, move one cell at a time in planned maintenance windows with a rollback, then tighten rules after a log review. No plan can promise an outage-free project, but this one plans every stop.
How long does OT network segmentation take?
It varies by plant size, and we found no sourced standard duration for the whole project. The pace depends on how many cells you move and how often production can stop. As an illustration only, one calendar moves six cells across twelve weeks, but your controls engineer and production supervisor should set the real dates.
What should be done first when segmenting an OT network?
Start by inventorying devices and traffic paths through passive observation, with no active scanning of legacy controllers. Record the device, vendor, protocol, who talks to whom, how often, and how critical it is. Walk the floor with your controls engineers, because they know which equipment the network map is most likely to miss.
Do you need to replace legacy PLCs to segment the network?
No, swapping out the controller isn’t a precondition. For a device that can’t take a new address, options include address translation at the cell edge, an industrial firewall in front of the cell, or leaving it in place and isolating it. Ask the equipment vendor which option it supports before anyone changes the device.
How do you roll back a network change on the plant floor?
Write the rollback before the window opens, with a configuration backup, a time limit and a named trigger such as a failed stop/go check. If the trigger fires or the limit passes, restore the saved configuration and tell production. The change-approval template records each of these fields, so the team follows one written path.
What is a maintenance window and how is it used for network changes?
A maintenance window is a stop agreed in advance with production. Network changes happen inside it, so any interruption is planned and short. Each window has pre-checks, steps, stop/go criteria and a rollback with a time limit, and a named person can stop the change at any point if a check fails.
Key Terminology
Maintenance window: A stop agreed in advance with production, used for changes that can interrupt a flow. Each window has pre-checks, steps, stop/go criteria and a rollback with a time limit. Phases 3, 4 and 5 each use planned windows.
Rollback: A written way back to the last working configuration. It names a trigger, a saved configuration backup and a time limit. The change-approval template holds all three fields.
Cell: A line or machine group that moves into a zone as one unit. It’s the unit of work in Phase 4. Six cells across twelve weeks is the illustrative pace used in the calendar.
Zone: A group of cells that share the same allowed flows and the same owner. Traffic between zones passes through the boundary and follows the allowed-flows table. A conduit is the path that carries that traffic.
Passive discovery: Observing network traffic without sending probes to devices. It adds no scan traffic to fragile equipment. Phase 1 uses it, and no other discovery method, before any change is made.
Book a free IT assessment and we’ll document your office network, firewall and the boundary to production, so your team starts the plan from a written map.
About the author
Slobodan Krsmanovic, the CEO of TechProComp, brings over 25 years of deep-rooted experience in the IT industry. As the author driving our insightful posts, Slobodan embodies a steadfast commitment to client-centric service, fostering respectful and secure collaborations across all business scales.