TTechclick ⚡ XP 0% All lessons
Nozomi · OT/IoT Security · Deployment ArchitectureInteractive · L1 / L2 / L3

Nozomi Deployment Architecture — Guardian Placement, SPAN vs TAP & Purdue Rollout

Deploying Nozomi Networks means deciding where Guardian sensors sit across the Purdue hierarchy, whether to feed them via a SPAN port or a TAP, how to right-size each zone, and how Guardians connect up to CMC or Vantage for multi-site management. This lesson maps every placement decision — from Level 2 control networks to the IT/OT DMZ — and walks a phased rollout that keeps OT operations untouched.

📅 2026-06-18 · ⏱ 16 min · 5 infographics · live sensor demo · 🏷 10-Q assessment + AI Tutor inline

⚡ Quick Answer

How to deploy Nozomi Networks Guardian sensors across the Purdue model (Levels 0–5), choose SPAN vs TAP, size per zone, connect to CMC or Vantage, and run a phased OT security rollout across multiple sites.

🎯 By the end you will be able to

Read as:

Pick where you want to start

1

Purdue & Guardian

Where sensors sit across Levels 0–5 and the DMZ.

2

SPAN vs TAP & sizing

Traffic collection, packet-loss risk, right-sizing per zone.

3

CMC & Vantage topology

On-prem aggregation vs SaaS; Arc coverage gaps.

4

Multi-site & phased rollout

Phase 1–4 deployment, SIEM integration, expansion.

🧠 Warm-up — 3 questions, no score

Just notice which ones make you pause. We answer all three inside the lesson.

1. Where is the first Guardian sensor usually placed in a new OT deployment?

Answered in Purdue & Guardian.

2. What is the key advantage of a TAP over a SPAN port?

Answered in SPAN vs TAP & sizing.

3. What does CMC do that Vantage SaaS cannot?

Answered in CMC & Vantage topology.

Most engineers think…

Most people picture OT sensor deployment as 'one box anywhere on the network'. In a Purdue-model OT environment, placement is everything — and getting it wrong means blind spots in the zones that matter most.

A Nozomi Networks deployment is a layered topology: Guardian sensors at the right Purdue-level aggregation switches (not everywhere — not on every device), each fed by a SPAN port or a TAP, all feeding up to a CMC or Vantage. Knowing which Purdue level gets a sensor, when to upgrade from SPAN to TAP, and how to roll out without alarming OT operators — that is what the next 16 minutes covers.

① The Purdue model — which levels get a Guardian sensor

The Purdue Reference Model (Levels 0–5 plus the IT/OT DMZ at Level 3.5) defines the zones of an industrial network. Guardian sensors are not placed on every device — they sit at aggregation switches where traffic from an entire zone crosses one point.

The primary Guardian placement zones are: Level 2 (HMIs, SCADA servers, historians, engineering workstations — the richest traffic for asset discovery and anomaly detection); Level 3 (MES, operations servers, site historians — north–south traffic between control and business zones); and Level 3.5, the IT/OT DMZ (firewalls, jump servers, OPC proxies — critical for detecting cross-boundary anomalies and firewall-bypass). Levels 0–1 (PLCs, RTUs, field instruments) are usually covered by the Level 2 SPAN, since field-bus traffic converges at the Level 2 switch anyway.

Levels 4–5 and beyond

The enterprise LAN (Levels 4–5) is normally covered by IT security tooling. A Guardian at the Level 4 switch can catch OT-to-corporate lateral movement, but is optional in most deployments. Guardian Air (the wireless variant) is added wherever Wi-Fi, Bluetooth, or cellular OT devices operate — often outside the standard Purdue wired hierarchy.

LegendGuardian sensor / primary monitored zone (royal)network level / pipeline stagediagram headingdiagram background panelsupporting label / caption
Figure 1 — Purdue levels & Guardian placement
Guardian sensors sit at aggregation switches — primarily Level 2, Level 3, and the IT/OT DMZ — not on every device.Purdue levels & Guardian placementLevel 4–5Enterprise LAN — IT tools + optional GuardianLevel 3.5 DMZFirewalls, OPC proxy — Guardian recommendedLevel 3MES, historians — Guardian for north–southLevel 2HMI, SCADA — primary Guardian zoneLevel 0–1PLCs, RTUs — covered via Level 2 SPAN
Guardian sensors sit at aggregation switches — primarily Level 2, Level 3, and the IT/OT DMZ — not on every device.
Figure 2 — Traffic flow: field device to Vantage
A PLC packet travels from the field bus through the Level 2 SPAN to Guardian, then up to CMC or Vantage.Traffic flow: field device to VantagePLC / RTULevel 0–1 fieldLevel 2 switchSPAN or TAPGuardianPassive DPICMC / VantageAggregation
A PLC packet travels from the field bus through the Level 2 SPAN to Guardian, then up to CMC or Vantage.
Quick check · Q1 of 10 · Understand

In most Nozomi deployments, the primary Guardian placement zone is…

Correct: b. Level 2 (supervisory) is the primary Guardian zone because the HMI-PLC and HMI-historian traffic converges at the Level 2 switch, giving the richest asset discovery and anomaly detection from the fewest sensor placements. Levels 0–1 field devices are usually covered via the Level 2 SPAN.
👉 So far: Guardian sensors belong at Purdue Level 2 (HMI/SCADA), Level 3 (site operations), and Level 3.5 (IT/OT DMZ) aggregation switches — not on every device. Levels 0–1 are covered via the Level 2 SPAN.

② SPAN vs TAP — how Guardian gets its traffic feed

Guardian is completely passive: it never injects packets into the OT network. Traffic reaches it one of two ways. A SPAN port (mirror port) is a feature built into most managed switches — the switch copies selected port traffic to a monitoring port. SPAN is free and easy to configure, but many switch models allow only one SPAN destination at a time, and under heavy load the switch may drop mirrored frames before Guardian sees them.

A network TAP (Test Access Point) is a dedicated hardware device placed inline on a link. It copies 100% of frames to Guardian with no packet loss, regardless of link utilisation — the physical copy is passive and cannot be saturated by the live traffic. TAPs are recommended for high-value chokepoints: the IT/OT DMZ uplink, the historian server connection, or any link where a packet-dropping SPAN would create blind spots in asset inventory or forensics.

Sizing per zone

Guardian sizing (appliance model or virtual spec) scales with link bandwidth and asset density per zone. The guiding principle is one Guardian per discrete network zone or VLAN. A typical mid-size substation might deploy two Guardians: one for the Level 2 control zone and one for the Level 3/DMZ zone. Virtual Guardians (VMware, Hyper-V, container) are common in sites where rack space is limited.

Figure 3 — SPAN port vs network TAP
Both feed Guardian passively — SPAN is free but can drop frames; TAP costs more but guarantees full capture.SPAN port vs network TAPSPAN (mirror port)Built into managed switches — noFree to configure; needs switchCan drop mirrored frames underUsually one SPAN destination perNetwork TAPDedicated hardware inline on theZero packet loss at any linkHigher upfront cost; requiresRecommended for DMZ and critical
Both feed Guardian passively — SPAN is free but can drop frames; TAP costs more but guarantees full capture.
📡
Guardian sensor
tap to flip

Nozomi's core passive network sensor — passive DPI, asset discovery, anomaly and threat detection. Deploys as appliance, VM, or container; never touches OT traffic.

🏭
CMC
tap to flip

Central Management Console — on-prem or virtual aggregator for multiple Guardians. Ideal for air-gapped, sovereign, and regulated sites that cannot send data to the cloud.

☁️
Vantage SaaS
tap to flip

Cloud-native SaaS management platform. Aggregates Guardians and Arc across many global sites with no management hardware — adds Vantage IQ (AI analytics) for faster triage.

🖥️
Arc endpoint
tap to flip

Lightweight host sensor adding user, process, USB, and session context — fills coverage gaps in zones that have no SPAN port and cannot be seen by passive Guardian monitoring.

One Guardian per zone, not per device

The classic over-deployment mistake is trying to put a sensor on every switch. The right model is one Guardian per discrete network zone or VLAN — placed at the aggregation point where all zone traffic converges. This minimises sensor count while maximising coverage.

Quick check · Q2 of 10 · Apply

The IT/OT DMZ uplink switch is heavily loaded at peak. Which collection method should you use for the Guardian feed?

Correct: a. A heavily loaded switch risks dropping SPAN mirrored frames, creating blind spots in asset inventory and forensics. A hardware TAP copies 100% of frames passively with no packet loss — the right choice for high-value, high-utilisation links like the IT/OT DMZ uplink.
👉 So far: SPAN is free but can drop frames under load; a TAP is hardware-cost but guarantees 100% capture. Use TAPs on high-value, high-utilisation chokepoints like the IT/OT DMZ uplink.

③ CMC & Vantage — the management topology above Guardian

Individual Guardians are powerful, but multi-site OT programmes need a central management layer. Nozomi offers two paths. The Central Management Console (CMC) is an on-prem or virtual appliance that aggregates data from many Guardians across one or more sites — ideal for air-gapped, sovereign, or heavily regulated environments (power utilities, defence, water) where OT metadata cannot leave the site perimeter. CMC provides centralised policy, cross-site alerts, and consolidated reporting without any cloud dependency.

Vantage is Nozomi's cloud-native SaaS alternative: Guardians (or a CMC) send data to Vantage over encrypted HTTPS. Vantage scales effortlessly across dozens of global sites, adds Vantage IQ (AI-assisted triage and correlation), and requires no management-layer hardware. For hybrid estates — some sites air-gapped, some cloud-connected — CMC aggregates air-gapped Guardians and forwards summary/alert data on to Vantage, giving a single pane of glass globally.

Arc fills the coverage gaps

Arc, Nozomi's lightweight host sensor, complements Guardian by reaching segments that have no SPAN port — micro-segmented zones, isolated workstations, or point-to-point serial links visible only from the host. Arc adds user, process, USB, and session context that passive network monitoring cannot see.

Figure 4 — Guardian topology — sensors to management
Each Guardian reports to CMC (on-prem) or Vantage (SaaS), with Arc filling host-visibility gaps.Guardian topology — sensors to managementCMC / VantageCentral managementGuardian (L2)Guardian (L3)Guardian (DMZ)Guardian AirArc (host)Site 2 CMC
Each Guardian reports to CMC (on-prem) or Vantage (SaaS), with Arc filling host-visibility gaps.
'We will just use Vantage everywhere'

Vantage SaaS is ideal for cloud-connected sites, but air-gapped OT environments (power substations, nuclear, water treatment, defence) cannot send traffic metadata to the cloud. CMC is the correct on-prem aggregation layer for those sites. Always confirm cloud-connectivity policy before choosing Vantage.

▶ Watch a rogue device get discovered passively by Guardian

How a new unauthorised PLC is discovered through the Level 2 SPAN, identified by Guardian, and surfaced in Vantage. Press Play for the healthy path, then Break it to see the classic SPAN failure.

① PLC connectsAn unauthorised PLC is cabled into the Level 2 switch and starts broadcasting its protocol identity (e.g. Modbus/TCP) to the HMI.
② SPAN mirrorsThe managed Level 2 switch mirrors the new PLC traffic to Guardian's monitoring interface via the SPAN port.
③ Guardian DPIGuardian's passive DPI engine parses the Modbus/TCP frames, extracts the device identity (vendor, model, IP, MAC) and classifies it as a new Level 2 OT asset.
④ Alert in VantageGuardian raises a 'new unauthorised device' alert in Vantage with full asset profile — operator reviews and either authorises or quarantines.
Press Play to step through the healthy discovery path. Then press Break it.
Quick check · Q3 of 10 · Analyze

A power utility operates three substations that are fully air-gapped — no internet or cloud access. Which management topology fits?

Correct: c. CMC is the on-prem/virtual aggregation layer designed for air-gapped, sovereign, and regulated environments. Vantage SaaS requires cloud connectivity that the substations do not have. CMC consolidates the Guardians at each site with no internet dependency.
👉 So far: CMC aggregates Guardians on-prem for air-gapped/sovereign sites; Vantage SaaS scales across cloud-connected sites. Arc fills host-visibility gaps in zones without a SPAN port.

④ Multi-site & phased rollout — deploying without alarming OT

A successful Nozomi rollout follows four phases. Phase 1 — Passive discovery: deploy Guardian in monitor-only mode; let it build a complete asset inventory and learn the network baseline over 2–4 weeks. Zero alerts, zero OT impact — operators do not notice it. Phase 2 — Baselining & tuning: review the discovered asset list, map assets to Purdue levels and zones in CMC/Vantage, tune detection thresholds, suppress known-good behaviour, and reduce false positives before any alerting is enabled.

Phase 3 — Operational alerting: enable active alert policies. Integrate with SIEM/SOAR (Splunk, Microsoft Sentinel, IBM QRadar) and ticketing (ServiceNow, Jira). At this point the SOC begins receiving OT alerts in the same console as IT alerts. Phase 4 — Site expansion: replicate the pattern to additional sites and zones, onboard new Guardians to the existing CMC or Vantage instance, and add Arc for host-visibility gaps. Each new site repeats Phases 1–3 before going live with alerting.

Integration hooks

Nozomi integrates with firewalls (Palo Alto, Fortinet), identity platforms, and vulnerability scanners via syslog, REST APIs, and certified connectors — the SOC correlation layer gets richer OT asset context without re-architecting the security stack.

Figure 5 — Phased Nozomi rollout — 4 phases
Each phase adds capability without touching OT operations until the baseline is stable.Phased Nozomi rollout — 4 phasesPhase 1Passive discoveryPhase 2Baseline & tunePhase 3Live alertingPhase 4Site expansion
Each phase adds capability without touching OT operations until the baseline is stable.

Priya Nair, OT security architect at PowerGrid India Pvt. Ltd. in Nagpur, faces this

After Phase 1 Guardian deployment at a substation, Vantage shows 40% of Level 2 assets as 'unknown' — vendor and firmware cannot be identified.

Likely cause

The SPAN port on the heavily loaded Level 2 switch is dropping mirrored frames during peak SCADA polling cycles; Guardian misses the initial broadcast packets that carry device identity.

Diagnosis

In Vantage, filter assets by site and sort by 'Classification confidence' — the unknown assets all appear in the same VLAN. Check Guardian packet-drop counters; they show significant loss during peak windows. SPAN oversubscription is the root cause.

Vantage ▸ Assets ▸ Filter by site ▸ Classification + Guardian ▸ Diagnostics ▸ Packet counters
Fix

Replace the SPAN connection with a hardware TAP on the two most critical uplinks to Guardian. No switch config change needed after the TAP is installed — Guardian begins receiving every frame.

Verify

Within hours of the TAP install, Vantage asset inventory drops to under 5% unknown. The previously 'unknown' PLCs are now fully identified with vendor, model, and firmware version.

High unknown-asset rate = SPAN packet loss, not a Guardian bug

After Phase 1, open the Vantage or CMC asset inventory and check the 'Classification confidence' column. High unknown-asset rates (above 10–15%) are a strong signal of SPAN packet loss — not a Guardian limitation. Switch to a TAP on the affected link and re-run the discovery phase.

Quick check · Q4 of 10 · Evaluate

An OT team is nervous about enabling Nozomi alerts because past tool deployments triggered nuisance alerts during shift changes. What is the correct first phase?

Correct: d. Phase 1 is always passive discovery and baselining with no active alerts. This respects the OT team's concern — Guardian learns what normal looks like (shift changes, scheduled polling cycles, maintenance windows) before any alerting fires, dramatically reducing false positives.
👉 So far: A safe OT rollout is four phases: passive discovery → baseline and tune → operational alerting → expand to more sites. Never skip the baseline phase — it prevents the false-positive storms that erode OT team trust.

🤖 Ask the AI Tutor

Tap any question — instant, scoped to this lesson. No login, no waiting.

Pre-curated from vendor docs + community Q&A, scoped to this lesson. For a live prod issue, paste your export into chat.techclick.in.

📝 Wrap-up assessment — six more

You've answered 4 inline. Six left. 70% (7 of 10) marks the lesson complete on your profile. Tap Submit all answers at the end.

Q5 · Remember

Which Purdue level is the primary Guardian placement zone in most OT deployments?

Correct: a. Level 2 (supervisory) is where HMIs, SCADA servers, historians, and engineering workstations communicate. The aggregation switch here sees the richest OT traffic for asset discovery and anomaly detection from the fewest sensors. Levels 0–1 field devices are typically covered via the Level 2 SPAN.
Q6 · Understand

Why can a SPAN port fail to deliver all traffic to Guardian under heavy load?

Correct: c. Under switch load, the internal fabric prioritises live traffic. Mirrored (SPAN) frames are lower priority and can be silently dropped — Guardian receives an incomplete stream and may miss device-identity packets, causing high unknown-asset rates. A hardware TAP avoids this because it physically copies the signal before the switch fabric.
Q7 · Apply

A water treatment plant must keep all OT monitoring data on-site due to regulatory requirements. Which management topology should be used?

Correct: b. CMC is the on-prem/virtual aggregation layer built for regulated, air-gapped, and sovereign environments. It consolidates Guardian data entirely on-site with no cloud dependency. Vantage SaaS would require sending OT metadata to the cloud, violating the regulatory requirement.
Q8 · Analyze

An OT security team finds that 30% of Level 2 assets appear as 'unknown' in Vantage after two weeks of Guardian monitoring. The most likely root cause is…

Correct: d. A high unknown-asset rate after initial discovery strongly indicates SPAN packet loss. Guardian misses the initial broadcast/announcement packets that carry device identity when the SPAN drops frames under load. The fix is to replace the SPAN with a TAP on the affected uplinks.
Q9 · Evaluate

A site already has CMC aggregating its three Guardians. The CISO wants global visibility across 12 sites worldwide, some air-gapped. What is the best architecture?

Correct: b. The hybrid topology is the correct answer: CMC stays at air-gapped or sovereign sites (no cloud dependency), cloud-connected sites send directly to Vantage, and CMC can forward alert/summary data to Vantage so the CISO sees all 12 sites in one dashboard. This respects the air-gap constraint while giving global scale.
Q10 · Evaluate

An OT manager insists on skipping the baseline phase and enabling all Guardian alerts from day one. The strongest counter-argument is…

Correct: b. Without a baseline, every normal OT behaviour — shift-change traffic spikes, scheduled SCADA polling, planned maintenance — appears anomalous and triggers alerts. The resulting false-positive storm overwhelms the OT team and destroys confidence in the tool. The baseline phase (Phase 1) exists specifically to learn normal before alerting on abnormal.
Lesson complete — saved to your profile.
Almost! You need 70% (7 of 10) — re-read the path that tripped you up and tap "Try again".

🧠 In your own words

Type one line: why does a Nozomi deployment place Guardian at Level 2 rather than on every switch in the plant? Then compare with the expert version.

Expert version: Guardian is a passive sensor that needs to see traffic, not be everywhere traffic travels. Level 2 (supervisory) is the natural aggregation point where HMI-to-PLC, HMI-to-historian, and engineering-workstation traffic all converge on one switch — so a single Guardian on the Level 2 SPAN sees the entire control zone. Adding sensors on every switch would multiply hardware cost and management overhead without adding proportional visibility; the coverage principle is one Guardian per discrete network zone, not per device or per switch.

🗣 Teach a friend

Best way to lock it in — explain it in one line to a teammate. Tap to generate a paste-ready summary.

📖 Glossary

Guardian
Nozomi's core passive network sensor — passive DPI, asset discovery, anomaly and threat detection, vulnerability assessment. Deploys as appliance, VM, or container.
CMC (Central Management Console)
On-prem or virtual aggregator that consolidates multiple Guardian sensors across one or more sites without cloud dependency — the choice for air-gapped and sovereign environments.
Vantage
Nozomi's cloud-native SaaS management platform — aggregates Guardians and Arc, single pane of glass, scales across many sites, adds Vantage IQ (AI analytics).
Arc
Lightweight Nozomi host sensor adding user, process, USB, and session context — fills passive network monitoring blind spots in zones without a SPAN port.
SPAN port
Switch mirror port that copies selected port traffic to a monitoring port; free and easy but can drop mirrored frames under heavy switch load.
Network TAP
Dedicated hardware device that passively copies 100% of frames on a link to Guardian with zero packet loss — recommended for high-value chokepoints.
Purdue Model
Reference architecture for industrial control systems, Levels 0–5: field devices (0–1), supervisory (2), site operations (3), IT/OT DMZ (3.5), enterprise (4–5).
Guardian Air
Wireless Guardian variant for Wi-Fi, Bluetooth, cellular, and drone-spectrum OT visibility — complements wired Guardian in sites with wireless OT devices.

📚 Sources

  1. Nozomi Networks — Guardian sensor product page & data sheet. nozominetworks.com/products/guardian
  2. Nozomi Networks — Vantage SaaS platform overview & Vantage IQ. nozominetworks.com/products/vantage
  3. Nozomi Networks — Central Management Console (CMC) overview. nozominetworks.com/products/central-management-console
  4. Nozomi Networks — Arc endpoint sensor. nozominetworks.com/products/arc
  5. NIST SP 800-82 Rev 3 — Guide to Operational Technology (OT) Security: Purdue Reference Model for ICS. nist.gov
  6. ISA/IEC 62443 — Industrial Automation & Control Systems (IACS) security standards: zone and conduit model. isa.org

What's next?

Got the deployment topology? Next, go deep on how Guardian's hybrid detection engine builds its behaviour baseline, fires anomaly alerts, and how Nozomi Labs Threat Intelligence feeds enrich detection in real time.