Roadmap/Cisco Security Platform Lab — side by side + T1-T4 v0.2·07·2026
§ Roadmap Cisco Security Platform Lab · What's live, what's next

Every layer of Cisco Secure AI Factory. Same products. Free tier. Running end-to-end in a home lab. The Secure AI Factory reference architecture — Access, AI Runtime, Data, Compute, Network, wrapped by Security and Observability — is what Cisco+NVIDIA jointly recommend to enterprises building AI infrastructure. This lab is that same stack, downsized to what any DC SE can rebuild at zero incremental cost.

Live in tenant: Secure Firewall 1210CE · FTD 7.6 FDM · SCC AI Defense ThousandEyes · Talos
§ Architecture Cisco Secure AI Factory with NVIDIA — side by side with the home-lab version

Same reference architecture Cisco+NVIDIA co-market to enterprise DC customers. Left column is what runs in a real Secure AI Factory deployment. Right column is what runs in this lab. Where the products are identical, they're flagged SAME. For the layer-only view without the comparison, see /reference/.

Cisco Secure AI Factory with NVIDIA
Reference · lab is a representation
Observability wrap ThousandEyes·Talos intel·Splunk (future)·Cisco XDR (future)
L1

Access & Applications

User → chatbot UI
Secure AI Factory Enterprise IdP · Duo · Umbrella · TLS ingress · AI Defense on API edge
This lab SameDuo (free) · Umbrella (eval) · Caddy TLS ingress on OCI · AI Defense demo host
L2

AI Runtime

Inference · models
Secure AI Factory NVIDIA AI Enterprise · NIM · Nemotron · OpenShell · Agent Toolkit
This lab Same NIMNVIDIA NIM Free (hosted API · up to 5,000 credits total: 1K signup + 4K biz-email upgrade · 40 RPM) · Nemotron chat model
L3

Data

Sources · RAG
Secure AI Factory Enterprise data lake · vector DBs (Pinecone / pgvector) · RAG index · certified storage array
This lab NetBox (network inventory as chatbot data source) · PostgreSQL · pgvector for future RAG
L4

Compute

CPU · GPU · substrate
Secure AI Factory Cisco UCS C885A M8 + NVIDIA HGX H200 SXM (or AMD MI300X/MI350X OAM) · Cisco Unified Edge (edge tier) · BlueField-3 DPU
This lab OCI Free Tier VMs (2 AMD + Ampere A1 · 2 OCPU / 12 GB) · Cisco UCS C220 M8 (same rack, no GPU)
L5

Network Fabric

L3 · SDN · edge
Secure AI Factory Cisco N9100 (Spectrum-6) · Silicon One G300/P200 · Nexus One (mgmt + Hyperfabric) · east/west policy
This lab Same FTDCisco Secure Firewall 1210CE (FTD 7.6 — same software as 3100/4200) · home LAN switch
Security wrap Cisco AI Defense·Cisco Hybrid Mesh Firewall on BlueField DPUs·NeMo Guardrails (agent)·Cisco XDR (future)
Operational context · Cisco Cloud Control Controlled Availability · 2026-06-02

The lab is the bench. Cloud Control will be the cockpit.

Cisco Cloud Control is Cisco's AI-native, agentic operating platform spanning Networking, Security, Compute, Observability, and Collaboration — the foundation of Cisco's AgenticOps model. It entered Controlled Availability (US) on 2026-06-02 at Cisco Live, with Global Availability to follow. When customers run this portfolio at production scale, Cloud Control is the cockpit. This lab isn't a substitute; it's the hands-on foundation SEs need to be fluent in the portfolio it will orchestrate. Build the stack here now. Be ready to operate it there.

§ Free-tier providers Four providers, stitched into the architecture above

The Secure AI Factory layers above run on four free-tier substrates. Any SE on the team can reproduce the whole architecture by signing up at four URLs. Full walk-through in the OCI Free Tier Lab guide.

Cloud services

$0/mo · perpetual
Oracle Cloud Infrastructure · Always-Free

Hosts every public-facing surface — this site (batman4ny.github.io until pages was ready; roadmap now lives here) and the AI Defense demo. One AMD instance plus one ARM instance covers the distributed lab's public face at zero cost.

Currently deployedbatman4ny.github.io (GitHub Pages) · aidefense-demo.uppernyack.com (OCI Ampere ARM)
Compute2× AMD E2.1.Micro + 2 OCPU / 12 GB ARM A1
Storage200 GB block · 20 GB object
Egress10 TB / mo

AI inference

Free · rate-limited
NVIDIA NIM · Build.nvidia.com Free Tier

Managed API endpoints for open-weights models — Nemotron, Llama, embedding stacks. No local GPU needed. Signup grants 1,000 credits (up to 5,000 total by upgrading to a business email under a free 90-day NVIDIA AI Enterprise trial); credits are a one-time allocation, not monthly-replenishing. Rate limit is 40 RPM by default.

Currently deployedNemotron chat backend for the NetBox chatbot demo
Accessbuild.nvidia.com API key
CreditsUp to 5,000 total (one-time · 1K signup + 4K biz-email)
Rate limit40 RPM default per model
Runs whereNVIDIA-hosted (no local GPU)

On-prem lab

One-time HW · $0 recurring
Home rack · Cisco SE-provided hardware

Where the story earns the "hybrid" label. The 1210CE is one node — real enterprise gateway with FTD 7.6 and full evaluation licensing. Adjacent: a UCS server for compute workloads and a Raspberry-Pi-based OOB console. Build walk-through in the 1210CE Build Guide (WIP).

Currently deployedSecure Firewall 1210CE (FTD 7.6, FDM-managed, SCC-onboarding pending) · ConsolePi (rack OOB)
Firewall1210CE · 90-day eval, full features
ComputeUCS C220 M8 (planned same rack)
ConsolePi 3B+ · USB-C to FW · ser2net :9000

Observability

$0 (Cisco-provided)
ThousandEyes · Talos · Cisco tenant

The layer that sees every hop the request takes. Local ThousandEyes agent at the home lab, plus 200+ ThousandEyes cloud agents worldwide, plus Talos intelligence feeding IPS at every FTD instance. Splunk and XDR slot in here later as they roll out.

Currently deployedThousandEyes agent on home LAN · Talos feeds active in 1210CE FTD
Local agentRaspberry Pi at home
Cloud agents200+ cities via Cisco tenant
Intel feedsTalos IPS/URL updates — automatic
Rate limit240 req/min TE API
§ Rollout T1 – T4 · What's next, in order
T1

Same tenant, direct integration

Next up · SCC-native

Products that already share the SCC pane. Each is an "enable a switch, gain a story" move — no new logins for customers, immediate cross-product correlation.

Cisco XDR

Correlates FW + endpoint + email + cloud. Orchestration layer above every other Tier 1 product.

Umbrella

Filters at DNS layer + SASE tunnel from FTD. Blocks threats before they hit the firewall.

Duo

Authenticates admin access to FDM, SCC, everything. MFA + device trust in one move.

Secure Endpoint

Closes the endpoint side of the kill chain. Feeds XDR with process/file telemetry FW can't see.

Secure Email Threat Defense

Covers the email attack vector. Most breaches still start there — completes the "every door" story.

Talos intel

Powers IPS + URL filtering by default. Not a product to sell — proof of platform depth.

T2

Observability & telemetry

The "why is it slow" answer

Firewall vendors get blamed for problems they don't own. These products prove where the fault actually lives — the SE's most re-used demo.

ThousandEyes

Sees internet + SaaS + FW paths end-to-end. Answers "is it the ISP, Cloudflare, the app, or my box?" in one screen.

Splunk

Ingests FTD syslog + Snort events. Since Cisco owns it, the "platform SIEM" answer for enterprise DC.

AppDynamics · Splunk Observability

Extends observability to the app layer (now part of Splunk Observability suite). Rounds out the FSO story.

T3

Cloud-native security

Where DC meets K8s

Different problem space than perimeter FW — east/west inside clusters. Same platform pitch. Pairs cleanly with the AI Defense story already in the tenant.

Isovalent

Enforces workload-to-workload policy in K8s via eBPF (Cilium / Tetragon). Cisco acquired 2024. Complements FTD's north/south.

Panoptica · Outshift

Covers CNAPP — cloud posture, container, K8s, app. Sits above Isovalent in the stack.

Multicloud Defense

Protects multi-cloud network traffic (Valtix acquisition, 2023). Extends FTD-style policy across AWS, Azure, GCP.

T4

DC breadth & AI infrastructure

Adjacent but on-brand

Not core to the security pitch, but every DC SE needs credible answers when customers ask "what about…". The lab covers them by proximity.

Cisco Secure AI Factory

Validates AI-DC reference architectures (NVIDIA partnership). The "how do we build AI infra securely" answer.

UCS + Intersight

Manages a UCS C220 M8 in the same rack. Same ConsolePi already consoles it — one lab, two products.

Nexus One

Cockpits DC network fabric (subsumes Nexus Dashboard + Hyperfabric). Where switch + UCS stories converge.

15+
The point

Fifteen-plus Cisco products, spread across the five architectural layers of Secure AI Factory, running end-to-end from free tiers any DC SE can rebuild. When a peer asks "what would Secure AI Factory look like at our scale?" — the answer opens in a browser tab, not a PDF.