Known Limits

Capacity and Scope

• The 7‑day NIS2 evidence export promise assumes:
  • ≤ 100 sites in scope
  • Standard SNMP v2c/v3 and Redfish collectors (read‑only)
  • Outbound egress permitted to EU endpoints (TLS 443)
  • Baseline configuration delivered on Day 0
• Evaluation fit is strongest at 20–100 sites; < 5 sites are typically disqualified.

Protocols and Integrations

• Supported at GA: SNMP v2c/v3, Redfish, IPMI (read‑only), HTTPS health checks, syslog ingestion (metadata)
• Not yet supported at GA: BACnet/Modbus building systems (post‑pilot roadmap), remote control actions (two‑person rule planned), broad SCIM user provisioning (limited support; roadmap)
• API and webhooks are available; connector catalog is intentionally small at GA to protect reliability.

Data and Residency

• Telemetry is designed to contain zero PII. Misconfiguration that injects PII into free‑text fields is the customer's responsibility and should be avoided.
• EU‑only hosting with named providers; no cross‑border transfer unless explicitly configured by the customer for external backups.
• Evidence exports include control mapping, lineage, and a reproducible hash.

Offline Behavior and Continuity

• Default offline buffer is ≥ 24 hours per agent; GA SLO validated at 30/60/120‑minute outage windows with lossless backfill.
• Extremely long outages beyond local storage capacity will not retain full‑fidelity metrics.
• Continuity log is append‑only and hash‑chained; if local system clocks drift heavily, continuity ranges may display skew until NTP resync.

Security Model

• SSO included at all paid tiers (OIDC/SAML). Broad SCIM/Just‑in‑Time auto‑provisioning is limited at GA.
• Audit log is append‑only and tamper‑evident; administrative deletes are not supported by design.
• Agent runs with least privileges required for the selected collectors; root access is not required for SNMP/Redfish polling.

Performance Budgets

• Typical agent budgets under GA workloads:
  • CPU: < 3% average on a 2‑core x86_64 host
  • RAM: < 150 MB
  • Disk buffer: 200–500 MB per 24 h (depends on device count)
  • Egress: 50–200 MB per site per day (depends on sampling)
• Budgets may vary with device density and sampling rates; acceptance testing verifies ceilings per environment.

Device and OS Support

• Collectors validated on common Linux distributions (Ubuntu LTS, Debian stable) and containerized deployments.
• Windows endpoints are monitored via SNMP/Redfish where available; native WMI collectors are not GA at this time.
• At least one SNMP PDU and one Redfish‑capable server are required for pilot proof.

Reliability and SLA Notes

• Product is designed for monitoring and compliance evidence; it is not a real‑time control system.
• Ingest lag SLO is measured as P95 over a rolling window; transient spikes can occur during backfill after outages.
• The 7‑day export promise is void if preconditions are not met; once rectified, the promise applies to the remaining scope.

Browser and Network Requirements

• Supported browsers: current Chrome, Edge, Firefox, Safari
• TLS 1.2+ outbound required to EU endpoints
• Accurate NTP time sync required on monitored hosts and collectors
• Proxy environments must allow websockets or HTTP/2 for live views; fallback polling increases lag

Roadmap Items (post‑GA)

• Additional adapters: BACnet/Modbus for building systems
• Advanced anomaly detection beyond thresholds
• Remote actions (reboot/outlet/setpoints) with a two‑person rule
• Expanded SCIM and granular SSO mappings
• Carbon‑aware scheduling and energy orchestration

Known Vendor Quirks

• Some PDU models report non‑monotonic uptime counters over SNMP after power events; continuity logic accounts for this but may display short "step" artifacts.
• Certain Redfish implementations rate‑limit aggressively; sampling intervals may be increased to maintain reliability.
• NAT hairpinning can block local dashboards; prefer direct outbound patterns from the agent.