Infrastructure analytics platform
How many more nodes can you safely install? Rack 12 is at 78%.
Catch problems before they become outages — anomaly detection that learns what "normal" looks like for every device, every hour of the day, and capacity answers computed from real utilization instead of nameplate ratings.
Shared trial workspace — read-only, resets nightly. No signup, no install.
- Demo-org PUE
- 1.58
- Source: seeded demo number — not a production claim, and not the tour-live sustainability reading (~1.5).
- Demo fleet under management
- 224 devices
- Source: demo-org inventory — 3 datacenters, 11 racks, 224 devices with 30 days of backfilled telemetry (053).
- Infra cutover checks
- 3 / 3 green
- Source: infra phase (047) — apex static serves zero-JS, console answers on platform.nodx-ops.com, legacy host redirects.
Nameplate ratings decided your last install. The overage bill decided the one before.
20%
of contracted power typically sits stranded — paid for, never used. NodX meters per-device p95 draw over 7 days of hourly telemetry so the headroom is measured, not guessed.
3σ
Static thresholds page you after the damage. Streaming z-scores against 24 hour-of-day baselines per device catch drift while there is still time to act — and transients that do not persist never fire.
30 d / 1 y / 2 y
Raw samples kept 30 days, hourly rollups 1 year, daily rollups 2 years. Every chart and report reads from the same retention ladder — nothing silently aged out.
Sense → Stream → Alert. Three steps, no professional-services engagement.
1 · Sense
One lightweight agent per server reads sysfs, IPMI, and GPUs every 15 s and posts a single object per reading. BMCs and legacy gear arrive over the same envelope via Redfish and SNMP pollers; facility sensors publish MQTT.
2 · Stream
Readings land in a Redis queue and a Go worker writes them to a TimescaleDB hypertable — deduplicated on (timestamp, device, metric), feeding 24 hourly baseline accumulators per device-metric as they arrive.
3 · Alert
Rules evaluate every 10 s — threshold or z-score, with hysteresis, escalation, and stale-signal auto-resolve. Firing incidents dispatch HMAC-signed webhooks to Slack, Teams, Discord, or PagerDuty.
- HTTP
- MQTT
- Redfish
- SNMP
Six screens your operators will actually open.
Anomaly detection
3σ deviations against learned hour-of-day baselines, duration-filtered. See it fire on real spikes in the alerts console.
Capacity planning
Per-device p95 power profiles, stranded-power math, and a node-fit count per rack. Worked example on the capacity screen.
Live PUE
Site input power ÷ IT power, rolling average plus p95, on a gauge —auditable to the meter readings.
Rack heatmaps
U-slot tiles colored cyan→red from the latest telemetry per slot, click through to sparklines. Open any rack in the explorer.
Alert lifecycle
firing → acknowledged → escalated → resolved, with ack-timeout escalation and deadband auto-resolve. Rules and incidents live in one console.
Bulk import
500-device JSON or NetBox/Device42 CSV imports with mandatory dry-run preview and skip/merge/replace conflicts — migrations without the spreadsheet week.
Three deployment versions.
NodX ships as the same platform in three forms. Pick by network posture, not by feature checklist — the telemetry pipeline, alerting engine, and dashboards are identical across all three.
Standard SaaS
Multi-tenant cloud operations dashboard on platform.nodx-ops.com. No hardware to run, agents report outbound over HTTPS to the managed ingestion edge.
On-Premise
Air-gapped enterprise installation on your own hosts, no phone-home. Licensed by offline rack keycard, with a one-command support bundle for diagnostics.
Lite Agent
Ultra-light telemetry runner for single nodes. Reads sysfs, IPMI, and GPU sensors locally and pipes samples out with negligible overhead, spooling to disk across network drops.
Open where it counts.
The telemetry agent and the pilot operator kit are public on GitHub — read the pipeline, inspect the install, or run it against your own gear before you talk to us. No PDF datasheets; the code is the documentation.
Inside the live product in under a minute. No signup, no sales call required.
Deploy the ultra-light telemetry runner locally in under 60 seconds.
One-command install
curl -sSL https://nodx-ops.com | sh -s -- --key=nx_live_trial_053
Verify: nodx-agent status --check-pipeline
No framework required
Zero dependencies. Pure shell script. No Node, no Python, no Docker.
What NodX does not do (yet)
- No power control. Remote power-cycle and outlet switching stay with the PDU vendors — NodX monitors everything and controls nothing.
- No 3D twin. The rack view is a 2D U-slot heatmap with live values, not a walkthrough model. Cheaper truth over prettier models.
- No what-if simulation. Capacity answers come from measured p95 history, not pre-install impact models or feed-failure drills.
Gaps held honestly per the 2026 DCIM landscape review — they bound the claims above. If any of these is your buying criterion, say so on the contact form and we will tell you plainly whether the roadmap covers it.
Inside the live product in under a minute. No signup, no sales call required.
Frequently asked
How does the NodX telemetry agent capture hardware metrics without impacting system performance?
NodX utilizes a lightweight compiled agent that isolates telemetry processing loops to non-critical virtual column allocations, preventing agent execution paths from ever competing with primary host infrastructure compute cycles.
Does NodX support auto-discovery of devices on the network?
Yes — passive, not a noisy sweep. A discovery listener watches ARP/LLDP/mDNS traffic and an SNMP fingerprinter identifies known vendor seeds; everything else arrives via agent self-registration, API, or bulk CSV/JSON import. No broadcast storms on your subnets.
What makes NodX's pricing model different from Zabbix or Prometheus?
NodX uses flat per-rack pricing instead of per-device or per-GB consumption meters. Consumption meters scale your bill with your telemetry volume — the more you watch, the more you pay. Flat per-rack keeps full-fidelity monitoring the default instead of a budget decision.
Is SCIM (automated user provisioning) available for all NodX versions?
Yes. SCIM support is now included across Standard SaaS, On‑Premise, and Lite Agent deployments. Enterprise‑only restrictions have been removed, enabling consistent single‑sign‑on provisioning for every tier.