What goes in a data center operations dashboard in Metabase?

A data center operations dashboard watches the facility itself: rack space, power draw and PUE, cooling and temperature, facility incidents, change windows, and hardware failures. It is the room-and-power counterpart to a server monitoring dashboard — servers tell you a host is struggling, this page tells you whether the hall it sits in is the reason — and Metabase builds it from DCIM, BMS, and sensor exports in your warehouse.

What does a data center operations dashboard look like?

Here’s the layout this guide builds. Facility KPIs open the page — space, power, PUE, and open incidents in one row — then power and cooling, the two constraints that actually cap a data center, and finally the operations layer: incidents, change windows, and the hardware failure mix that explains a bad month.

Which cards belong on a data center operations dashboard?

The eight below cover the facility’s three stories: how full it is, whether power and cooling are holding, and what operations did about it.

• Rack space utilization by data hall (row)
• Power headroom — IT load against contracted capacity (progress)
• IT load and PUE by week — load as bars, PUE on its own axis (combo)
• Cold-aisle temperature by hall, against the setpoint (line)
• Environmental alerts per week — temperature, humidity, water (stacked bar)
• Facility incidents by category (bar)
• Hardware failures by component, year to date (donut)
• Change windows this month, with hall, type, and status (table)

What data does the dashboard need?

• power_readings — IT load and total facility kWh per feed per interval, from the BMS, UPS, and smart PDUs.
• sensor_readings — temperature and humidity per sensor with hall, row, and aisle position.
• racks — rack inventory with hall, U capacity, U used, and power budget per rack.
• facility_incidents — incident ID, category (power, cooling, network, hardware, physical access), opened/resolved timestamps.
• change_windows — change ID, hall, type, scheduled window, and completion status from the change system.

How do you build it?

1. Export power and sensor data from the DCIM/BMS to the warehouse on a 5-minute schedule, keyed by feed, sensor, and hall — PUE and temperature cards are just rollups of these two tables.
2. Compute PUE in one shared model as facility kWh over IT kWh per day (the SQL below), so the KPI scalar and the weekly combo chart cannot drift apart.
3. Load rack inventory with U-used and power budget per rack, and build the space and headroom cards from it, grouped by hall.
4. Sync facility incidents and the change calendar from your ticketing system, categorized, so the operations section joins cleanly to halls.
5. Add filters for data hall, incident category, and date range, and set the environment cards to refresh every 15 minutes for the NOC display.

Example card SQL

SELECT
r.reading_date,
r.data_hall,
ROUND(SUM(r.it_load_kwh), 0)              AS it_load_kwh,
ROUND(SUM(r.facility_kwh), 0)             AS facility_kwh,
ROUND(SUM(r.facility_kwh)
  / NULLIF(SUM(r.it_load_kwh), 0), 2)     AS pue,
ROUND(AVG(t.cold_aisle_temp_c), 1)        AS avg_cold_aisle_c,
MAX(t.cold_aisle_temp_c)                  AS max_cold_aisle_c
FROM power_readings r
LEFT JOIN (
SELECT
  reading_date,
  data_hall,
  AVG(temp_c) AS cold_aisle_temp_c
FROM sensor_readings
WHERE sensor_type = 'cold_aisle'
GROUP BY reading_date, data_hall
) t USING (reading_date, data_hall)
WHERE r.reading_date >= CURRENT_DATE - 90
GROUP BY r.reading_date, r.data_hall
ORDER BY r.reading_date, r.data_hall;

Related

Metrics

Integrations

Dashboards

FAQ