Data Center Monitoring

MQTT for Data Center Monitoring

PDUs, cooling units, UPS systems and rack sensors report power, temperature and humidity across every hall and site. Edge gateways publish those values once over MQTT, and your DCIM, NOC alerting and energy reporting each subscribe to what they need.

No card required

What you can build

Build your data center monitoring on ConnectMQ

You build the DCIM views, alerts and reports, ConnectMQ moves the data. Edge gateways in each hall read PDUs, cooling units, UPS systems and rack sensors and publish on change; your DCIM, capacity planning, NOC alerting and energy reporting each subscribe in parallel. Your BMS stays in control of the equipment.

Your halls and sites
Rack PDUsdc/{site}/{hall}/rack-{id}/pdu/power
CRAC / CRAH unitsdc/{site}/{hall}/crac-{id}/state
Rack temperature & humiditydc/{site}/{hall}/rack-{id}/env
UPS systemsdc/{site}/ups-{id}/status
ConnectMQ
  • TLS on 8883
  • per-device ACL
  • threshold alerts
  • retained messages
What you build on top
DCIM dashboards
Capacity planning
NOC alerting
Energy & PUE reporting
What you get

Data Center Monitoring on ConnectMQ

  • Power, cooling and rack environment data from every hall on one broker
  • Published on change — no central poller hammering every device
  • Retained last values, so a new dashboard is complete straight away
  • Separate credentials and topic permissions for every site
example topic design
dc/dfw1/hall-b/rack-b14/env
dc/dfw1/hall-b/rack-b14/pdu-a/power
dc/dfw1/hall-b/crac-02/state         (retained)
Report by exception

Stop polling every PDU and CRAC

The usual set-up is a collector at each site that polls every PDU, cooling unit and UPS over SNMP or Modbus, every minute or so, whether anything changed or not. With ConnectMQ, an edge gateway in each hall still reads the equipment locally but publishes only when a value changes — and every system that needs it subscribes.

On-site collectors polling

  • Every PDU, CRAC and UPS is polled on a fixed cycle
  • A hot spot waits for the next poll before anyone sees it
  • Each new tool adds another poller or another export
  • Every site’s data sits in its own collector

ConnectMQ

  • Edge gateways publish on change, over TLS
  • DCIM, NOC alerting and reporting all subscribe to the same topics
  • Retained last values and rules & alerts on the broker
  • Per-site credentials and topic permissions
Real-life example

Cool down a hot spot before the servers throttle

A common data center MQTT setup: an edge gateway in the hall publishes rack temperatures on change, the DCIM in the NOC subscribes, and the site’s BMS gateway listens for cooling setpoints. Hall and NOC both connect out to ConnectMQ over TLS; the BMS stays in charge of the cooling units.

ConnectMQMQTT broker · TLS 8883Incident INC-2291 openedHALL B · B-1424.1 °C31.2 °C ▲28.4 °C ▼INC-2291CRAC-02 18 °COperations centreDCIM · subscribed to every hallCold aisleHALL BB-12B-13B-14B-15B-14 · Inlet24.1 °C27.8 °C31.2 °C28.4 °C26.4 °C20 °C18 °C45 %85 %CRAC-02EDGE-BBMS-BBACnetData hall BEdge gateway + BMS gatewaydc/dfw1/hall-b/rack-b14/env {"inlet":31.2}dc/dfw1/hall-b/rack-b14/env {"inlet":31.2}dc/dfw1/hall-b/crac-02/set {"supply":18}dc/dfw1/hall-b/crac-02/set {"supply":18}dc/dfw1/hall-b/crac-02/state {"supply":18,"fan":85}NOC alerts · nowRack B-14 at 31.2 °C — CRAC-02 set to 18 °CConnectMQMQTT broker · TLS 8883Incident INC-2291 openedHALL B · B-1424.1 °C31.2 °C ▲28.4 °C ▼INC-2291CRAC-02 18 °COperations centreDCIM · subscribed to every hallCold aisleHALL BB-12B-13B-14B-15B-14 · Inlet24.1 °C27.8 °C31.2 °C28.4 °C26.4 °C20 °C18 °C45 %85 %CRAC-02EDGE-BBMS-BBACnetData hall BEdge gateway + BMS gatewaydc/dfw1/hall-b/rack-b14/env {"inlet":31.2}dc/dfw1/hall-b/rack-b14/env {"inlet":31.2}dc/dfw1/hall-b/crac-02/set {"supply":18}dc/dfw1/hall-b/crac-02/set {"supply":18}dc/dfw1/hall-b/crac-02/state {"supply":18,"fan":85}NOC alerts · nowRack B-14 at 31.2 °C — CRAC-02 set to 18 °C
  1. 03:14 Edge gateway in hall B

    Reads the rack sensors on the local network and publishes only when a value changes: rack B-14’s inlet air has reached 31.2 °C.

    dc/dfw1/hall-b/rack-b14/env → {"inlet":31.2,"rh":38}
  2. 03:14 DCIM in the NOC

    Subscribed to every hall through ConnectMQ. Its rule for inlets above 27 °C opens incident INC-2291 and requests an 18 °C supply-air setpoint on CRAC-02.

    dc/dfw1/hall-b/crac-02/set → {"supply":18}
  3. 03:15 BMS gateway → CRAC-02

    The only client allowed to write to hall B’s cooling: it applies the setpoint over BACnet, CRAC-02 raises its fan to 85 % and the gateway publishes the new state.

    dc/dfw1/hall-b/crac-02/state → {"supply":18,"fan":85}
  4. 03:21 On-call engineer

    Gets the NOC notification with the values. A few minutes later rack B-14 is back at 26.4 °C, and the incident shows every step.

When a new NOC screen connects, it subscribes to dc/dfw1/hall-b/+/env and gets every rack’s last reading at once — the edge gateway publishes every reading with the retain flag, so nothing waits for the next change.

Pilot

Pilot one data hall this week

Start on the Free plan with no card. Basic covers up to 100 concurrent connections — enough for the edge and BMS gateways of a few halls and the systems that subscribe to them. Move to Scale when you add sites.