Skip to content
Dualistic

DCIMate by Dualistic

Agentic Datacenters

DCIMate places workloads by energy, powers down idle hardware on its own, and runs a thermo-fluid twin of the room. Validated end to end on production-class hardware.

Book a call See the twin

Validated on real hardware

Results that hold up to measurement

How DCIMate did it
−31% Energy per unit of work Measured at the wall, p < 0.05
+44% Throughput Same jobs and hardware
−52% Facility energy when idle Measured, autonomous power-off
~−41% Annual electricity bill Modeled, conservative

How it works

Connect, simulate, optimize

One agent inside the datacenter, one physics model of the room, and a scheduler that acts on both.

0.01

Connect

A lightweight agent, dcim-claw, runs inside your datacenter and continuously streams metrics from every sensor, server, and network device into DCIMate.

  • 01
    Temperature sensors
    Rack-level and ambient temperature, sampled every second.
  • 02
    PDU and power meters
    Per-device power draw and cumulative energy consumption.
  • 03
    Network and compute
    Switch utilization, CPU load, and memory metrics, streamed live.
dcim-claw · ingest illustrative
DCIMate dcim-claw T° TEMP W POWER ≋ NET ⊡ CPU ❄ HVAC ▦ MEM
digital-twin · airflow-sim illustrative

0.02

Simulate

DCIMate builds a live digital twin of your datacenter, a real-time model that mirrors airflow, thermal dynamics, and workload distribution from sensor to ceiling.

  • 01
    Thermal mapping
    Color-coded rack heatmaps updated in real time. Spot hotspots before they cause failures.
  • 02
    Airflow modeling
    Simulate cold and hot aisle dynamics, and predict temperature under different load scenarios.
  • 03
    Predictive analysis
    Run what-if scenarios: add a rack, spike load, and see the thermal impact instantly.

0.03

Optimize

Agents continuously analyze datacenter state and rebalance workloads, preventing overheating, cutting energy waste, and alerting you before issues arise.

  • 01
    Workload rebalancing
    Automatically migrate tasks from overloaded servers to idle ones.
  • 02
    Proactive alerts
    Get notified before thermal thresholds are breached, not after a failure.
  • 03
    Energy efficiency
    Reduce PUE by consolidating workloads and spinning down underutilized hardware.
workload-monitor illustrative
unbalanced 6 nodes
91%
S1
12%
S2
88%
S3
9%
S4
94%
S5
17%
S6
Energy saved
+0%

Digital twin

The cluster we measured, as you see it

This is the cabinet from the DCIMate twin, not an illustration of one. The eight nodes in the frame are the eight machines the validation campaign actually instrumented.

A server cabinet from the DCIMate digital twin: eight compute nodes racked below a top-of-rack switch.
Instrumented nodes
8
Physical cores
472
Measured idle draw
2.34 kW

Platform

What DCIMate does

A control plane for energy-aware scheduling, thermal simulation, and agentic operations.

FIG.01

Workload optimization

Agents place and migrate jobs by energy cost, not just by free cores, and power down what nothing is using.

FIG.02

Thermo-fluid digital twin

A physics model of the room predicts airflow and rack inlet temperature, with an ML surrogate for fast inference.

FIG.03

dcim-claw agent

Query sensor data, trigger operations, and get answers about the facility in plain language.

Everything DCIMate does

Contact

Talk to us about your datacenter

Tell us what you run and where the energy goes. We will tell you what DCIMate would do with it.

Office
via Crispi 56, 34126 Trieste, Italy
Email
info@dualistic.it
Phone
+39 040 9719258
LinkedIn
Giuseppe Piero Brandino, CEO

Contact us

Email us Call us

We reply to everything. What we do with your message: privacy policy.