K8S scheduling
by energy, price
and forecast
K8S scheduling
by energy, price
and forecast
K8S scheduling
by energy, price
and forecast
Intelligent Orchestration runs workloads where they cost least, scales them before peaks and moves them off failing nodes.
Intelligent Orchestration runs workloads where they cost least, scales them before peaks and moves them off failing nodes.
Where does a new application run?
Placement score per site, 0–100 · example data
CPU and memory
Energy supply
Electricity price
Placement weighs energy supply and electricity price
Placement weighs energy supply and electricity price
The orchestrator scores every site on CPU, memory, energy supply and electricity price. It runs each workload where power costs least. You pay less for electricity. More of your power comes from renewable sources.
The orchestrator scores every site on CPU, memory, energy supply and electricity price. It runs each workload where power costs least. You pay less for electricity. More of your power comes from renewable sources.
Every site runs its own cluster and control plane
Every site runs its own cluster and control plane
The sites connect peer to peer, with no central controller. A site that loses its connection keeps managing its own workloads. If a site fails, the other sites start its applications. Your services keep running on unstable networks and at remote sites.
The sites connect peer to peer, with no central controller. A site that loses its connection keeps managing its own workloads. If a site fails, the other sites start its applications. Your services keep running on unstable networks and at remote sites.
Reserved capacity matches actual usage
Reserved capacity matches actual usage
Kubernetes workloads leave 69% of their reserved CPU unused (CAST AI, 2026). The orchestrator measures what each workload uses and cuts its reservation to fit. It packs workloads onto fewer nodes and switches idle nodes off until demand grows. You run the same workloads on fewer nodes and pay for less power.
Kubernetes workloads leave 69% of their reserved CPU unused (CAST AI, 2026). The orchestrator measures what each workload uses and cuts its reservation to fit. It packs workloads onto fewer nodes and switches idle nodes off until demand grows. You run the same workloads on fewer nodes and pay for less power.
One job, vCPU
vCPU reserved, used, and reserved after the resize · example data
The orchestrator forecasts load, energy and node health
The orchestrator forecasts load, energy and node health
It learns from the cluster’s history. It acts before load peaks, energy changes and failing nodes affect your applications.
It learns from the cluster’s history. It acts before load peaks, energy changes and failing nodes affect your applications.
Autoscaling
Autoscaling
Traffic has peaked every Monday at 9:00 for four weeks. The orchestrator adds capacity at 8:30. The capacity is ready when the peak starts.
Traffic has peaked every Monday at 9:00 for four weeks. The orchestrator adds capacity at 8:30. The capacity is ready when the peak starts.
Monday load forecast
Load and capacity, % of maximum capacity · half-hour slots · example data
Load forecast
Capacity
Autoscaling
Traffic has peaked every Monday at 9:00 for four weeks. The orchestrator adds capacity at 8:30. The capacity is ready when the peak starts.
Placement by forecast
Placement by forecast
A job must finish by the end of the day. The forecast shows the most renewable energy from 14:00 to 17:00. The orchestrator holds the job until 14:00. The job runs in these hours and finishes before the end of the day.
A job must finish by the end of the day. The forecast shows the most renewable energy from 14:00 to 17:00. The orchestrator holds the job until 14:00. The job runs in these hours and finishes before the end of the day.
Renewable energy forecast
Renewable share of grid power, % per hour · example data
Placement by forecast
A job must finish by the end of the day. The forecast shows the most renewable energy from 14:00 to 17:00. The orchestrator holds the job until 14:00. The job runs in these hours and finishes before the end of the day.
Self-healing
Self-healing
The health score of node 3 falls to 41 out of 100. The orchestrator moves its workloads to node 7 before latency grows. Your applications keep their normal response time.
The health score of node 3 falls to 41 out of 100. The orchestrator moves its workloads to node 7 before latency grows. Your applications keep their normal response time.
Node health score
Score 0–100, every 15 minutes · example data
Self-healing
The health score of node 3 falls to 41 out of 100. The orchestrator moves its workloads to node 7 before latency grows. Your applications keep their normal response time.
Request an Intelligent Orchestration demo
Request an Intelligent Orchestration demo
We will show how Intelligent Orchestration distributes your workloads across your sites by priority, resources, energy and price.
We will show how Intelligent Orchestration distributes your workloads across your sites by priority, resources, energy and price.
Request a demo