NewCompare CPU & GPU pricing across AWS, Azure & GCP

Kubernetes Cluster Autoscaling

Remove the nodes you do not need, plan capacity and see why pods wait across multicloud clusters. Works with any autoscaler.

Book a demoGet started

Companies who slashed their Kubernetes
spend using DevZero

DATABAHN
Starburst
Fi
Outerbounds
Codilas
personality pool
Onnitech
OpenObserve
Parsimo
Dentira
DATABAHN
Starburst
Fi
Outerbounds
Codilas
personality pool
Onnitech
OpenObserve
Parsimo
Dentira

Compact Idle Nodes Automatically

DevZero removes the nodes your cluster does not need. Every 15 minutes it finds empty or underutilized nodes whose pods fit on the rest of the group, checks for unschedulable pods, then scales down and deletes them.

Control How Compaction Runs

  • Choose a consolidation policy and how long a node waits before it is removed
  • Set disruption budgets with a cron schedule
  • Remove empty nodes first, then replace or consolidate the rest

Plan Capacity Before You Scale

Choose which capacity each pool may use and set limits on what it can spend, so the right nodes are ready when pods arrive.

Spot, On-Demand And Reserved Capacity

  • Allow spot, on-demand or reserved capacity on each pool
  • Review suggestions that weigh replicas, disruption budgets and startup time before you move a pool to spot

Static Capacity And Budgets

  • Run a static pool with a fixed node count
  • Set disruption budgets to limit how many nodes change at once

Manage Node Groups By Workload

DevZero reads the workloads on each node group before it proposes a change.

Simulate Before You Change

  • Select nodes to retire and see where their pods land
  • See which pods would need new nodes
  • Check Reserved Instance and Savings Plan coverage on AWS and Google Cloud
  • Run it read only, so nothing changes in your cluster

Reviewed Suggestions

  • See the workloads behind each suggestion
  • Accept a suggestion when you are ready, nothing applies on its own
  • Find duplicate node policies, also across clusters

See What Your Nodes And Workloads Cost

Check how your node pools and workloads use what they request, and what each one costs. Works on clusters running Karpenter, Cluster Autoscaler or both.

Node Analysis

  • Track node count, monthly projection and cost for the period
  • Find the most underutilized node and your overall underutilization
  • Check how much of your spend a Savings Plan or Reserved Instance covers

Workload Analysis

  • Compare requested and used CPU and memory for every workload
  • See which few workloads drive most of your cost

Scale Clusters Your Way

Node policy features that cut idle capacity across your clusters.

  • Cluster Autoscaling

    DevZero adds nodes when pods cannot schedule and removes empty or underutilized nodes.

  • Flexible Instance Shapes

    Allow a range of instance families and sizes in a node policy, so no node group is tied to one shape.

  • Savings Plan Awareness

    Check whether a Reserved Instance or Savings Plan covers a node before you move families.

  • Spot-to-Spot Consolidation

    DevZero replaces spot nodes with cheaper spot nodes, not only consolidates onto on-demand capacity.

  • Manage Existing Karpenter

    Bring your existing Karpenter NodePools and NodeClasses into DevZero as node policies.

  • Static Capacity

    Keep a fixed number of nodes in a pool when workloads need capacity held ready.

Keep Your Current Autoscaler

DevZero works with the autoscaler you run today, and you can simulate the move first.

  • Any Autoscaler

    DevZero works with Karpenter and Cluster Autoscaler. You review proposed node policies before anything applies.

  • Open-Source Karpenter Compatible

    You keep your NodePool names and the same NodePool and NodeClass resources.

  • Multicloud

    DevZero runs on AWS, Azure and Google Cloud. You write one node policy and it compiles for each cloud.

Results From Teams Running DevZero

What teams report after running DevZero on their clusters.

Databahn

About 75%

cost reduction on an AWS cluster

  • 29-36 manual node pools consolidated into 1-2 automated pools per cluster
  • Core provisioning reduced 40-50%, from 1,000s of cores to 500-600
“We thought about going with open-source Karpenter, but I wasn't comfortable with the production outage risk. DevZero's Karpenter-based approach felt familiar, but with the safety and multi-cloud support we needed.”
Yash Jagdale, Founding Engineer, Databahn
Read the Databahn case study

Fi Money

Up to 67%

lower Kubernetes costs on non-production clusters

  • 89% less overprovisioning
“Achieving 67% cost reductions on non-production clusters alone, before we'd even touched production, made the ROI case impossible to argue with.”
Prasanna Ranganathan, Founding Engineer, Fi Money
Read the Fi Money case study

Personality Pool

60%

reduction in total AWS infrastructure spend within 30 days of deployment

  • Daily spend dropped 30% within 24 hours of applying the first recommendations
“We started applying DevZero's recommendations on day 5, and within 24 hours our daily spend dropped by 30%. By day 30, we hit 60% total savings. That's faster ROI than any other infrastructure investment we've made.”
Lauren Glass Mullins, CEO, Personality Pool
Read the Personality Pool case study

Last updated

Frequently Asked Questions

Questions from platform teams evaluating cluster autoscaling.

What is Kubernetes cluster autoscaling?

It adds nodes when pods cannot be scheduled and removes nodes that are empty or underutilized.

How is DevZero different from the cluster autoscaler?

Cluster Autoscaler resizes node groups you define in advance. DevZero's Karpenter build picks instance types within the limits of a node policy, and DevZero adds suggestions, simulation and scheduling visibility on top.

Do I have to replace Cluster Autoscaler?

No. The Scheduling Health view works on clusters running Cluster Autoscaler. When you choose to move, a guided migration shows the proposed node policies before anything is applied, and switching Cluster Autoscaler off is reversible.

Can I move from open-source Karpenter?

Yes. A guided migration turns your NodePools and NodeClasses into node policies that start disabled, and your NodePools keep their names. The old Karpenter controller is scaled to zero, not uninstalled.

Which clouds are supported?

AWS, Azure and Google Cloud. One node policy compiles for each cloud, and node group simulation covers AWS and Google Cloud today.

Do suggestions change my cluster automatically?

No. A person reviews and accepts each suggestion before it changes a node policy. For NodePools that your own Karpenter manages, you apply the change in your own manifests.

Does consolidation disrupt running pods?

Consolidation drains a node, so its pods are evicted and rescheduled. Karpenter evicts through the Kubernetes eviction API, which respects PodDisruptionBudgets, and node policy disruption budgets cap how many nodes go at once.

See It on Your Clusters.Book a Demo.

Walk through node policies, suggestions and scheduling health with the DevZero team.

Book a demoGet started