Kubernetes Auto-Scaling

Status: Coming soon

Kubernetes Auto-Scaling will optimize pod-level resource allocation and horizontal scaling parameters across your Kubernetes clusters.

What it will do

Most Kubernetes clusters run with resource requests and limits set once at deploy time and never revisited. This leads to significant over-provisioning: pods requesting far more CPU and memory than they actually use, and horizontal scaling thresholds that don't reflect real demand patterns.

Pump's Kubernetes Auto-Scaling will address this by:

1. Analyzing pod utilization. Pump collects CPU and memory usage histograms per pod over time, similar to the approach used by the Kubernetes Vertical Pod Autoscaler (VPA).

2. Computing optimal sizing. Based on observed usage patterns, Pump computes a "desired state" for each pod's resource requests and limits.

3. Gradual right-sizing. Rather than applying changes all at once, Pump uses a staged approach to reduce resource allocations incrementally, reducing the risk of performance impact.

4. HPA/KEDA recalibration. Pump evaluates your horizontal scaling thresholds and recommends adjustments to better match actual demand curves.

How it relates to other Pump Save mechanisms

Kubernetes Auto-Scaling targets a different layer of waste than the Commitment Planner or Infrastructure Recommendations:

Mechanism	Layer
Commitment Planner	Pricing: pay less per unit of compute
Infrastructure Recommendations	VM sizing: run smaller or fewer VMs
Kubernetes Auto-Scaling	Pod sizing: use less compute per pod within your existing VMs

These stack additively. Smaller pods mean fewer required nodes, which means your commitment coverage stretches further.

When it will be available

Kubernetes Auto-Scaling is currently in limited availability. If you're interested in early access, contact your account team or reach out to [e-mail].