Product reference pages are available in English only.
Network Services

VM Auto Scaling

What Is It?

VM Auto Scaling automatically adjusts the number of VM instances based on demand. Define scaling policies based on metrics (CPU utilization, network traffic) to scale out during high traffic and scale in during low usage. The service integrates with the Managed Application Load Balancer (ALB), automatically registering and deregistering scaled instances with ALB target groups to distribute traffic; VM Auto Scaling has no equivalent automatic integration with the Managed Network Load Balancer (NLB).

Quick Facts

Aspect Details
Type Automatic VM instance scaling service
Scaling Direction Scale out (increase) and scale in (decrease)
Trigger Types Metric-based (CPU utilization, network traffic)
Integration Managed ALB, Managed NLB
Target Resources Compute Engine VM replicas in Auto Scaling Groups
Management DCD, Cloud API

What You Can Do

Metric-Based Scaling

Define scaling policies based on CPU utilization or network traffic metrics (incoming/outgoing bytes or packets). Automatically add instances when thresholds are exceeded and remove them when demand decreases.

Policy-Based Scaling Configuration

Configure metric-based scaling policies with customizable evaluation intervals and cooldown periods. Scale out before anticipated traffic spikes and scale in during known quiet periods.

Integration with Load Balancers

Automatically register and deregister new instances with Managed ALB target groups. Traffic distributes to scaled instances without manual configuration; no equivalent automatic registration exists for the Managed Network Load Balancer (NLB).

Capacity Planning

Define minimum and maximum instance counts to ensure baseline capacity while capping costs. Scaling actions always remain within these defined boundaries.

Auto Scaling Groups

Group Compute Engine VM replicas into Auto Scaling Groups with shared configuration. All instances in the group scale together and are managed uniformly according to the defined scaling policies.

Replica Management

Configure minimum and maximum replica counts; the service adds or removes replicas only when a configured scaling policy (metric-based threshold or manual action) is triggered. VM Auto Scaling does not independently detect and replace an individual failed or terminated replica outside of a triggered scaling action.

Best For

Scenario Why It Fits
Variable traffic applications Automatically handle traffic spikes and quiet periods
Cost optimization Scale down during low usage to reduce running instance costs
High availability Maintain minimum instance count for redundancy (VM Auto Scaling is currently in Early Access; IONOS recommends limiting production-critical high-availability use until General Availability)
Predictable patterns Pre-scale capacity through policy tuning
Event-driven workloads Scale out for promotions, launches, or seasonal peaks
Development/staging environments Scale to minimum of one instance during off-hours to minimize costs (scale-to-zero is not supported)

Key Considerations

Billing & Costs

  • Main billing: Billed for the running VM replica instances
  • Scaling costs: Pay only for instances while they run
  • No Auto Scaling service fee: No separate charge for scaling functionality

Limitations

  • Scaling is horizontal only: it adds and removes VM replicas and does not vertically resize a running VM. Replica configuration changes apply to new replicas, not to existing ones
  • Can optionally integrate with Managed ALB or NLB for traffic distribution (recommended but not required)
  • Minimum and maximum instance limits must be defined
  • Scaling actions have cooldown periods to prevent rapid oscillation
  • Metric-based scaling depends on monitoring data availability
  • Scale-to-zero is not supported

Management Options

  • Data Center Designer (DCD)
  • Cloud API
  • Configure Auto Scaling Groups
  • Define metric-based scaling policies (one metric policy per group)
  • Set minimum and maximum replica counts (the actual count is managed automatically by the service)