Scale

Autoscaling

Add and remove instances automatically as traffic, CPU, and memory change.

Start for free Talk to sales

web-frontend

2 instances · 1 CPU
6 instances · 2 CPU
18 instances · 4 CPU
2 instances · 1 CPU

Capacity that follows demand

Set the limits once. The platform handles every spike and quiet hour in between.

CPU and memory targets
Scale on average CPU, memory, or both, with thresholds you choose per service.
Min and max instances
Keep a warm floor for fast responses and a ceiling that protects your budget.
Scale-down cooldowns
Avoid flapping with cooldown windows that wait for traffic to settle.
Load-balanced by default
New instances join the pool as soon as they pass health checks.

How autoscaling works

Three checks run every few seconds, so capacity is always one step ahead of traffic.

Min 2
Max 20
CPU 70%
Memory 80%

Configuration

Autoscaling in a few lines

Turn it on in the dashboard, or keep it in your config file next to the rest of your stack.

Per-service limits and targets
Change limits without a redeploy
Same config for staging and production

          
            
                1
                services:
              
                2
                  - type: web
              
                3
                    name: web-frontend
              
                4
                    runtime: node
              
                5
                    autoscaling:
              
                6
                      minInstances: 2
              
                7
                      maxInstances: 20
              
                8
                      targetCPUPercent: 70
              
                9
                      targetMemoryPercent: 80

          
            
                1
                # Update limits for a running service
              
                2
                $ cloud scale web-frontend --min 2 --max 20 --cpu 70
              
                3
                ✓ Autoscaling updated for web-frontend

When to use autoscaling

Autoscaling fits any service with traffic that changes over the day: storefronts, public APIs, webhooks, and AI endpoints that see bursts after a launch or a campaign. You pay for the instances you run, so a quiet night costs less than a busy afternoon.

What scales and what stays fixed

Web services and private services scale horizontally. Each new instance runs the same build and joins the same private network, so there is nothing to reconfigure. Background workers can scale on queue depth, and databases scale separately with their own plans.

Horizontal scaling for web and private services
Queue-based scaling for background workers
Independent limits for every environment

Keeping costs predictable

Your maximum instance count is a hard ceiling. When a service reaches it, you get an alert and the service keeps running at that size. Usage is billed per second, and the usage page shows exactly when each instance started and stopped.

Autoscaling

Capacity that follows demand

CPU and memory targets

Min and max instances

Scale-down cooldowns

Load-balanced by default

How autoscaling works

Autoscaling in a few lines

When to use autoscaling

What scales and what stays fixed

Keeping costs predictable

Works well with

Zero-downtime deploys

Private networking

Infrastructure as code

Push code today. Be live before your coffee cools.