Documentation

Shared or Dedicated nodes.

Every environment picks its tenancy model when you create it. Learn the difference, the trade-offs, and when it's worth mixing both.

You're not locked in: change an environment's tenancy later, or run a mixture across your environments.

Shared nodes

Cheaper out of the box

Your build runs alongside other tenants on shared capacity, isolated by container boundaries. It's the default tenancy, scales to zero when idle, and needs no capacity planning. You configure an amount of compute for your workload — for example 2 CPU cores, 0.1 CPU cores, or a specified amount of RAM — and pay for the resources assigned to your running workloads. Stop a workload and billing for it stops immediately; bring it back up and billing resumes.

Good for
  • Marketing sites & landing pages
  • Web apps and admin panels
  • APIs with light or bursty traffic
  • Staging, preview & internal tools
Pay for what you use Scales to zero Fastest to launch
Dedicated nodes (Isolated Compute)

Built for scale

Lease a whole node in a fixed size, just for your build: no noisy neighbors, and performance that stays predictable as traffic grows. You pay for the node itself, not per-millisecond execution — while it's provisioned, you're billed for the whole node regardless of utilisation. Deprovision everything on the node and billing stops; provision it again and billing resumes.

Good for
  • Production APIs & microservices at scale
  • Databases & other stateful services
  • High-traffic web apps
  • Compliance-sensitive workloads
No noisy neighbors Predictable performance Instant self-service
Most teams start every environment on shared nodes, then move specific ones to dedicated as traffic or compliance needs grow. For example, shared nodes for staging and the marketing site, dedicated nodes for the production API and database.
Billing example: an organisation runs five workloads on shared nodes and stops one. Billing for that workload stops immediately, and the organisation continues paying only for the four still running — so a stopped workload's own infrastructure cost can reach $0, even though the organisation's total bill may still reflect other running workloads, seats, or plan costs.

See where compute fits in the full platform walkthrough →

Ready to Lyft?

Deploy your first build today.

Create your account, and pick shared or dedicated tenancy for every environment as you go.

Start deploying — free Talk to sales