Skip to main content

Before You Start

Before opening the canvas, be clear on four things. Having concrete answers allows you to set meaningful simulation parameters and evaluate recommendations objectively.

Traffic profile​

What is your expected baseline RPS? What does a peak spike look like — 2×? 10×?

For a Netflix-style recommendation API, this might be 10,000 requests per second at baseline with 10× spikes during content launch events. For an internal admin tool, it might be 50 RPS constant.

Latency budget​

What are your p50 and p99 targets end to end? A customer-facing checkout API might target p99 under 250ms. A background data processing pipeline might tolerate multiple seconds.

Cost envelope​

Do you have a monthly infrastructure budget? PinPole's live cost estimate makes this concrete during simulation — at 1,000 RPS, a typical CloudFront + API Gateway + Lambda + DynamoDB stack runs approximately $16,000–$17,000/month. At 10,000 RPS with provisioned concurrency and multi-region components, costs scale proportionally.

Resilience requirements​

What is your acceptable failure mode? Active-active, active-passive, or best-effort? This determines whether you need multi-region architecture, circuit breakers, dead-letter queues, and failover routing from the start.

If you are optimising an existing AWS environment​

  • Note the services currently in use and their approximate configuration (Lambda memory, RDS instance type, concurrency limits, etc.)
  • Identify known pain points — cold start latency, throttling events, cost surprises — that the simulation should surface
  • If the AWS Digital Twin feature is available for your account (upcoming — see Upcoming Features), use it to import the live topology directly rather than recreating it manually