Course contentsShow
← Cloud and Infrastructure
Lesson 479 of 2,9899. Serverless ComputeFree lesson

Reserved and Provisioned Concurrency

Pre-warming function instances to eliminate cold starts and guarantee capacity for critical workloads.

Reserved and Provisioned Concurrency

What you'll learn: How to eliminate cold starts and guarantee capacity for critical serverless functions by pre-warming instances.

The Cold Start Problem

You've already learned that serverless functions can experience cold starts when no warm instance is available. For latency-sensitive applications (like user-facing APIs), even a few seconds of delay is unacceptable. You've also learned about Function Concurrency Limits, which cap how many instances can run simultaneously. But what if you need guaranteed capacity with zero cold starts?

Pre-Warming Your Functions

Reserved Concurrency allocates a dedicated portion of your account's concurrency limit to a specific function. Think of it like reserving seats at a restaurant — those seats are yours, but they're not pre-occupied until guests arrive. This prevents other functions from consuming all available capacity but doesn't eliminate cold starts.

Provisioned Concurrency goes further: it keeps a specified number of function instances initialized and ready at all times. These instances are pre-warmed, have already run initialization code, and can respond immediately. It's like having a car idling in your driveway, engine running, ready to drive instantly rather than waiting for it to start.

When to Use It

Use provisioned concurrency when:

  • Your function handles user-facing requests requiring sub-100ms latency
  • You have predictable traffic patterns (scheduled workloads)
  • Cold start duration significantly impacts user experience

Trade-offs: Provisioned concurrency costs more because you're paying for idle capacity, not just execution time. Use it strategically for critical paths only.

Key Takeaway: Reserved concurrency guarantees capacity allocation, while provisioned concurrency pre-warms instances to eliminate cold starts entirely — ensuring predictable, low-latency performance for critical serverless workloads.