> ## Documentation Index
> Fetch the complete documentation index at: https://help.nops.io/llms.txt
> Use this file to discover all available pages before exploring further.

# Reduce Lambda Provisioned Concurrency

> Remove or reduce Lambda provisioned concurrency that is used too little to pay for itself.

Provisioned concurrency keeps Lambda environments warm so functions start without a cold start. You pay for it every hour it is configured, plus a lower rate for the time it is used. When it is used too little, the same invocations would cost less on demand. nOps compares what you pay with what the same work would cost without it.

| | |
| - | - |
| **Cloud** | AWS |
| **Category** | Compute |
| **Service** | AWS Lambda |
| **Action** | Modify |
| **Effort** | Low |
| **Restart needed** | No |
| **Can be rolled back** | Yes |
| **Analysis window** | Last 7 days |
| **Source** | nOps |

## Why it matters

Provisioned concurrency is often sized for a launch or a peak and left in place. Below a certain utilization it costs more than it saves.

## How nOps finds it

nOps reads the last 7 days of Lambda billing for each function's provisioned concurrency and the compute time run on it. The saving is the provisioned concurrency charge plus its compute charge, minus what that compute would cost at the on-demand rate, using list prices in the region with your discount carried through. A function is flagged when the saving is positive (its utilization is below the break-even point) and at least $5 a month, or at least $3.75 when the 30-day average saving is \$5 or more. Usage covered by Savings Plans is left out.

## Before you start

Permission to change provisioned concurrency (`lambda:PutProvisionedConcurrencyConfig`, `lambda:DeleteProvisionedConcurrencyConfig`).

## How to implement

1. Check whether the function is latency-sensitive and how often it sees cold starts without provisioned concurrency.
2. Look at its CloudWatch metric `ProvisionedConcurrencyUtilization` to pick a lower level, or decide to remove it.
3. Lower it (`aws lambda put-provisioned-concurrency-config --function-name NAME --qualifier ALIAS --provisioned-concurrent-executions N`) or remove it (`aws lambda delete-provisioned-concurrency-config --function-name NAME --qualifier ALIAS`).
4. If it is managed by Application Auto Scaling or a template, change it there instead.

## Risks and considerations

* Without provisioned concurrency, some requests see cold starts, which adds latency.
* A scheduled scale-up for known peaks can keep the benefit at lower cost.
* To roll back, set provisioned concurrency again on the alias or version.

## FAQs

> [accordion] What is the break-even point?
>
> The utilization at which provisioned concurrency costs the same as running the same work on demand. Below it, on demand is cheaper.

## Related

* [Switch Lambda Function to Arm64](https://help.nops.io/inform/supported-recommendations/switch-lambda-function-to-arm64)
* [Rightsize Lambda Function](https://help.nops.io/inform/supported-recommendations/rightsize-lambda-function)
* Move Steady High-Volume Lambda to Managed Instances
* [Supported recommendations](https://help.nops.io/inform/supported-recommendations.md)


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.