AWS Lambda Managed Instances with Java 25 and AWS SAM - Part 3 Create Lambda function with LMI compute type

Source: Dev.to

Introduction

In part 1 of the series we explained the ideas behind AWS Lambda Managed Instances (LMI) and introduced our sample application.
In part 2 we described what a Lambda Capacity Provider is and how to create it with AWS SAM.

In this article we cover how to create Lambda functions and attach them to the already‑created Capacity Provider.

When creating a Lambda function with the LMI compute type via the AWS Lambda console, an Additional configuration section appears with two choices:

• Lambda (default) – the traditional compute type.
• Lambda managed instance – the new LMI option.

The compute type must be chosen up front and cannot be changed later because each type uses a different concurrency model.

We’ll use AWS SAM as infrastructure‑as‑code (IaC). The sample application is available in the GitHub repository aws‑lambda‑java‑25‑lmi. The relevant SAM template sections are shown below.

Global Function Settings

Globals:
  Function:
    Runtime: java25
    MemorySize: 2048
    CapacityProviderConfig:
      Arn: !Sub ${MyCapacityProvider.Arn}
      ExecutionEnvironmentMemoryGiBPerVCpu: 2
      PerExecutionEnvironmentMaxConcurrency: 32
    FunctionScalingConfig:
      MinExecutionEnvironments: 1
      MaxExecutionEnvironments: 3

• Runtime – java25 (the managed Java 25 runtime).
• MemorySize – 2048 MB (the current minimum supported RAM for LMI).
• CapacityProviderConfig – selecting the LMI compute type by referencing an existing capacity provider.
• FunctionScalingConfig – defines the scaling limits for the function.

Supported Runtimes for LMI

Additional runtimes will be added in the future.

CapacityProviderConfig Details

The PerExecutionEnvironmentMaxConcurrency setting distinguishes LMI from the default compute type:

• Default (Firecracker microVM) – one request per execution environment; memory can range from 128 MB to 10 GB.
• LMI – multiple concurrent requests per environment (e.g., 32) but requires at least 2048 MB of memory.

You can tune memory and concurrency to optimise price‑performance, monitoring memory consumption as the application scales.

FunctionScalingConfig Details

These settings control the scaling behaviour of a Lambda Managed Instances function. See the AWS documentation for more on Scaling Lambda Managed Instances.

When you run sam deploy -g, publishing a function version occurs automatically. The new version becomes active on the capacity provider instances once published.

Viewing the Configuration

After deployment you can inspect the capacity provider and function scaling settings in the Configuration tab of the Lambda console.

If you set both MinExecutionEnvironments and MaxExecutionEnvironments to 0, the function is effectively disabled, avoiding any runtime costs. With the values 1 and 3 (as in the IaC example), you’ll see the underlying EC2 instances on the Amazon EC2 service page. These instances are identified by the Operator property starting with scaler.lambda. Direct access (e.g., SSH) to these LMI instances is not permitted.

In the Capacity providers view, selecting your provider (e.g., CapacityProviderForJava25LMI) and opening the Function versions tab shows all Lambda functions and their versions linked to that provider.

Next Steps

In the next part of the series we will cover:

• Monitoring Lambda Managed Instances
• Currently unsupported features
• Ongoing challenges
• Pricing considerations

---