AWS re:Invent 2025 - Lambda Managed Instances: EC2 Power with Serverless Simplicity (CNS382)

AWS re:Invent 2025 - Lambda Managed Instances: EC2 Power with Serverless Simplicity

Overview

• AWS has pioneered serverless computing with AWS Lambda, continuously innovating and expanding the capabilities over the past 10 years.
• Customers have widely adopted serverless to run modern applications, benefiting from the programming model, scalability, and cost-efficiency.
• However, some use cases still require more control over the underlying compute infrastructure, leading to architectural shifts and increased operational overhead.
• Lambda Managed Instances (LMI) is a new solution that provides the same serverless programming model with more control over the compute resources.

Lambda Managed Instances (LMI)

• LMI allows you to run AWS Lambda functions on EC2 instances of your choice within your own AWS account.
• Key features:
  • Access to over 400 different EC2 instance types across general-purpose, compute-optimized, and memory-optimized families
  • AWS handles the lifecycle management, OS patching, and other operational tasks for the EC2 instances
  • Ability to apply EC2 pricing constructs like Savings Plans, Reserved Instances, and custom agreements
  • Support for multi-concurrency invocations on the EC2 instances

Using LMI

1. Create a Capacity Provider:
   • Defines the VPC configuration, allowed instance types, and scaling settings for the EC2 instances
   • Provides an IAM role for Lambda to manage the instances in your account
2. Create a Function:
   • Associate the function with the Capacity Provider to designate it as an LMI function
   • Supports the same packaging formats, runtimes, and features as Lambda default
   • Configure the function memory size and CPU ratio
3. Publish a Function Version:
   • Triggers the deployment of the function on the EC2 instances within the Capacity Provider
   • Lambda will launch the instances, initialize the execution environments, and make the function available for invocation

Differences from Lambda Default

1. Concurrency:
   • LMI supports multi-concurrency, allowing multiple invocations to execute simultaneously within a single execution environment
   • Requires consideration of thread safety best practices
2. Scaling:
   • LMI uses asynchronous, resource-based scaling, monitoring CPU and memory utilization to scale instances and execution environments
   • No cold starts, as execution environments are pre-initialized
3. Security Boundary:
   • LMI functions run in your own AWS account, providing a single-tenant environment
   • VM-level isolation can be achieved by using separate Capacity Providers

Partner Integrations and Tooling

• DataDog provides observability and monitoring for LMI instances
• Sedimentation's Autonomous Optimization platform helps migrate and optimize LMI deployments
• AWS AppConfig, CloudWatch Lambda Insights, and AWS Serverless Application Model (SAM) support LMI
• Pricing includes EC2 instance costs plus a 15% management fee

Key Takeaways

• LMI is not a replacement for Lambda default, but rather a complementary offering for specific use cases:
  • High-traffic, steady-state workloads with predictable traffic patterns
  • Applications with specialized compute, memory, or network requirements
• LMI provides the same serverless programming model with more control over the underlying compute infrastructure and the ability to leverage EC2 pricing constructs.
• The multi-concurrency, asynchronous scaling, and single-tenant security model of LMI differentiate it from the Lambda default experience.
• LMI is supported by a growing ecosystem of observability, migration, and automation tools, making it a viable option for enterprises looking to optimize their serverless deployments.