shiichan

AWS Lambda Managed Instances gets SOCI lazy loading, cutting container startup by up to 70%

Hi everyone, it's Shii-chan! Today's news is about Lambda container startup getting up to 70% faster!

AWS What's New aws.amazon.com

What was announced?

In AWS What's New, AWS announced that AWS Lambda now supports lazy loading with Seekable OCI (SOCI). It applies to Event Functions packaged as OCI images running on Lambda Managed Instances.

The story so far

Compute-intensive workloads such as data processing pipelines, high-throughput API backends, and batch computation can have container images of several gigabytes. Downloading and preparing the full image could take several minutes when provisioning new execution environments.

What changes

With SOCI, Lambda fetches only the files needed at each stage of execution environment startup. Because the whole image isn't downloaded upfront, functions start faster as you scale, and AWS says startup times drop by up to 70%.

Dive Deep

Getting started takes three steps:

  • Create a SOCI index (a table of contents for the files in your image layers) for your container image
  • Push the indexed image to Amazon ECR
  • When creating or updating your function, set the image URI to the SOCI-indexed tag (for example, my-function:v1-soci) and publish a function version

Lambda automatically detects the SOCI index and enables lazy loading, with no additional configuration.

If you use Kiro, Claude Code, or Codex, you can install the Lambda Managed Instances skill from the Agent Toolkit for AWS to configure and deploy functions through your agent.

It is available in all AWS commercial Regions where Lambda Managed Instances is offered. See the Lambda Managed Instances developer guide for details.

Wrap-up

  • SOCI lazy loading is available for OCI-image functions on Lambda Managed Instances
  • Only the needed files are fetched, cutting startup time by up to 70%
  • Create a SOCI index and push it to ECR; no extra configuration is needed

This is for Lambda users running heavy workloads from multi-gigabyte images!