Performance and operational optimizations

AL2023 uses the latest drivers for Elastic Network Adapter (ENA) and Elastic Fabric Adapter (EFA) devices. AL2023 focuses on performance and functionality backports for hardware in Amazon EC2 infrastructure.

Kernel live patching is available for the x86_64 and aarch64 instance types. This reduces the need for frequent reboots.

All kernel build and runtime configurations include many of the same performance and operational optimizations of AL2.

AL2023 packages are built with compiler optimizations (-O2) enabled by default

AL2023 packages are built requiring x86-64v2 for x86-64 systems (-march=x86-64-v2), and Graviton 2 or higher for aarch64 (-march=armv8.2-a+crypto -mtune=neoverse-n1).

AL2023 packages are built with auto-vectorization enabled (-ftree-vectorize).

AL2023 packages are built with Link Time Optimization (LTO) enabled.

AL2023 uses the updated versions of Rust, Clang/LLVM, and Go.

Select backports to major system components include several performance improvements for running on Amazon EC2 infrastructure, especially Graviton instances.

AL2023 is integrated with several AWS services and features. This includes the AWS CLI, SSM Agent, Amazon Kinesis Agent, and CloudFormation.

AL2023 uses Amazon Corretto as the Java Development Kit (JDK).

AL2023 provides database engines and programming language runtime updates to newer versions as they're released by upstream projects. Programming language runtimes with new versions are added when they're released.

The base AL2023 AMI and container images are frequently updated to support patching instance replacement.

Kernel updates are included in AL2023 AMI updates. This means that you don't need to use commands such as yum update and reboot to update your kernel.

In addition to the standard AL2023 AMI, a minimal AMI and container image is also available. Choose the minimal AMI to run an environment with the minimal number of packages that's required to run your service.

By default, AL2023 AMIs and containers are locked to a specific version of the package repositories. There's no auto-update when they're launched. This means that you're always in control of when you ingest any package update. You can always test in a beta/gamma environment before rolling out to production. If there's a problem, you can use the pre-validated rollback path.