Store the Docker images in ECR and enable the local Docker layer cache in CodeBuild

Why the Answer Is Correct

The requirement is to improve build performance and minimize cost when producing large Docker images used across multiple builds. Two mechanisms together address it. Storing the images in an Amazon Elastic Container Registry repository makes them centrally available so a later build can pull an existing image instead of rebuilding everything from source, and this repository is where the artifacts persist across builds (A). Enabling the local Docker layer cache for CodeBuild makes the Docker daemon reuse the layers it has already built rather than pulling and rebuilding them, which is what removes the redundant work from each build; as DKM noted, this is the Docker local layer cache feature and it is what speeds up the build process by reusing layers (A). Together the repository and the cache are what reduce both build time and cost. A is the correct answer.

Why the Other Options Are Wrong

B caches the Docker images in an Amazon S3 bucket available across build hosts and expires the cache with an S3 Lifecycle policy. Docker's layer cache lives on the local filesystem of the build environment; storing image data in an arbitrary S3 bucket does not make Docker reuse those layers, so there is no build-time performance benefit. C stores the images in ECR but modifies the CodeBuild project's runtime configuration to always use the most recent image version. Pulling the most recent version on every build is the opposite of caching, since it guarantees nothing is reused and therefore loses the performance improvement the question asks for. D creates custom AMIs that contain the cached Docker images and launches EC2 instances from those AMIs in the build. This requires building and maintaining a custom AMI for each image change, which introduces AMI build time and lifecycle management and is substantially more expensive and operationally heavy than a build-time layer cache. A is correct.

Community Comment Notes

Community voted A unanimously. DKM explained that the local Docker layer cache stores build cache layers on the local filesystem and can significantly speed up the build process by reusing layers that have already been built, which is the performance mechanism. teo2157 described both halves, pushing built images to ECR after a successful build and pulling them at the start of each build to use as a cache. Ky_24 noted that ECR provides centralized storage so images can be maintained and reused across builds, reducing the need to rebuild base images repeatedly. No alternative received support.