How can you improve the availability of a stateless web-based API deployed on a single Compute Engine instance in the europe-west2-a zone when the current Service Level Indicator (SLI) for availability is below the target Service Level Objective (SLO)? The API experiences frequent request timeouts due to high traffic and memory exhaustion.