We use Amazon S3 because it provides a reliable, fast and relatively cheap storage system for storing archived application data and CDN-optimized resources. The Glacier storage classes also allow us to store data for longer periods of time (e.g. logs) with lower costs.