A talk given at the SF Hadoop User group on May 3 2016. The YARN shared cache is a feature that aims to reduce the overhead of resource localization without requiring application developers to coordinate manually. Leveraging the shared cache, applications can securely share resources localized by a different application or even a different user. In this talk, we will give an overview of the shared cache architecture and a solid foundation for developers who want to build YARN applications that leverage this feature. Additionally, we will discuss some of the benefits we have seen from using the shared cache at Twitter.