Over the past several years, Google has developed the successor to Google File System, Colossus, to support both Google products and Google Cloud services. Colossus is a distributed file system designed to support heterogeneous workloads ranging from transaction processing to analytics at hyperscale. In this post, Dean Hildebrand and Denis Serenyi provide an overview of the architecture of Colossus, which relies upon a scalable metadata service and stores file system metadata in BigTable. They also describe how Colossus powers Cloud Storage, including leveraging the Colossus control plane to manage a shared storage pool, thereby enabling disaggregation of resources. In addition, they discuss how Colossus leverages flash and disk storage to avoid wasteful overprovisioning and provides fault tolerance and transparent recovery while leveraging different storage hardware.