What is the name of Google's early distributed file system?

Google's early distributed file system was the Google File System, commonly abbreviated GFS.

GFS was designed for Google's enormous data-processing needs, especially search indexing and other workloads running across large clusters of inexpensive commodity computers. Instead of assuming that individual machines would always work, it treated hardware failures as normal and used replication to keep data available when servers or disks failed.

The system divided files into large fixed-size chunks, traditionally 64 megabytes, and stored those chunks on machines called chunkservers. A master coordinated the namespace and tracked chunk locations. This architecture favored high throughput and large sequential reads or appends over the low-latency, POSIX-style behavior expected from a conventional desktop file system.

GFS grew out of Google's earlier BigFiles work and was described publicly in a landmark 2003 paper. It was eventually replaced inside Google by Colossus, but its design influenced later distributed-storage systems. HDFS is similar in purpose, but it is Apache Hadoop's separate open-source file system.