Fractal - Distributed File System

Fractal is a distributed file system designed for large-scale, data-intensive workloads, drawing architectural inspiration from systems like the Hadoop Distributed File System (HDFS) and Google File System (GFS). Fractal adopts a CP (Consistency and Partition Tolerance) model, deliberately prioritizing data integrity and strong consistency over availability by rejecting operations whenever the required quorum or replica counts cannot be safely satisfied. The entire ecosystem is operated via a standalone CLI that exposes commands for file management (create, read, update, and burn) alongside diagnostic utilities such as list and status to monitor cluster health. 

A centralized NameNode maintains the cluster namespace and replica locations, persisting its state to a lightweight filesystem image, while raw data blocks are distributed across the physical disks of worker DataNodes. Rather than routing payload bytes through the master, Fractal utilizes a distributed pull model: clients query the NameNode solely for a chunk location blueprint, after which all heavy chunk streaming occurs concurrently and directly with the DataNodes to maximize bandwidth utilization. 

Incoming files are split into fixed-size chunks and distributed using a rack-aware allocation strategy that guarantees multi-rack fault isolation. The system operates under an immutable storage paradigm where file updates generate a timestamped replica set before executing an atomic metadata swap, rather than modifying bytes in place. Continuous heartbeat exchanges and quorum consensus protocols allow Fractal to safely coordinate multi-client concurrency, actively detect dead workers, and transparently repair stale or missing replicas on routine failure.