Spark utilizes a distributed computing model called RDD (Resilient Distributed Datasets), which allows data to be divided into partitions and processed in parallel across a cluster of machines. RDDs are immutable and fault-tolerant, enabling Spark to handle failures and automatically recreate lost partitions.

At a high level, Spark works by creating a directed acyclic graph (DAG) of transformations on RDDs. When an action is called, Spark optimizes and schedules the DAG execution by breaking it into stages and tasks. Each task is assigned to an available executor, which can run in parallel across multiple machines.