While many distributed systems can operate in standalone or single-node mode, they are inherently designed for distributed processing and they carry the architectural overhead. Even in Spark's Standalone mode you need to run both Spark master and worker processes on a single machine.