Big Data systems (Spark / Hadoop over S3) really pioneered this, as the original MapReduce paradigm had fairly tight coupling between compute and data.