Batch processing
Processing large volumes of data in bounded runs rather than event by event. The oldest data processing model there is — punch card jobs were batch — and still the backbone of most data platforms.
The archetype is the nightly job: the day's transactions are read, transformed and written to a data warehouse. The advantages are tangible — high throughput because overhead is amortized across millions of rows, simple error handling because a failed job is simply rerun, and deterministic results you can reproduce. The price is latency: the answer is only as fresh as the last run. The tools are MapReduce, Spark, dbt and orchestrators like Airflow. The opposite is stream processing with Kafka and Flink. The Lambda architecture combined both, while modern practice more often runs micro-batches until someone actually proves that seconds matter.