In a project at DataStream, two senior engineers, Alex and Taylor, had differing approaches in redesigning the event processing pipeline to handle increased data volumes.
Alex advocated for a batch-processing solution using Apache Spark for cost efficiency and simpler implementation, while Taylor pushed for a real-time streaming approach with Apache Kafka and Flink to minimize latency.
To resolve the disagreement, a structured decision-making workshop was organized, with a comparison framework to assess objective criteria like performance, complexity, scalability, and cost projections.
The workshop led to a hybrid approach, using Kafka for ingestion, processing time-sensitive events with Flink, and handling the majority with scheduled Spark jobs.