From ‘Dataslows’ to Dataflows: The Gen2 Performance Revolution in Microsoft Fabric

Microsoft Fabric's Dataflows Gen2 has received significant performance and cost optimization enhancements for ETL workloads. These updates include a revised pricing model that reduces compute unit (CU) costs for longer jobs, a new "Modern Evaluator" query engine, and a "Partitioned Compute" feature for parallel processing. The author conducts a benchmark test processing 29 million records to compare the performance of the old Dataflow Gen1 against three configurations of Dataflow Gen2. The results show that enabling the Modern Evaluator provides the most substantial improvement, achieving a 30% time savings compared to the old dataflow and a 20% savings over the standard Dataflow Gen2, which also translates to significant cost reductions.