Fabian Lehmann defended his dissertation, “Adaptive Scheduling of Dynamic Workflows” with distinction on June 29th, 2026. Fabian was a member of subproject B5 in FONDA Phase I. His work examines the tradeoff between workflow portability and efficiency, and introduces WONDERS, a best of both worlds optimization strategy for Nextflow workflows.
In order for scientific workflows to be portable, workflow management systems such as Nextflow cannot rely on hard-coded assumptions about the dataset or underlying infrastructure to improve efficiency. WONDERS utilizes the Common Workflow Scheduling Interface (CWSI), developed by Fabian, to bring workflow context to the resource manager, then combines his three novel approaches for workflow optimization: WOW, PONDER, and SCALE.
WOW is a scheduling approach which reduces network congestion within a cluster by basing scheduling decisions primarily on the co-location of tasks and data. PONDER and SCALE are tools which predict and minimize the ongoing use of memory (PONDER) and CPU (SCALE) over the course of a workflow execution. The combination of the three techniques in WONDERS results in super-additive improvements in workflow makespan (average reduction ~50% compared to nf-core default settings) when tested on a variety of Nextflow workflows using real-world data from remote sensing and bioinformatics.
Congratulations Fabian!
