flexFS-Backed Nextflow Framework for Scalable, Multimodal Bioinformatics Pipeline Execution
C. Desjardins, S. Moore, S. Sarangi, G. Planthaber. Poster presented at Bio IT World 2026 May 19-21; Boston, MA
Today’s AI applications are fundamentally different from traditional software. They’re data-hungry, parallel by design, and performance-sensitive in ways that expose every weakness in conventional storage systems. Enter flexFS, the high performance file system for AI.
flexFS throughput scales indefinitely as cluster size grows, but both EFS and FSx for Lustre hit a performance ceiling. FSx for Lustre throughput can be increased, but only by increasing capacity – and cost.
Scale your science, not your spend. We benchmarked cryo-EM analysis using flexFS against FSx for Lustre and EFS. Results show that Lustre costs 3.5x flexFS and is 21% slower for end-to-end performance. EFS costs at least 1.75x flexFS and is 29% slower for end-to-end performance. We also found that flexFS is faster at each step […]