Data-driven optimization and AI-powered orchestration for Nextflow with Fovus on AWS

This article introduces Fovus, an AWS partner solution that optimizes Nextflow genomics pipelines through per-process benchmarking and AI-powered resource allocation on AWS.

• Static resource configs cause 70-85% cost overspend across heterogeneous pipeline steps
• Fovus benchmarks each process individually to determine optimal EC2 instance types and configurations
• Memory checkpointing enables Spot Instance use for long-running processes with automatic failover
• nf-core/rnaseq achieves $0.70 per sample cost on Spot Instances versus unoptimized deployments
• nf-core/sarek 30x WGS costs $6.15 on Spot Instances with near-identical walltime to on-demand
• Deploys within customer AWS account with no data leaving account boundary
• Free pilot workspace available for benchmarking and configuration generation

Fovus delivers significant cost savings and performance improvements for genomics workloads by replacing static configurations with data-driven, per-process optimization.