Fig. 5

Benchmark results. A Total wallclock duration of each pipeline. BiocMAP completes in roughly 1 day, not including downtime between the first and second module, while nf-core/methylseq completes in roughly 7 days. B Total CPU utilization by pipeline, in CPU hours. nf-core/methylseq uses more than 14 times the CPU resources as BiocMAP, at 1527 CPU hours. BiocMAP uses less than 8.7 GPU hours total (not shown) during alignment, suggesting the greater CPU usage of nf-core/methylseq is not explained by BiocMAP’s use of GPUs. C Maximal concurrent CPU usage at any point during pipeline execution. BiocMAP utilized 26 CPUs at its most demanding moment, while nf-core/methylseq used 72. Concurrent CPU usage is configurable, but was designed to be unconstrained during this benchmark. D Total memory usage by each pipeline in TB-hours. nf-core/methylseq used approximately 12.5 times more memory hours than BiocMAP over the full course of execution. E Maximal concurrent memory usage in GB across all processes, at any point during pipeline execution. BiocMAP used 341 GB of memory, approximately 4.7 times more than nf-core/methylseq, at each of their most memory-intensive moments. Note that concurrent memory usage is configurable, but was intentionally unconstrained for this benchmark