Skip to content

aligner_benchmarks.md

Latest commit

 

History

History
executable file
·
67 lines (40 loc) · 1.61 KB

aligner_benchmarks.md

File metadata and controls

executable file
·
67 lines (40 loc) · 1.61 KB

Comparison of BWA-mem2 ert and Sentieon BWA + Sorting

Optimizations Applied

jemalloc and numa!

Use them. Enable Huge Pages. Increase performance substantially.

Pipe to Sort

Avoid hitting disk to minimize/eliminate expensive multi tmp file bam sort.

Hardware

AWS EC2 r6* family 128vcp/1024GB memory instances. (choosing least expensive spot price from pool of 8 instance types meeting these criteria).

Results

Sentieon BWA + Sort

extremely fast re-implementation of bwa.

  • command:


30x genome, speed to map to b37:



30x genome, speed to map to hg38:



BWA-mem2 + Sort




re-compiled using intel compiler tuned to the target EC2 instances to be run on.




    

  • command:
    • 





30x genome, speed to map to b37:



30x genome, speed to map to hg38:



BWA-mem2 ert + Sort




fork of bwa-mem2, adding an Enumerated Radix Tree for an additional speed increase.
re-compiled using intel compiler tuned to the target EC2 instances to be run on.




    

  • command:
    • 





30x genome, speed to map to b37:



30x genome, speed to map to hg38:



HiSat2 (coming)