Background: We are setting up the new storage for an HPC Compute cluster for applied statistics, bioinformatics, and genomics. Contrary to our expectations, an active-active configuration did not improve I/O performance compared to failover under various I/O loads.
Question: Under what scenarios would you fiddle with device-mapper multipath configuration e.g. set up an active-active configuration or other optimizations? Is best practice to use the default settings and avoid optimization unless strictly necessary, as with many other things in the world of software engineering and systems administration?