I am measuring read-write performance of large blocks (from 1-20 MB) in
SUSE Linux and have same results no matter
how many outstanding IOs are issued (from 1-40).

I noticed in the mailing list archive that there was previous discussion
about this. However, I couldn't find the final outcome. What is the
current state of the code with regards to varying # of outstanding IOs in Linux? Also, what about # of Workers?