There are two reasons in theory. I'll explain both for generality, but
in your case it really is the second reason:
1) bigger numbers don't mean faster disks, it means more parallel
disks. That number represents how many requests the disk can handle in
parallel. There are cases in which the disk can have very little
parallelism, but each request is very fast so it can sustain a very
high IOPS with a very low number.
2) it timed out. scylla_io_setup uses a statistical process, and if it
times out before it can complete the evaluation, then the results will
be given with the data points it currently has. The results are
refined in many phases, and the early phases are really innacurate.
>> > email to
scylladb-user...@googlegroups.com.
>> > To post to this group, send email to
scyllad...@googlegroups.com.
>> email to
scylladb-user...@googlegroups.com.
>> To post to this group, send email to
scyllad...@googlegroups.com.
> --
> You received this message because you are subscribed to the Google Groups
> "ScyllaDB users" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to
scylladb-user...@googlegroups.com.
> To post to this group, send email to
scyllad...@googlegroups.com.
>
https://groups.google.com/d/msgid/scylladb-users/CANiF8TSciYJsC79UzLeB75wbRf2hYSVknPT3Lrw3RfjZgoXb5A%40mail.gmail.com.