Comment 12 for bug 633686

Thank you for the numbers.

Here's what I can see (which "calibrates" my faster/slower expectations):

1) flash drives are ~2x slower than hard disks (see the "total" numbers)

2) its _NOT_ rpmdb syncing or redundancy that is to blame (see dbadd/dbget/dbput
entries. Your anomalous results are due to "warm cache" != "cold cache" and other
peripheral issues, the numbers are approximately equal, as they should be,
depending on where you have chosen to place the rpmdb).

3) installs (where most of the syscall issues are located) are ~2x slower.
OTOH, "uncompress" (which is likely summed into install, don't get tricked)
is more than 4x slower.

That is already useful info to direct attention to where a fix
might be attempted.