Comment 7 for bug 633686

Noone disputes what you are seeing on a USB flash drive.

What isn't clear is how to fix the issue because the root cause
cannot be seen in a yum trace. All that can be confirmed
from your observations is that indeed, USB operations
are S-L-O-O-W, and noone is disputing that at all.

There are two "likely" causes of slow, each with different "cures":

1) redundant database operations sync'd to disk

2) file system operations.

What is needed is some way to guage which is the S-L-O-W-E-R
issue.

Something like a strace (which WILL be quite large and slow, run on
a single, not 40, package upgrade) is needed with timestamps on the
system calls. It will be possible to tell from context whether its
database reads/writes or file system operations that are the root cause.

Note that there's not much that can be done by RPM to "cure" 2) file system
operations. An RPM installer MUST read and write content. There is a chance
that stat(2) is the culprit, and there's a chance that some stat(2) operations
can be avoided.

But there's no way to tell until the S-L-O-O-W syscalls are identified in context.