Can anyone suggest a good way to measure USB drive speed in a repeatable and effective way?

I know that what I "copy" and how I "copy" and the filesystem in use make a difference in anything I try to measure.

I know that I can:

• choose some file or folder tree
• use various tools (cp, tar, rsync, etc) to move the data
• use time command to collect details
• do this more than once

One "benchmark" I've used in the past was to create a source tree for some application and use make clean; time make in that tree. Lots of I/O of various sorts and some computation and very repeatable.

Since I'm not seeking a gold standard benchmark, I can do what I like, but I'd like to do something (a)meaningful in general for myself and others, and (b)avoid inventing what probably already exixts. I hope someone, here, can suggest an existing tool kit.

Cheers,
~~~ 0;-Dan

Hi SaintDanBert,

Have you heard of bonnie++ at all? ( http://linux.die.net/man/8/bonnie++ )
I the tool is designed for benchmarking hard drives. I have personally used it for placing load on one in the past and it will print out statistics in the end. One thing that caught me out is that with default options, it will create an 8GB file in the current working directory. You may need to manually specify the size to create and number of files.
Hope that helps.

Kind Regard,
Netsurf

I've use dd in the past:
copy from /dev/zero to a file on my disk to measure disk write speed
copy from a file on my disk to /dev/null to measure disk read speed

NB, you need to make the io size match what your applications would use. and you need to use a number of IOs big enough that Linux does not cache everything in ram.
Use bs and count options for this.

bonnie++ is on the fedora repository, so I'll try it

I have no idea about I/O sizes and caching. Any recommendations where to learn?
Are averages and other statistics kept somewhere by the various file systems that
I could read from /proc or /sys?

~~~ 0;-Dan