| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| |
|
|
|
|
|
|
|
|
|
|
|
|
| |
The test in recent versions only worked by mounting and unmounting
partitions where the test is to be run. This was done to correctly
measure the time taken to run the tests.
However, this meant the tests can't be run on systems that don't
have a spare partition (like my development machine). Adding
support for running the test in the current working dir here so
that the program can be tested. Of course, the results should
not be trusted.
Signed-off-by: Amit Shah <amit.shah@redhat.com>
|
| |
|
|
|
|
|
|
|
|
|
| |
A few runs that resulted in bad timings for chunk-sized writes led Jan
Kara and Mike Galbraith to discuss why it might be so and Mike spotted
the error in the code -- we allocated a buffer of the size of the file
instead of the size of the chunk -- in my case, 4G instead of 4K.
Thanks Mike.
Signed-off-by: Amit Shah <amit.shah@redhat.com>
|
| |
|
|
| |
Signed-off-by: Amit Shah <amit.shah@redhat.com>
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
The results from running the test via the run_tests.sh script are not
yet formatted the way this gnuplot script needs them to be, but when
formatted so, this script can be used to draw the bar graphs.
My knowledge on gnuplot is elementary; just about 1h worth of searching
and reading the demo files and this is a modified script I picked up
from the gnuplot website. More customisations are surely possible, like
piping the resulting image to image processors and dumping a png
directly instead of 'copy to clipboard'->paste in gimp->save.
The run_results.txt file that this script parses should look like this:
filesystem posix-fallocate mmap chunk-4096 chunk-8192 posix-fallocate mmap chunk-4096 chunk-8192
ext2 74 96 761 81 33 37 37 36
ext3-writeback 87 97 202 93 34 40 39 36
The first set of 4 columns contains seconds; the 2nd set contains fragments.
Signed-off-by: Amit Shah <amit.shah@redhat.com>
|
| |
|
|
|
|
|
| |
This script takes care of creating a new file system, adjusting various parameters
for the fs when needed for different cases and runs the test.
Signed-off-by: Amit Shah <amit.shah@redhat.com>
|
| |
|
|
| |
Signed-off-by: Amit Shah <amit.shah@redhat.com>
|
| |
|
|
|
|
|
|
| |
Flush out data in buffers / caches not belonging to our test. This data can
figure in our timings if the kernel decides to write them out during
our testing, so it's best to get it out beforehand.
Signed-off-by: Amit Shah <amit.shah@redhat.com>
|
| |
|
|
|
|
|
|
|
|
|
|
| |
* Support for mouting and unmounting file systems. This will
flush out data for sure -- so the times we collect will
reflect the time needed to actually write everything we want to
to the disk.
* Add pre-run and post-run functions that take care of the common
functionality needed by each test run
Signed-off-by: Amit Shah <amit.shah@redhat.com>
|
| |
|
|
| |
Signed-off-by: Amit Shah <amit.shah@redhat.com>
|
| |
|
|
| |
Signed-off-by: Amit Shah <amit.shah@redhat.com>
|
| |
|
|
| |
Signed-off-by: Amit Shah <amit.shah@redhat.com>
|
| |
|
|
| |
Signed-off-by: Amit Shah <amit.shah@redhat.com>
|
| |
|
|
| |
Signed-off-by: Amit Shah <amit.shah@redhat.com>
|
| |
|
|
| |
Signed-off-by: Amit Shah <amit.shah@redhat.com>
|
| |
|
|
| |
Signed-off-by: Amit Shah <amit.shah@redhat.com>
|
|
|
This program measures performance of creating a file and zeroing it using
a few different methods:
1. using the posix_fallocate() syscall
2. using mmap() and then memset()
3. writing chunks to the file (currently 4k and 8k-sized chunks)
directly calling the Linux fallocate() syscall doesn't yet work; a bug has
been filed earlier for this in the Red Hat Bugzilla, #485487
A few results from running this program on my laptop and on my desktop are put
in the results.txt file.
Signed-off-by: Amit Shah <amit.shah@redhat.com>
|
