tdg5/blog

Jekyll-ified blog

blog's Issues

Unexpected operator

Running the script runs with the following error:

sup2069@debian-pc:~$ sudo sh scripts/dd.obs.test.sh /dev/null
scripts/dd.obs.test.sh: 17: [: -ne: unexpected operator
block size : transfer rate
scripts/dd.obs.test.sh: 37: [: -eq: unexpected operator

When omitting the cache clear (using your website script from 2015, it works). http://blog.tdg5.com/tuning-dd-block-size/

Presently shows optimal bs for the virtual mem "device"

First, thanks for your paper on this; I liked reading it! Actually 'been some time since it was on my TODO list; which happened today --viva summer time :-)

It seems both of your tests are letting the kernel page cache or virtual memory subsystem takes the lead. It's kind of clearer on the "find out optimal input block size", so I'll report here. Apologies for doing so before I could dig in your scripts.

1. On a fast USB stick (ok fast but)

$ ./dd_ibs_test.sh /media/SDX64_2/dd_ibs_testfile                                                                                                                                                                                    
Generating test file...
block size : transfer rate
     512 : 894 MB/s
    1024 : 1 GB/s
    2048 : 4 GB/s
    4096 : 6 GB/s
    8192 : 5 GB/s
   16384 : 0 GB/s
   32768 : 3 GB/s
   65536 : 2 GB/s
  131072 : 3 GB/s
  262144 : 1 GB/s
  524288 : 1 GB/s
 1048576 : 9 GB/s
 2097152 : 9 GB/s
 4194304 : 7 GB/s
 8388608 : 4 GB/s
16777216 : 0 GB/s
33554432 : 3 GB/s
67108864 : 9 GB/s

2. On SSD

$ sh $HOME/.local/scripts/dd_ibs_test.sh ~/dd_ibs_testfile                                                                                                                                                                                    
Generating test file...
block size : transfer rate
     512 : 896 MB/s
    1024 : 7 GB/s
    2048 : 0 GB/s
    4096 : 9 GB/s
    8192 : 5 GB/s
   16384 : 9 GB/s
   32768 : 4 GB/s
   65536 : 6 GB/s
  131072 : 8 GB/s
  262144 : 5 GB/s
  524288 : 9 GB/s
 1048576 : 7 GB/s
 2097152 : 8 GB/s
 4194304 : 7 GB/s
 8388608 : 1 GB/s
16777216 : 8 GB/s
33554432 : 5 GB/s
67108864 : 8 GB/s

Test system: Ivy 3.3 GHz with Arch Linux.
Linux llewellyn 4.7.2-1-ck #1 SMP PREEMPT Sat Aug 20 14:07:40 EDT 2016 x86_64 GNU/Linux

TRANSFER_RATE may be incorrect with different locale in dd test scripts

If the used locale has e.g. comma (,) as decimal separator, the script does not show correct value. Better to run the dd command as LC_NUMERIC=C dd ...

Problems with scripts/dd_ibs_test.sh and scripts/dd_obs_test.sh

First, thanks for the handy scripts! Needed to tweak them to make them work though...

Testing this with bash 4.3 on Raspbian, the scripts silently exit without running the tests.

Seems the following line is the problem:
[ -e "$TEST_FILE" ]; TEST_FILE_EXISTS=$?

This test appears to be entirely redundant (and for some reason it causes an early exit as well).

With that, the last line would also simplify from
[ $TEST_FILE_EXISTS -ne 0 ] && rm $TEST_FILE
to
rm $TEST_FILE

Now, I wonder what the purpose of that problematic line above was, exactly? It seems like if a file exists 'ibs' will early exit anyway (with a message), and 'obs' doesn't even have the (working) existence check seen in 'ibs'. Thoughts?