Hello
Since everyone is talking about disk benchmark and the results on
different kernel version I would like to ask a question about this.
I noticed when using bonnie that I could sometimes get very different
results (> 20%) using the same kernel and disks only that the second
test (slower one) is a long time after the first one. Or if I run
bonnie on a freshly formated disks would always give me the best
results. I know that the disk is fastest at the beginning and the
more data you have filling up the disk the slower it will be. But
what about fragmentation? If I constantly create files of different
size and delete some of them will I not be getting a file system that
looks something like this:
xxxxx--------xxxx-xxxxx---xx-xxxxx-xxx----xxxxxxxxx-----xxx-------
where "x" is data and "-" is free. And if I start bonnie on such
a disk it will fill up the free spaces with say "O":
xxxxxOOOOOOOOxxxxOxxxxxOOOxxOxxxxxOxxxOO--xxxxxxxxx-----xxx-------
Will this not influence the performance a lot since the head of
the disk has to walk all over the disk? Thus making comparisons
practically useless since you never know the state of fragmentation?
Hope this is not to much off topic.
Holger
