Comment by Sami_Lehtinen
10 years ago
I've usually done intentional fragmentation by filling disk with small files and deleting those while growing a new file to claim those blocks. Works basically with every file system. Some examples where I've used such method.
People often claim that fragmentation doesn't affect SSD drives, but that's not true: http://www.sami-lehtinen.net/blog/ssd-file-fragmentation-myt...
This is slightly related. How contiguously growing files are allocated on different file systems: http://www.sami-lehtinen.net/blog/test-btrfs-ext4-ntfs-simpl...
On the other hand, your worst fragmented case still gets ~200MB/s with a random read, which is around the best case with a sequential read for most HDDs today. A 4k random read on a HDD will be 1-2MB/s at most.
Relatively speaking the SSD slows down to ~50% of its max speed with fragmentation, but the HDD will be down to around 1%. So fragmentation affects SSDs somewhat, but HDDs are affected much more severely.
(Which SSD was it, and how was the test setup? That's important for comparison purposes, as the layout of the filesystem blocks and how they correspond to the NAND eraseblocks/pages has a huge effect on what fragmentation will do.)