12 cpus_allowed_policy=split
14 # For the pmemblk engine:
16 # IOs always complete immediately
17 # IOs are always direct
25 # Unlink can be used to remove the files when done, but if you are
26 # using serial runs with stonewall, and you want the files to be created
27 # only once and unlinked only at the very end, then put the unlink=1
28 # in the last group. This is the method demonstrated here.
30 # Note that if you have a read-only group and if the files will be
31 # newly created, then all of the data will read back as zero and the
32 # read will be optimized, yielding performance that is different from
33 # that of reading non-zero blocks (or unoptimized zero blocks).
37 # The pmemblk engine does IO to files in a DAX-mounted filesystem.
38 # The filesystem should be created on an NVDIMM (e.g /dev/pmem0)
39 # and then mounted with the '-o dax' option. Note that the engine
40 # accesses the underlying NVDIMM directly, bypassing the kernel block
41 # layer, so the usual filesystem/disk performance monitoring tools such
42 # as iostat will not provide useful data.
44 # Here we specify a test file on each of two NVDIMMs. The first
45 # number after the file name is the block size in bytes (4096 bytes
46 # in this example). The second number is the size of the file to
47 # create in MiB (1 GiB in this example); note that the actual usable
48 # space available to fio will be less than this as libpmemblk requires
49 # some space for metadata.
51 # Currently, the minimum block size is 512 bytes and the minimum file
52 # size is about 17 MiB (these are libpmemblk requirements).
54 # While both files in this example have the same block size and file
55 # size, this is not required.
57 filename=/pmem0/fio-test,4096,1024
58 filename=/pmem1/fio-test,4096,1024
68 # We're done, so unlink the file: