Im starting to use XFS because I want to create a highly parallel disk array and experimenting on just 2 drives for now.
I had thought that if I used a logical volume group to join two physical volumes and then put XFS over the top, XFS would take advantage of both drives by having each handle one of the requests, whereas raid 0 would force both drives to handle both requests. Unfortunately, I just tried this and iostat shows that all the data is being written to /dev/sdb
, whilst /dev/sdc
is doing nothing. Have I misunderstood the point of XFS? Perhaps the advantage of parallelization only shows as the disks become full and operations start to occur on files that are already on different drives?
For reference the two commands I ran in parallel were:
dd if=/dev/zero of=10000MB.img bs=1M count=10000 conv=fdatasync
dd if=/dev/zero of=10000MB2.img bs=1M count=10000 conv=fdatasync