Create a file system with a small AG size:

# mkfs.xfs -d agsize=64m /dev/sdb7 > /dev/null
# mount /dev/sdb7 /test
# chmod 777 /test

Create ten 10MB files concurrently in two directories:

# cd /test
# mkdir a b
# for dir in a b; do
> for file in `seq 0 9`; do
> xfs_mkfile 10m $dir/$file
> done & 
> done; wait 2>/dev/null

Note that the all the inodes in the same directory are in the same AG:

# ls -sid * */*

What about the file data? Use xfs_bmap -v to examine the extents of each of the files:

# xfs_bmap –v a/*
# xfs_bmap –v b/*

Note that once the original AG was used, the files from both directories are interleaved in the next allocation group. Any read-ahead for files in directory a that assumes the files are contiguous on disk will start reading files from directory b instead.