
160 High Performance Parallel I/O
0
PP
2
file access phase
file system
P
0
’s file domain P
2
’s file domain
a 2D global array
subarrays in the
file offset view of
0
P P
1
start
P
2
end
P
3
aggregate access region
Two−phase I/O
I/O aggregators
data partitioning of
file space
0
P P
1 0
P P
1 0
P P
1
P
2
P
3
P
2
P
3
end
disp
start
request aggregation phase
FIGURE 13.3: A 5 × 8 2D array partitioned among 4 MPI processes in a
block–block pattern and its data layout in the file. The data layout in the file
follows the array’s canonical order. The bottom part describes the two-phase
I/O operation carried out in MPI collective I/O.
At each checkpoint, the intention considers writing the 2D array as