Exchange Performance Comparison

Exchange is another one of PMB's Parallel Transfer Benchmarks. Exchange is a communications pattern that often occurs in grid splitting
algorithms (boundary exchanges). The group of processes is seen as a periodic chain, and each process exchanges data with both left and right neighbor
in the chain.

The turnover count is 4 messages per sample (2 in, 2 out) for each process.

2-node Cluster Results
4-node Cluster Results
8-node Cluster Results
กก

2-node Latency (millisecond) Performance Results

#bytes  #reps   linux           dom0            domU(noSMP)     domU(SMP)
--------------------------------------------------------------------------

2-node Throughput (MB/s) Performance Results

#bytes  #reps   linux           dom0            domU(noSMP)     domU(SMP)
--------------------------------------------------------------------------

4-node Latency (millisecond) Performance Results

#bytes  #reps   linux           dom0            domU(noSMP)     domU(SMP)
--------------------------------------------------------------------------

4-node Throughput (MB/s) Performance Results

#bytes  #reps   linux           dom0            domU(noSMP)     domU(SMP)
--------------------------------------------------------------------------

8-node Latency (millisecond) Performance Results


#bytes  #reps   linux           dom0            domU(noSMP)     domU(SMP)
--------------------------------------------------------------------------
0       1000        281.54          429.36          565.25          973.46
1       1000        281.19          435.82          485.76          877.18
2       1000        285.75          433.47          482.67         1105.28
4       1000        285.90          435.85          486.66          820.09
8       1000        285.00          436.05          485.44          967.67
16      1000        304.95          438.99          620.95         1190.52
32      1000        286.06          445.63          631.08         1138.39
64      1000        287.67          450.57          502.23          888.69
128     1000        295.64          464.49          516.16          904.27
256     1000        342.10          511.82          565.61         1100.40
512     1000        366.18          632.50          657.13         2101.90
1024    1000        498.45          760.31          736.25         3239.72
2048    1000        779.28         1189.53         2699.41        11729.42
4096    1000       1155.15         1336.72         1530.59        12422.06
8192    1000       2257.39         2753.39         7470.11        22706.57
16384   1000       4199.82         4465.29        16938.53        30219.32
32768   1000      11746.70        12464.87        29844.82        72139.23
65536   640       24965.25        28936.41        42195.51       116714.49
131072  320       64966.33        64104.88       101934.76       231995.28
262144  160      132964.32       130781.39       143293.85       374124.26
524288  80       261793.36       263867.58       255780.38       651676.86
1048576 40       510482.33       520217.42       542601.23      1128027.18
2097152 20      1078979.11      1061722.18      1060231.26      2580126.19
4194304 10      2222642.98      2224997.34      2259047.59      4805922.79

8-node Throughput (MB/sec) Performance Results


#bytes  #reps   linux           dom0            domU(noSMP)     domU(SMP)
--------------------------------------------------------------------------
0       1000          0.00            0.00            0.00            0.00
1       1000          0.01            0.01            0.01            0.00
2       1000          0.03            0.02            0.02            0.01
4       1000          0.05            0.04            0.03            0.02
8       1000          0.11            0.07            0.06            0.03
16      1000          0.20            0.14            0.10            0.05
32      1000          0.43            0.27            0.20            0.11
64      1000          0.85            0.54            0.49            0.28
128     1000          1.65            1.05            0.95            0.54
256     1000          2.89            1.91            1.72            0.88
512     1000          5.33            3.09            2.97            0.97
1024    1000          7.86            5.14            5.30            1.28
2048    1000         10.02            6.57            2.99            0.68
4096    1000         13.52           11.69           10.21            1.25
8192    1000         13.85           11.37            4.18            1.44
16384   1000         14.88           13.99            3.69            2.06
32768   1000         10.64           10.02            4.19            1.73
65536   640          10.01            8.64            5.94            2.15
131072  320           7.69            7.79            4.91            2.15
262144  160           7.51            7.64            6.98            2.67
524288  80            7.61            7.55            7.78            3.05
1048576 40            7.77            7.62            7.30            3.50
2097152 20            7.28            7.41            7.42            3.02
4194304 10            6.93            6.98            6.87            3.13