Why observed different performance on the replica pairs in distribute-replicate volume in Red Hat Gluster Server?

Issue

• Why observed different performance on the replica pairs in distribute-replicate volume in Red Hat Gluster Server?

• The NFS write performance is vary depending on the mount location. What's making so different performance?

# for i in `seq 1 5`;do dd if=/dev/zero of=/mnt/nfs_v001/file-$i bs=8K count=131972;done
131972+0 records in
131972+0 records out
1081114624 bytes (1.1 GB) copied, 10.1513 s, 106 MB/s
131972+0 records in
131972+0 records out
1081114624 bytes (1.1 GB) copied, 19.1493 s, 56.5 MB/s
131972+0 records in
131972+0 records out
1081114624 bytes (1.1 GB) copied, 9.94428 s, 109 MB/s
131972+0 records in
131972+0 records out
1081114624 bytes (1.1 GB) copied, 19.1462 s, 56.5 MB/s
131972+0 records in
131972+0 records out
1081114624 bytes (1.1 GB) copied, 19.1306 s, 56.5 MB/s