doing some NFS benchmarks. server and client are on the same hypervisor. mounted and exported a ramdisk. writes were done before reads to prevent caching.
server write
dd if=/dev/zero of=/mnt/ramdisk/zeros bs=10k count=50000
50000+0 records in
50000+0 records out
512000000 bytes (512 MB, 488 MiB) copied, 0.717472 s, 714 MB/s
server read
dd if=/mnt/ramdisk/zeros of=/dev/null bs=10k count=50000
50000+0 records in
50000+0 records out
512000000 bytes (512 MB, 488 MiB) copied, 0.555382 s, 922 MB/s
client write
dd if=/dev/zero of=nfs_ramdisk/zeros bs=10k count=50000
50000+0 records in
50000+0 records out
512000000 bytes (512 MB, 488 MiB) copied, 2.75018 s, 186 MB/s
client read
dd if=nfs_ramdisk/zeros of=/dev/null bs=10k count=50000
50000+0 records in
50000+0 records out
512000000 bytes (512 MB, 488 MiB) copied, 14.0721 s, 36.4 MB/s
186 MB/s write is pretty good; i was only expected to get 1 Gb/s max but I guess the virtual netdev can handle more than that. interesting how the client read is significantly slower, though. thoughts?