ok, here we go...
i think the reason why the test failed in the past wasn't the overall transfer-rate but the slowdowns... so the goal is a constant stream not a very high throughput (i did a test record with TNT-Serie HD and needed about 15Mbit/s aka ~2MByte/s and during testing i reached 9MByte/s)
to trace this i opened TaskManager on my Server PC, in general you will need some kind of network monitor with a throughput graph...
*first we disable nfsmount and record_to_nwshare scripts and reboot tv
*ssh to tv and create a mountpoint for testing
then we only have to repeat the test procedure and watch our network graph / test results
it goes like
*mount the nfs share with test settings
*check if mount has been done with the specified parameter
*copy over / write some test-files
*unmount again to be ready to remount with different settings
Code: Select all
# /bin/busybox mount -t nfs -o soft,udp,async,nolock,rsize=32768,wsize=8192 192.168.1.4:/nfsshare /tmp/mnt
# /bin/busybox mount
(...)
192.168.1.4:/nfsshare on /tmp/mnt type nfs (rw,vers=3,rsize=32768,wsize=8192,soft,nointr,nolock,proto=udp,timeo=7,retrans=3,sec=sys,addr=192.168.1.4)
# dd if=/dev/zero of=/tmp/mnt/ddfile bs=1000 count=512000
512000+0 records in
512000+0 records out
512000000 bytes (512 MB) copied, 58.0756 s, 8.8 MB/s
# /bin/busybox umount /tmp/mnt
as you may notice the mount has been made without the async option, undesirable but no disaster...
now play a bit with rsize and wsize (for reading and writing direction)
start with 1024 and then go on doubling the values (1024, 2048, 4096, 8192, 16384, 32768)
stop when the network utilization doesn't increase from test to test
you may also test different filesizes (up to 2GB perhaps, its multiplied of bs for blocksize and count)
also switch between tcp and udp...
above you see the settings i ended up with