Performance Tuning

Run3, Analysis + next steps

By adding process affinity things have changed
●
The bottleneck is now on the transmit side
●
core5 on TX is 100%, core5 on RX side is handling the load

Try moving process affinities around (already done)

Change the code (if you can)
●
Default netperf method uses send() which copies data around
●
Try TCP_SENDFILE which use the sendfile() system call

Try bigger MTU
●
Currently at 1500