From: David Jones <jones@mc...> - 2003-10-13 16:45:59
What have people gotten the best net performance from when using UML (eg:
switch daemon, tap devices, tap w/ bridge, etc.)?
For my UML cluster, I've just been using tap devices for each UML instance (no bridging). I
get transfer rates of 5.18MB/s between two UMLs on the same host and
6.5MB/s between two UMLs on separate hosts. As a reference, I get
10.2MB/s between the two hosts. This is over a 100Mbit connection and
using iPerf for benchmarking.
I'm hoping to get a transfer rate closer to 10MB/s for the UMLs since that
seems to be the biggest bottleneck in my setup.