> From other comments, it seems like this device is just not that good.
Yes :-) I have two in case someone wants one! I have since purchased an axen(4) device and am seeing improved performance. > The only suggestions I would have to debug it are: > > - Test in UDP mode also. TCP is extremely sensitive to loss. > iperf/TCP measures how well you can do TCP which is what you usually > care about, and if it's close to line rate (even 350/1000), you know > all is well. But if it's slow, you don't know if each packet is > slow with little loss, or loss often, or loss above a certain rate, > or ? > > - Look at "netstat -i" and error counters, and also do "netstat -s" > before and after and diff them, on a system with nothing else > happening (maybe even single user). Anything that changes that you > don't expect is interesting and I've found a lot of issues this way. > > - Run tcpdump and save to a file, and use tcpdump2xplot from > graphics/xplot-devel to visualize it, after patching up the perl > program or editing the trace to make it look like tcpdump used to > look. This is a lot of effort, and requires you to not only learn > to deal with that software, but to really understand TCP. But it > can be very illuminating as to what is going on. > > You can also select one stream and just examine the text output. Thank you for these suggestions. I'll keep them in mind for when I have extended time to explore. -- matth...@fastmail.us
