It is east to forget which direction of the data payload when doing an IperfSpeed test. Does the data from from Server to Client or from Client to Server?. Well the answer from what I have read is from Client to Server. So this 'hack' is simply to modify the Iperf Speed html page to to show such a reminder.
- Here is part of a screenshot of what would be the result:
The procedure is simple:
- Ssh into the node that has IperfSpeed installed ("ssh -p 2222 root@nodename")
- Copy then Paste the following text into the ssh terminal and hit 'enter':
sed -i '/Run a Iperf Speed Test/c \ Run a Iperf Speed Test (Client sends data payload)' /www/iperfspeed/index.html
(The number of spaces in the middle is 28.)
The result can be immediately verified.
If someone wants to post a more elegant way of doing this, I would appreciate it. The 28 spaces seem like a goofy way to do this but it works. :)
- Mike
Hi, Mike
How about
sed -i 's,Run a Iperf Speed Test,Run an Iperf Speed Test <i>Client sends data payload</i>,' index.html
sed -i '/Run a Iperf Speed Test/c \ Run a Iperf Speed Test (Client sends data payload)' /www/iperfspeed/index.html
sed -i 's/Run a Iperf Speed Test/c \ Run a Iperf Speed Test (Client sends data payload)' /www/iperfspeed/index.html
I think you omitted the 's' (substitute).
3s, Chuck
Yes, nice simplification Chuck. Thanks.
I also like keeping the parentheses in it as this:
sed -i 's,Run a Iperf Speed Test,Run an Iperf Speed Test <i>(Client sends data payload)</i>,' index.html
73 - Mike
Hi, Tom:
I don't know.
I tried to view rc_stats, but I think the path changed in the recent nightlies.
cat /sys/kernel/debug/ieee80211/phy0/netdev:wlan0/stations/*/rc_stats
Chuck
When I use a throughput estimation tool like iperf3 it seems like I'm always comparing apples to oranges. Differences in hardware as well as what the CPU/Memory are doing on each end might account for some of the discrepancy. In our case the client is doing all the heavy lifting, so if you're using a client node with more CPU/Memory or one that is idle all the time the stats may look better. You could play with the client command line arguments to iperf3. The "-O 3" argument will omit the first three seconds from the stats, which typically raises the throughput because you're avoiding the ‘slow start’ portion of a TCP test. When I use my busier node (Hub) as the client I often get lower throughput numbers depending on how busy it and the network are at the time.
