I have recently purchased a NetComm NP285 Turbo HomePlug Twin Pack from TradeMe, which are Ethernet over Power bridges. Information regarding standards and so forth can be found at the HomePlug Powerline Alliance. In brief, these particular adaptors work at up to 85Mbps (earlier version was 14Mbps and up and coming version at 200Mbps).
This post is trying to give you some information about the performance I was able to attain in my flat, which is quite old, both with regard to typical round-trip-time and throughput.
So, here is a picture of the adaptors, and a photo of the mains board, which should give you some idea of the age of the wiring (1950s) in our four-person flat.
One adaptor was plugged directly into the wall. The adaptor is quite wide, so you won’t be able to plug into an adjacent socket. Thus, it is likely that a multi-box or double-plug would be necessary; this is not the recommended configuration, but it wouldn’t be very lifelike otherwise; and is necessary in my particular environment. The other adaptor was connected to a 6-plug multi-box, along with a printer (on standby), and an old iMac G3. The TV was also on, and I had a load of washing going, so the power lines would not have been clean or quiet, especially with a large motor in a washing machine going at the time; actually its probably nearer to worst-case end of the spectrum than to lab-conditions.
I shall test both round-trip-time and throughput. I can’t meaningfully do these at the same time, so I shall run the round-trip test first and throughput test in sequence, repeatedly. The round-trip-time will be done with my wrel tool (see the Public Files section of my site). The throughput will not be done with some naïve file-transfer, but with ttcp.
Results
while true
do
ttcp -p 2000 -t -s -n 8192 -f m 169.254.192.198 2>&1 | grep 'real seconds'
~/bin/wrel 169.254.192.198 homeplug.`date +%H%M%S`.png 1000 "HomePlug single hop" 1492
done
The throughput (each runthrough taking about 32 to 48 seconds) took between 10 and 15Mbps while the washing machine was on. Ironically, it was a little slower (9 or even 8Mbps) before the washing was put on; however this may have been due to other circumstances.
With regard to the round-trip-time, the following distribution is very typical, with a small set of peaks taking over 100msec.
Conclusion
The results were not satisfying, although coping with such old wiring certainly wouldn’t help matters. This is a) not surprising to my flatmate, who has more knowledge of electrical engineering than I do, and b) still much greater than our Internet connection, which is where all the traffic on our wireless connection ends up going.
The short of it is that it allows us to move our 802.11g (54Mbps max) wireless access point to a better location without having to run ethernet cable in an unsightful way, although it does mean a potention hit to local WiFi performance, it should mean that we can increase reliability of the WiFi with regard to range. This is important because if the AP is not very well situated, there can be places where it become very likely that a TCP packet will be dropped. Such an event is very disruptive to interactive traffic, such as SSH, so this is to be avoided.