4

While testing something with TFTP I noticed lots of timeouts (and slow speed as result) when I used my WLAN - and no problems when using a network cable. A quick test program sending/receiving UDP revealed that there are about 3-5% packets lost.

While it's obvious that WLAN has to be less reliable than LAN, I have no knowledge what loss rates are considered 'normal' - and when there is a need to further investigate the network infrastructure.

Are there 'typical' packet loss rates on WLAN (and other network technologies e.g. PowerLAN, WAN, ...?

Thanks

2 Answers2

1

I have no knowledge what loss rates are considered 'normal'

I don't think there is a normal - but I certainly think you're getting a poor deal. I'm currently at 835 retransmits out of 179943 packets with a really cheap USB wifi dongle - although I couldn't tell you if it's on the wifi hop where I'm losing data.

While Steve-o is correct in saying that you can mitigate the effects of packet loss via forward error correction, but he neglected to mention how to amend the default settings (AFAIK it's not possible). There are tools like udpcast if your only problem is moving data around and you're not bothered about applications/protocols (openvpn runs over UDP but no FEC option).

Or you might just hang around until QUIC is widely available.

Meanwhile I would suggest seeing if you can get better antennas at either end (you've not said what you're trying to connect).

symcbean
  • 19,931
  • 1
  • 29
  • 49
  • I would say terrifically poor. I am at the end of a point to multipoin WAN internet link and I get less than half a promilpercent e packet loss unless it is storming. WAY less. That is over a range around 3km from the provider. – TomTom Feb 25 '14 at 13:47
0

802.11a and g networks already have FEC incorporated into the protocol to guard against expected data loss, some discussion in this paper:

http://media.techtarget.com/searchMobileComputing/downloads/Finneran.pdf

However wireless networks are highly dependent upon environment and ultimately you have to choose what is an acceptable level. If there was a consistent level of noise you could design applications with technology to effectively hide the disruption and with such technology there is always a trade-off between performance and reliability.

For example you could always transmit double packets (XOR FEC) but you can only reach 50% maximum channel capacity.

Steve-o
  • 829
  • 6
  • 12