Hi All, I am facing a strange packet loss behaviour in my network which is affecting my final project. I would like to ask a few questions for the same.
*TestBed*: 15 motes running the simple apps/RadioCountToLeds application which sends periodic(5 sec) brodcast messages, the* additions *are that I switch Off the Radio(to improve liftime of the mote) after SendDone and start it when the timer fires, also I have removed the receive interface. I am running the apps/BaseStation code for receiving all these messages and then check the packet loss through java. The nodes are kept close to each other. I am using tinyos-2.1. So, 15 motes sending a packet every 5 seconds and a BaseStation receiving. *Result*: Packet loss in high numbers and the more troubling statistics was that I lost upto 15 consecutive packets from a particular node. My application is based on calculating the presence of the motes in a particular period, lets say 30 sec(6 times the frequency), so 15 packets result into a lost mote which is not true. *Acknowledgements*: Are there ACK's for broadcast packets ? So, would my application wait for ack's and if it doesnt arrive retransmit ? If not, can i enable it somehow ? I read many replies on the mailing list but cudnt arrive to a conclusion. I have tried putting the following two methods(similar to TestPacketLink) but according to this I never lose any ack and packets are always ACKed. call PacketAcknowledgements.requestAck(&packet); call PacketAcknowledgements.wasAcked(bufPtr). Also, I had commented CFLAGS += -DCC2420_NO_ACKNOWLEDGEMENTS in MakeFile for the BaseStation. *BaseStation Queue*: This sounds absurd, but is it possible that BaseStation's receive queue cannot handle packets that arrive at more or less the same time, so the Radio receives the packet and sends ACK(if it does) but the Queue somehow looses some packets. I presume that cc2420 is fast enough to handle such slow rate here but can anybody prvovide the statistics for reception of colliding packets ? how close can they be ? *Tweak* - If I add a delay of 100ms in SendDone before switching Off the Radio, the packet loss is much much less and max number of consecutive packet loss from one node is 6. Further if I add a delay of 200/300 ms, there is a drastic decrease in packet loss and I dont loose more than 1-2 consecutive packets from a particular node. I cant understand whatcan be the reason behind the dramatic change ???? ACK's ? but 100/200/300 ms is a very big time. *Channel*: This test is done in a regular office environment where ofcourse there is wireless and other regular interferences, but nothing that can be pinpointed as 'the' problem. Will changing the default channel on which the motes are sending packets have any positive effect ? What amount of packet loss rate should be expected from the motes in such a simple environment and what can be the reasons ? *Bigger Picture: *Putting 200ms delay leads to an acceptable packet loss rate but decreases the lifetime of the nodes and also when I make the network a bit more complex e.g adding some nodes with MultiHopOscilloscope app which can also accept the broadcast messages from these RadioCountToLeds motes, the packet loss re-emerges. So understanding some of the above Questions is necessary for the bigger picture. Waiting for some hints, Gaurav -- --------------------------------------------------------------------------------------------------------------------- To accomplish great things, we must not only act, but also dream; not only plan, but also believe. - Anatole France
_______________________________________________ Tinyos-help mailing list [email protected] https://www.millennium.berkeley.edu/cgi-bin/mailman/listinfo/tinyos-help
