[Tinyos-help] MultiHopOscilloscope Packet Loss

[Gaurav Chandwani]

Hi Omprakash,

Thanks for the quick reply.

Viable path - I see the node lost behaviour even with nodes that are
directly connected(no hop, very close) to the root node.
Do you mean something else by viable path ?

How much traffic amounts to saturation of the channel ?
I have 10 nodes having MHO code sending a packet every 5 sec, i think this
should make very less traffic ?

If i put tests/TestNetwork on all the nodes, to see the debug
messages, i have to
enable a flag or these are simply printf statements ?
Right now, I have only root node connected to the PC, but as you
suggested, to check the debug messages,i can  connect other nodes too.
Will try this.

Regards,
G











On Wed, Jun 25, 2008 at 4:40 PM, Omprakash Gnawali <gnawali at usc.edu> wrote:
> On Wed, Jun 25, 2008 at 7:09 AM, Gaurav Chandwani
> <gaurav.chandwani at gmail.com> wrote:
>> Hi All,
>>
> ...
>> I saw a strange behaviour using SerialForwarder - When the application
>> is running for some hours, sometimes, i dont receive messages from a
>> particular node for some minutes and then it comes back again.  Its
>> reproduced many times.
>>
>> Has anybody seen this kind of behaviour ?
>> Since MHO is stable, so there should be a problem at my end but I am
>> not able to find where.
>
> If there is no viable path between the nodes and the root, CTP won't
> be able to deliver packets. Have you considered that possibility?
>
>> Since the nodes can reach root node directly, they will send message
>> only to it or also other nodes ?
>
> It can be either way depending on the link and path qualities.
>
>> Can there be a problem with root node sending of messages on UART or
>> some other queue overflow ? I am using baud rate of 115200.
>> Next, I will try to use printf to understand at which level the problem is.
>
> Queues should not overflow if you are not sending a lot of packets and
> if there are viable paths. It is possible to overflow a queue if a
> node is not able to drain it fast enough (too many retransmissions,
> for example)
>
>> As i read, the reliability of CTP should be 99.9% at low traffic rate,
>> but the 0.1% messages that are lost, are lost
>> periodically or there is a possibility that they can be lost in chunks
>> of messages ? and what will be high traffic rate ?
>
> If you are planning to saturate the channel, you should expect a lot of losses.
>
> If these motes are connected with a PC during the experiment, I
> suggest running tests/TestNetwork and looking at the debug messages
> logged by each mote to understand the cause for the loss you are
> seeing.
>
> - om_p
>



-- 
---------------------------------------------------------------------------------------------------------------------
To accomplish great things, we must not only act, but also dream; not
only plan, but also believe.
 - Anatole France