[Tinyos-help] CTP: ETX values vs. LPL wake-up interval

Manjunath Doddavenkatappa doddaven at comp.nus.edu.sg
Sun Nov 7 03:40:29 PST 2010


Dear Dr. Gnawali,

Sorry to get back to you so late, we had a paper deadline.

Everything works fine if I set LPL variables from the Makefile as in 
TestNetworkLpl. Problems arise only on trying to set LPL variables from 
the program using "setLocalWakeupInterval()". Just to verify in the later 
case, I retrived sleep interval value using "getLocalWakeupInterval()", 
the returned value is consistent with what I set it to.

Manjunath D

################################################################################################################
***************************************************************************************************************

On Fri, 22 Oct 2010, Omprakash Gnawali wrote:

> Are you using TestNetworkLpl and changing the flags in the Makefile to
> control various intervals? If not, you should tell me how to replicate
> your problem starting with TestNetworkLpl.
>
> - om_p
>
> On Thu, Oct 21, 2010 at 11:01 PM, Manjunath Doddavenkatappa
> <doddaven at comp.nus.edu.sg> wrote:
>>
>> Generally, yes. As ETX increases lossses also increase. I can observe the
>> fact that RetxmitTimer in CtpForwardingEngine being fired quite a few times
>> before sucessfully transmitting every packet.
>>
>> But, only a few times this does not hold with packets getting through
>> although ETX remains at a higher value.
>>
>> Please let me know if you want me to redo the experiments if you need any
>> other information.
>>
>> Regards,
>> Manjunath D
>>
>> ################################################################################################################
>> ***************************************************************************************************************
>>
>> On Thu, 21 Oct 2010, Omprakash Gnawali wrote:
>>
>>> On Tue, Oct 19, 2010 at 1:04 AM, Manjunath Doddavenkatappa
>>> <doddaven at comp.nus.edu.sg> wrote:
>>>>
>>>> Dear All,
>>>>
>>>>   We are not able to figure out a reason why ETX values in CTP increases
>>>> as
>>>> wake-up interval of the LPL (BoXMaC-One) increases. We have a simple
>>>> setup
>>>> of a sender and a root node placed adjacent to each other. Nodes use
>>>> maximum power of 0 dBm.
>>>>
>>>>   1) We are using IPI of 15s. We repeated the experiment at three
>>>> different places with sevral runs runs at every location. Different
>>>> wake-up intervals were tried back-to-back so that temporal variations are
>>>> minimized. We used wake-up intervals of 10, 20, 40, 100, and 1000 (ms).
>>>>
>>>>   2) We also used "LPL.setRemoteWakeupInterval(&packet, WAKEUP_INTERVAL +
>>>> 100)" but without any luck.
>>>>
>>>>   3) Our debug messages show that CTP's RetxmitTimer being fired quite a
>>>> few times before sucessfully transmitting every packet.
>>>>
>>>>   4) I first thought that the loss of data packets in the BoXMAC's
>>>> preamble could be the reason, but looking into the LPL code, I understand
>>>> that preamble packet losses are not fed into the Link Estimator.
>>>>
>>>>   5) In some experimental runs, particularly when the wake-up interval
>>>> is 1000ms, no packet go-through. Requires a hard reboot.
>>>>
>>>>  6) Searching the archieves for "CTP and LPL on tmote", I found that
>>>> similar problem being discussed sometime in Dec 2008. The discussion
>>>> suggests quite a few number of changes.
>>>>
>>>> http://mail.millennium.berkeley.edu/pipermail/tinyos-devel/2008-December/003510.html
>>>
>>>
>>> We should debug this. Do you also observe loss in delivery ratio along
>>> with high ETX values?
>>>
>>> - om_p
>>
>


More information about the Tinyos-help mailing list