[Tinyos-devel] retransmissions in networking protocols
koepke at tkn.tu-berlin.de
Thu Oct 4 08:05:38 PDT 2007
I'm talking about lost ACKs, not lost packets. Whenever only the ACK is
lost, a duplicate is created by the hop by hop recovery mechanism of the
network protocols. My question is: why is there no mechanism in place to
filter duplicates out?
Omprakash Gnawali wrote:
> In terms of software architecuture, are we talking about hop-by-hop
> retransmission or MAC level retransmission?
> In net2, we had a related discussion a while ago. Please also go to
> this page:
> and read the discussion with subject:
> "L2 retx and CTP FE"
> - om_p
>>I'd vote for including hop by hop filtering if we have hop by hop
>>retransmissions in the network protocols. Or remove retransmissions from
>>the network and let the MAC/LinkLayer do this kind of stunts.
>>Matt Welsh wrote:
>>>Actually these protocols perform retransmissions at the *link* layer,
>>>not the network layer (the latter would imply end-to-end).
>>>I'm firmly in the camp that link-layer ARQ is a necessity with these
>>>kinds of radios, due to interference and collisions (the MAC could
>>>perhaps be tuned to help avoid the latter). Better route selection -
>>>choosing only 'symmetric' links - can cut down on duplicate packets. One
>>>could imagine a node detecting many duplicates from its children telling
>>>those children to reduce their ARQ threshold. Even so I think filtering
>>>out duplicates hop-by-hop or at the base station is perfectly
>>>On Thu, 2007-10-04 at 14:02 +0200, Andreas Koepke wrote:
>>>>Hi guys in the net group,
>>>>looks like MultihopLqi and Ctp all do retransmissions on the network
>>>>level if the MAC reports a transmission failure. So whenever the MAC did
>>>>not receive an ACK the packet is duplicated. This happens quite often in
>>>>our testbed, because the TX power of the nodes seems to have a
>>>>What is your experience, should we do something about it?
>>>>Tinyos-devel mailing list
>>>>Tinyos-devel at Millennium.Berkeley.EDU
>>Tinyos-devel mailing list
>>Tinyos-devel at Millennium.Berkeley.EDU
More information about the Tinyos-devel