Hi Rahul, this sounds as if fragment forwarding would help already? (Am I reading this right that the discards are buffer space discards, not retransmission count packet losses?)
Grüße, Carsten > On Apr 6, 2017, at 15:39, Rahul Jadhav <[email protected]> wrote: > > Hi Carsten, > > Currently I m reassembling hop by hop. > > Problem begins with initial auth session itself!! I use PANA for network > access in single channel (with no RDC) mesh and the payload in most PANA > messages crosses 127B (especially considering the relay header overhead),,, > its all unicast with MAC layer ACK. None of my mcast packets cross MTU > boundary. Also an imp point is that the buffer space is only for one IPv6 > datagram (i.e. 1280B). > Some of the problems highlighted by Pascal such as, fragments getting > discarded prematurely for e.g. because of high node density causing reception > from different peers is hurting the end to end msg transmission badly. The > failure rate and thus the retransmission rate at app is very high,, so much > that we had to adopt a proprietary compressed PANA signalling mechanism to > improve convergence time. > > Regards, > Rahul > > On 6 April 2017 at 18:24, Carsten Bormann <[email protected]> wrote: > On Apr 6, 2017, at 10:57, Rahul Jadhav <[email protected]> wrote: > > > > Fragmentation at 6lo layer as it is today is almost unusable for a > > moderately sized network with lower MTUs such as 127B. > > Hi Rahul, > > please explain some more. > > What kind of problem do you experience? > > Is the load on your network predominantly composed of packets > 127 B? > > Are you using MAC-layer ACK? > > Do the problems you are seeing occur with Unicast or with Multicast? > > Are you already using fragment forwarding or are you reassembling hop by hop? > > Grüße, Carsten > > > _______________________________________________ > 6tisch mailing list > [email protected] > https://www.ietf.org/mailman/listinfo/6tisch _______________________________________________ 6tisch mailing list [email protected] https://www.ietf.org/mailman/listinfo/6tisch
