Repository : ssh://git@open-mesh.org/doc

On branches: backup-redmine/2017-07-13,master

commit 10be13b54f63a5ca2ed03b7bc0b072a49c4d1ce0 Author: Linus Lüssing linus.luessing@c0d3.blue Date: Sun Jul 11 04:59:52 2010 +0000

doc: batman-adv/ELP

10be13b54f63a5ca2ed03b7bc0b072a49c4d1ce0 batman-adv/ELP.textile | 7 ++++--- 1 file changed, 4 insertions(+), 3 deletions(-)

diff --git a/batman-adv/ELP.textile b/batman-adv/ELP.textile index 521467bc..0d3de933 100644 --- a/batman-adv/ELP.textile +++ b/batman-adv/ELP.textile @@ -10,7 +10,8 @@ The idea is to strip the link-quality calculations from the normal originator-me

== packet type: BAT_PACKET_NDP == {{{ -/* Neighbor discovery packet */ struct batman_packet_ndp { +/* Neighbor discovery packet */ +struct batman_packet_ndp { uint8_t packet_type; uint8_t version; /* batman version field */ uint8_t orig[6]; @@ -48,6 +49,6 @@ Every node could rebroadcast an OGM i.e. 3 times instead of just once to increas === Adaptive ndp interval ===

=== Probablistic flooding === -Although this does not optimize the convergence speed, it could greatly increase the efficiency of the flooding of broadcast/mulitcast-data packets. With ndp we now also know our neighbors' TQ-values on the same interface. With this information a node A can determine the probability, that a broadcast of node B has already reached all of A's neighbors: P(|N| = |N^{+}|) = \prod_{n \in N} TQ(B, iface, n) with N being the set of A's neighbors and N^{+} the set of neighbors successfully receiving the broadcast. +Although this does not optimize the convergence speed, it could greatly increase the efficiency of the flooding of broadcast/mulitcast-data packets. With ndp we now also know our neighbors' TQ-values on the same interface. With this information a node A can determine the probability, that a broadcast of node B has already reached all of A's neighbors: P(|N| = |N^+^|) = ∏_{n ∈ N} TQ(B, iface, n) with N being the set of A's neighbors and N^+^ the set of neighbors successfully receiving the broadcast.

-A node could then either decide with a threshold, if it might not rebroadcast (rebroadcast, if P(|N| = |N^{+}|) < 0.9 i.e.). Or more elegantly rebroadcast to a certain probability: 1 - P(|N| = |N^{+}|) (or 1 - P(|N| = |N^{+}|)^3 for a more conservative decision i.e.). \ No newline at end of file +A node could then either decide with a threshold, if it might not rebroadcast (rebroadcast, if P(|N| = |N^+^|) < 0.9 i.e.). Or more elegantly rebroadcast to a certain probability: 1 - P(|N| = |N^+^|) (or 1 - P(|N| = |N^+^|)^3^ for a more conservative decision i.e.). \ No newline at end of file