Hi all,
I'm doing a research project on the topic discussed. I think the solutions proposed can improve the performance of the protocol, especially its reactivity in case of link failures. But in my opinion the main cause of performance deficiency is the route flap problem documented in some research paper. This is a common problem, especially in dense network. My work focus on finding out some strategy to reduce this problem maintaining at the same time a good reactivity in route decision. The first strategy I will try, is to introduce an hysteresis in route change and then I will study more sophisticated technique.
As soon as I have some written material I will send it to the mailing list hoping to receive some suggestion or critiques.
PS. I'm also working on a theoretical analysis on routing overhead in term of network utilization. Soon I will post some results.
Bye.
2010/12/2 Antonio Quartulli ordex@ritirata.org
Hi all*,
On gio, dic 02, 2010 at 11:27:52 +0100, Linus Lüssing wrote:
Hi everyone,
thanks for having joined the IRC channel, Hlabishi and Chris, that's really speeding the discussions up a lot, I guess :). Just wanted to throw in another idea open for discussions which is refering to Hlabisihi's packet count weighting.
What do you guys think about pushing this concept even one step further and removing the local packet count window? And instead using a "smoother" exponential weighted average which got very popular in TCP for it's RTT estimations:
RQ_new = f(newseqno, lastseqno, RQ_old, α) = (1-α) * RQ_old + α * 1 / Δ(newseqno, lastseqno)
As in TCP, α is then the parameter for tweaking on how much more priority a newer packet shall have and is chosen between 0 (conservative) and 1 (opportunistic).
What do you think?
IMHO this is not the best solution. Probably it could better than the actual strategy, but, as TCP demonstrates, this is not a really nice way of evaluating the RQ.
Imho, since we have more and more information than TCP for evaluating the new RQ, we could try to make a better choice considering "more values": instead of using only the deltaSEQ and the oldRQ, we could try to apply a function that can weight also the previous history of the link (of course with a limit). Something like
newRQ = f(window[0], ...., window[N-1]) in which, each OGM (received or not) can be weighted in a different way depending on its position.
I hope I've been clear.
furland is probably working on this aspect.
Bye!
Cheers, Linus
On Wed, Dec 01, 2010 at 01:30:18PM +0100, Marek Lindner wrote:
On Wednesday 01 December 2010 10:46:10 hlabishi kobo wrote:
Thank you for considering this concept, i am working on the implementation now. I would love to attend your IRC channel discussions, just tell date and time.
The IRC meetings are less formal. Just join us when you have the time. People hang out there all the time.
Cheers, Marek
-- Antonio Quartulli
Ognuno di noi, da solo, non vale nulla Ernesto "Che" Guevara
-- Daniele Furlan (furland)