Hola todos after On Mon, Oct 15, 2007 at 04:14:36PM +0200, elektra wrote:
I'm sorry to say that 0.3beta is not usable at the moment.
and my tests with 0.3-beta_~rv720 showed path-detection pretty broken, i tried exp-0.3. Finally i installed rv772 on 14 nodes real-world, rural nicaragua "non-profit-WISP/community network" parallel to existing olsrd(0.4.10/0.5.2/0.5.4pre)/ batmand_0.2-rv502. the solar powered nodes run bmxd-rv772 for almost 36hrs now, with no significant instabilities(with lots of tunneled download-traffic, the tunnel get's a bit unstable, what Axel already fixed in rv774). CPU-usage on client nodes looks less than in batmand_0.2, and slightly more than olsrd (would be interesting to see that scaling in a cloud with >100 nodes (how's the "massive parallel vm simulation" going btw?)), but is significantly higher (about factor 2) on gateway node. Best path-detection and usability/stability of routes so far. All nodes seem visible everywhere anytime, which is the case in olsrd as well, but not in batmand_0.2. I'm located two hops from the GW where my server also is. So i'm streaming music from there to check the stability of the route: with only olsrd it gets silent (with 1024KB disk cache in mplayer, ~40sec) pretty regularly (because of "collapsing routing tables" somewhere), with only batmand_0.2 it's even worse (because the GW node "doesn't hear my OGMs" too well). Both together are pretty usable. So i tried with only bmxd and disk cache set to 32KB (>1.5sec) and in tree hours music stopped once for <2sec ! from the "user-feeling" it's the way best mesh-daemon i tried so far. for fairness i have to say that my olsrds are neither all up to date, nor pretty well configured, but i beleave olsrd is a "historically crippled design" and the evolution of batmand will show the possibilities of wireless mesh networking. this all needs some further testing and improvements, but i'm pretty sure all the good bits will find their way into 0.3-final. Thanks to Elektra, Marek, Axel and everyone else who put so much energy into this amazing peace of free software.
BatMan-eXperimental 0.3-alpha rv772 (compatibility version 5)
/' '\ / __^..^__/ \ / / _ \vv/ _ \ \ / / \ / \ May the bat guide your path ...
cheers
--Jan
jan, please can you do me a favor and report if the release version 0.5.4 of olsrd was more efficient? sven-ola fixed a hug huge bug (the fish eye algorithm did not work!).
We experienced ~ 200% increase in cpu efficiency and ~ 300% in network efficiency between 0.4.10 and 0.5.4. c.f. http://olsr.funkfeuer.at for graphs.
I believe the 0.5.4pre still had that bug .
thanks for helping with testing .
a.
PS: Concerning the massive emulation... i am working on it..
On Nov 2, 2007, at 6:17 AM, Jan Hetges wrote:
Hola todos after On Mon, Oct 15, 2007 at 04:14:36PM +0200, elektra wrote:
I'm sorry to say that 0.3beta is not usable at the moment.
and my tests with 0.3-beta_~rv720 showed path-detection pretty broken, i tried exp-0.3. Finally i installed rv772 on 14 nodes real-world, rural nicaragua "non-profit-WISP/community network" parallel to existing olsrd(0.4.10/0.5.2/0.5.4pre)/ batmand_0.2-rv502. the solar powered nodes run bmxd-rv772 for almost 36hrs now, with no significant instabilities(with lots of tunneled download-traffic, the tunnel get's a bit unstable, what Axel already fixed in rv774). CPU-usage on client nodes looks less than in batmand_0.2, and slightly more than olsrd (would be interesting to see that scaling in a cloud with >100 nodes (how's the "massive parallel vm simulation" going btw?)), but is significantly higher (about factor 2) on gateway node. Best path-detection and usability/stability of routes so far. All nodes seem visible everywhere anytime, which is the case in olsrd as well, but not in batmand_0.2. I'm located two hops from the GW where my server also is. So i'm streaming music from there to check the stability of the route: with only olsrd it gets silent (with 1024KB disk cache in mplayer, ~40sec) pretty regularly (because of "collapsing routing tables" somewhere), with only batmand_0.2 it's even worse (because the GW node "doesn't hear my OGMs" too well). Both together are pretty usable. So i tried with only bmxd and disk cache set to 32KB (>1.5sec) and in tree hours music stopped once for <2sec ! from the "user-feeling" it's the way best mesh-daemon i tried so far. for fairness i have to say that my olsrds are neither all up to date, nor pretty well configured, but i beleave olsrd is a "historically crippled design" and the evolution of batmand will show the possibilities of wireless mesh networking. this all needs some further testing and improvements, but i'm pretty sure all the good bits will find their way into 0.3-final. Thanks to Elektra, Marek, Axel and everyone else who put so much energy into this amazing peace of free software.
BatMan-eXperimental 0.3-alpha rv772 (compatibility version 5)
/' '\ / \__^..^__/ \ / / _ \vv/ _ \ \ / \/ \ / \May the bat guide your path ...
cheers
--Jan
B.A.T.M.A.N mailing list B.A.T.M.A.N@open-mesh.net https://list.open-mesh.net/mm/listinfo/b.a.t.m.a.n
--- there's no place like 127.0.0.1
Hello Jan -
indeed, your report is the wage for our work. Your application is one of our prime targets.
Cheers! cu elektra
b.a.t.m.a.n@lists.open-mesh.org
-
Aaron Kaplan -
elektra -
Jan Hetges