[Olsr-users] Scalability of Networks using Olsr_Switch

[Lo, Caleb K.]

Hi Henning and Markus,
                Thank you for your helpful replies and insights.
I did try running the test that I mentioned below (with olsr_switch) for olsrd-0.6.1, and I ran into the following error: "Error, cannot aquire [sic] OLSR lock '/var/run/olsrd-ipv4.lock', Another OLSR instance might be running. OLSR EXIT: Terminated"
So I'm not sure if olsr_switch is compatible with the newer olsrd versions that you listed.  As for the seg fault error that I mentioned below, I'm wondering if by simply being more generous in the memory allocation of a few pertinent arrays, I can support more distinct olsrd processes.
At this point, my intention is to see how many distinct olsrd processes can be supported on a single Linux kernel.  Ideally I'd like to have 100 distinct olsrd processes running simultaneously on the same kernel; from browsing the olrsd website, it seemed that olsr_switch would be a useful tool for me in that regard.
If my proposed bugfix (i.e. more generous memory allocation) doesn't work, I'll go with your suggestion of using olsrd-0.6.1 in conjunction with a few VMs.  I do see that Xen has been discussed on the mailing list, so that would be a good place for me to start: http://lists.olsr.org/pipermail/olsr-users/2009-April/003389.html

Caleb Lo

From: (spam-protected) [mailto:(spam-protected)] On Behalf Of Markus Kittenberger
Sent: Wednesday, December 15, 2010 8:12 PM
To: Lo, Caleb K.
Cc: (spam-protected)
Subject: Re: [Olsr-users] Scalability of Networks using Olsr_Switch

use a newer olsrd version (e.g. 0.6.1 is a current one),.
all 0.5.x wasn't very stable, maybe except the latest 0.5.6-Rx
(but hmm, i'm not 100% sure if the hemu code is still in 0.6.x)

but even 0.5.4 "might" scale to thousands,.. (depending on your cpu and ram limits) they where definitely used (somewhat reliable) in meshes of several hundred nodes,..

but they are buggy, an will segfault from time to time,..

Markus

p.s what is your goal intention?

On Thu, Dec 16, 2010 at 1:22 AM, Lo, Caleb K. <(spam-protected)<mailto:(spam-protected)>> wrote:
Hi everybody,
                I'm running olsrd-0.5.4, and I'm trying to simulate a large network using the olsr_switch application.  I've tried to set up a large network in host-emulation mode using the instructions in the "Running olsrd in host-emulation mode" section of: http://www.olsr.org/docs/README-Olsr-Switch.html
The command that I used was "./olsrd -d 1 -hemu 10.0.0.x", where x ranges from 1 to 17 (inclusive).  I'm using the olsrd.conf file that was provided in the tarball from the olsrd website, except that I disabled hysteresis based on the following run-time error that I observed: "Hysteresis and LinkQuality cannot both be active!  Deactivate one of them.  Bad configuration!"
Using GNU Screen, I started olsr_switch in one window, and then I started (sequentially) olsrd in 17 new windows.  When I started the 17th olsrd process, I observed the following error on the other processes: "Could not read whole packet (size 51592, read 31848).  Segmentation fault", and they all crashed.
Is this a fundamental limitation (in terms of scalability) of olsr_switch, or is this a bug that could be fixed by changing some appropriate constants in the olsr_switch and/or olsr directories?
Thank you very much.

Caleb Lo

--
Olsr-users mailing list
(spam-protected)<mailto:(spam-protected)>
http://lists.olsr.org/mailman/listinfo/olsr-users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.olsr.org/pipermail/olsr-users/attachments/20101216/b7aedfb6/attachment.html>