[Xorp-users] Using XORP for PIM-SM

Pavlin Radoslavov pavlin at ICSI.Berkeley.EDU
Thu Jun 5 15:28:37 PDT 2008


Ben Greear <greearb at candelatech.com> wrote:

> James Courtier-Dutton wrote:
> >> 2008/6/5 Ben Greear <greearb at candelatech.com>:
> >>> James Courtier-Dutton wrote:
> >>> 3) How many tunX interfaces can XORP handle?
> >>>
> >> Multicast will have a limit of 32 interfaces per table in my implementation.
> >>  It would require
> >> even more API change and kernel hacking to remove this limitation..and I
> >> don't need more,
> >> so I don't plan to do this.
> >>
> > 
> > Ben,
> > 
> > Thank you for the quick response. :-)
> > I am interested to understand why the 32 interfaces limit is there.
> > Which kernel API calls are limited to 32 interfaces?
> > I might want to add support for more than 32 interfaces. I am thinking
> > of maybe 1000 interfaces.
> > The total packet bandwidth over all the 1000 interfaces is likely to
> > be low, about 4MBytes/s.
> 
> The linux kernel uses a 32-bit number as a bitfield, and at least some of that
> is visible in the user-space to kernel API as well.

The 32-bit bitfield was probably used in older kernels and/or
userland programs.
Now the olist in the kernel for example should be specified in an
array of size MAXVIFS which can be redefined as appropriate.

Approx. 2 years ago I had an interaction with a fellow who also
wanted to increase the number of max. vifs in the Linux kernel.
I believe he was able to achieve few hundred vifs, but I don't know
whether he needed any more than this or whether he was hitting some
other limits.
The notes (and the discoveries re. other system limits) in the
xorp/ERRATA file on the subject are from that experiment with him.
In other words, it should be possible, but requires kernel and XORP
recompilation, and system parameters tweaking. See the IGMP/MLD and
PIM-SM entries in the xorp/ERRATA file for details.
If you find some other limits, please us know so we can update the
documentation.

Regards,
Pavlin

> According to comments, seems to be derived from ancient BSD logic,
> or something like that.
> 
> I don't plan on making these changes for my own needs, but
> I know people who do this sort of thing for hire if you are interested
> in paying for the feature...
> 
> Thanks,
> Ben
> 
> 
> > 
> > Kind Regards
> > 
> > James
> 
> 
> -- 
> Ben Greear <greearb at candelatech.com>
> Candela Technologies Inc  http://www.candelatech.com
> 
> _______________________________________________
> Xorp-users mailing list
> Xorp-users at xorp.org
> http://mailman.ICSI.Berkeley.EDU/mailman/listinfo/xorp-users



More information about the Xorp-users mailing list