Home
Reading
Searching
Subscribe
Sponsors
Statistics
Posting
Contact
Spam
Lists
Links
About
Hosting
Filtering
Features Download
Marketing
Archives
FAQ
Blog
 
Gmane
From: Antonio Quartulli <ordex-GaUfNO9RBHfsrOwW+9ziJQ <at> public.gmane.org>
Subject: Re: DHCP servers and Subnets in Wireless Mesh Networks [Mesh Architecture]
Newsgroups: gmane.org.freifunk.wlanware
Date: Friday 9th November 2012 14:39:40 UTC (over 3 years ago)
On Thu, Nov 08, 2012 at 11:28:35AM -0500, Naman Muley wrote:
 
>    I was reading on with the BATMAN-adv wikis. I understand the
functionality
>    to an extent. Here's my question with respect to the translation
table: 
>    I guess the translation table keeps record of the clients connected to
a
>    particular node. Now, my question is, doesn't this get exhaustive?
>    Typically looking at my deployment scenario, say a library, or a
lecture
>    theater, will have atleast 400 people at a place. (Not going into load
>    balancing on one node! ) will the size of the table not become large?
I
>    mean then performing a search operation every time a packet addressing
the
>    client comes and sending it, will that not become time consuming? As I
>    understand, in the wired counterpart, switches are supported by CAMs
which
>    are fast at search operations but a normal router working on it, won't
the
>    load become overwhelming very quickly? 
>

This is a good point.
For sure batman-adv, which is implemented in-kernel, cannot achieve the
same
performance as CAMs. Batman-adv uses its own hash table implementation to
perform this task, trying to speed it up as much as possible.
I think there are some examples out there where they use batman-adv with
~300
clients and they didn't report back any issue related to this. You should
also
keep in mind that batman-adv probably will not have to handle Gbps of
traffic,
as high-level wired bridges do. If you want to handle that high bandwidth,
maybe
the wireless medium is not the best approach at all?

> 
>      This is possible with batman-adv, but also with layer3
>      protocols (like OLSRd or Babel): simply because the clients will
become
>      part
>      of the node network segment. However, as already mentioned before my
>      reply, with
>      batman-adv the (mesh-un-aware) clients will have their DHCP packets
>      re-routed in
>      a smart way so to choose the best (from the batman-adv metric point
of
>      view)
>      DHCP server. In case of a layer 3 protocol, each node should also
run
>      the DHCP
>      server, unless you think to some a bit more complex solutions.
> 
>    Coming back to say an OLSR implementation, I don't really understand
well
>    enough how will clients move freely if there are multiple subnets.
Moving
>    from one MR supporting one sub-net to another will require an IP
handover.

I do not use OLSRd that frequently, but I think that in this case the
problem is
not OLSRd related anymore: here you probably need to take into
consideration
mobile IP and other stuff like that. I think there was an OLSRd plugin some
time
ago which was trying to help in this direction..but I don't know whether is
still exists and if it is actively developed or not.

Cheers,

-- 
Antonio Quartulli

..each of us alone is worth nothing..
Ernesto "Che" Guevara
 
CD: 16ms