Re: Hierarchical addressing (fwd)

[Lloyd Wood <L.Wood@surrey.ac.uk>]

On Thu, 5 Nov 1998, Shuqian Yan wrote:

> Below is a forwarded message about the hierarchical addressing format
> should be used for large multicast simulations (with more than 256 nodes
> in your graph...)

with more than _128_ nodes. The high bit is always reserved for
multicast, irrespective of whether multicast is used or not.

(This trips up a lot of people; I'm tempted to suggest something like

 $ns set-maximum-number-of-nodes n

 where n is rounded up to the nearest power of 2 and then doubled
 for the multicast bit to give the addressing space, and warnings are
 issued if n is exceeded at any point. If you're creating/destroying
 nodes via some complex routine or other, it could be a useful bounds
 check.)

On recent ns:

$ns set-address-format hierarchical 1 12

should give you a flat addressing space (of 2^11, 2048, nodes) where
the hierarchical routing flag should automatically be set to zero,
since there isn't really a hierarchy.

At least, this I what I've picked up.

L.


> ---------- Forwarded message ----------
> Date: Thu, 5 Nov 1998 15:44:58 -0800 (PST)
> From: Padmaparna Haldar <haldar@ISI.EDU>
> To: Shuqian Yan <syan@jupiter.nal.utoronto.ca>
> Subject: Re: Hierarchical addressing
> 
> 
> Shuqian,
> I think if you really dont care about hier rtg, it'd be convinient for you
> to just call Node expandaddr. this would give you an addr space of 30 bits
> with 8bits allocated for port and rest for nodeaddr.
> also be careful to allocate mcast address by calling "Node allocaddr".
> 
> Otherwise if you do want to use hier addressing you'd have to generate
> hierarchical node address (using the right topology generator/converter)
> and you'd have to set the EnableHierRt_ flag to 0 explicitly.
> 
> Obviously the first option seems more attractive/easy to me!
> cheers,
> --Padma.
> 
> 
> On Thu, 5 Nov 1998, Shuqian Yan wrote:
> 
> > Hi, Padma
> > 
> > My only intesnsion now is to expand the address space without any
> > hierarchical routing, so with "Node expandaddr" augmented in the
> > simulation script will do the work. or do I have to also disable the
> > EnableHierRt_ flag (set the flag to "0" since I am only have one level
> > hierarchy), I found one of the posting in the ns archives you've
> > mentioned this.
> > 
> > tks,
> > 
> > shuqian
> >  
> > 
> > On Thu, 5 Nov 1998, Padmaparna Haldar wrote:
> > 
> > > On Wed, 4 Nov 1998, Shuqian Yan wrote:
> > > 
> > > > Hi, Padma
> > > > 
> > > > My question: are all the mulicast routing protocols currently available in
> > > > NS supporting the hierarchical addressing scheme (as opposed to flat
> > > > addressing), thus one can simulate large graphs with more than 500 nodes? 
> > > 
> > > Hierarchical addressing should work for any routing protocol you use
> > > (unicast/mcast). However I'm not sure if you want to use hierarchical
> > > routing as well, which is supported in mcast extensions as well.
> > >  If you simulate large topologies using DM you should be
> > > able to do that by simply expanding addrspace thru "Node expandaddr".
> > > hierarchical routing can also be used for memory savings. 
> > > You can see examples for both in ~ns/tcl/ex
> > > 
> > > Hope this helps,
> > > --Padma.
> > > 
> > > 
> > > > 
> > > > If for example the DM protocol is only supporting flat addressing and I
> > > > want to simulate large graphs using available DM, what are the changes
> > > > should be made?

<L.Wood@surrey.ac.uk>PGP<http://www.ee.surrey.ac.uk/Personal/L.Wood/>