[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Active networking in ns



[I'm forwarding this to the list again]

>I see. In that case, mcast won't help. Have you looked at this?  They
>might have fixed classifier.cc in a way that packets marked as 'active'
>will be sucked up to the agent level.
>
>http://www.tascnets.com/panama/
>Active Networking Simulation Packages for ns versions 2.1b1 and 2.1b4 are
>available:
>     Active Network Package (v-ns2.1b5.tar.gz) (v-ns2.1b4.tar.gz)
>     (v-ns2.1b1.tar.Z) for ns:  Files for the LBNL network simulator ns
>     version 2 (available for download from LBNL).  The package include
>     C++ files that create Active Agents with basic functionality
>     (initialization and disabling) and a sample tcl script that
>     illustrates this basic behavior.  Directions for installing the
>     package can be found in the README file in the package.  

I have been studying this for a while in the past days, and there
are only two things about it:

- Although they intended to release this package as a 'generic'
extension to ns, in fact it is very focused towards AER, their
own active protocol; or at least towards active protocols for
multicasting. There is nothing wrong with this, but I want to
simulate a very large network (2000+ nodes) in which all the
nodes may have a built-in data cache (which I am not modeling
'in full' but for which I have to track at least which
requests are being cached at each node); so I need to keep
the nodes as simple as possible and remove all unneccesary
detail.

- Also, this solution involves a modified classifier.cc, which
I would like to avoid unless I'm forced to use it (same reason
as above).

Anyway if I don't find anything better suited to my needs, I
will end up using this package, or more probably I will use it
as the basis for a modified, more 'light-weight' version.

Thanks!

>> > DOes n1 intercept the packet, process it, and then decide what to
>> > do next (for example forward it down to n2) or n1 make a quick copy of
the
>> > packet and process it, while the packet continues to be fowarded down
>> > to n2? If it's the latter, you can probably use a multicast address for
>> 
>> Ops, I forgot to mention this. When n1 gets the packet and processes it,
>> it decides what to do next. In my case, it will either let it go towards
>> n2 or send it back to the original destination (n0), depending on
>> whether it can solve the request originated at n0.
>> 
>> In fact, active protocols are commonly used to do some 'custom routing
>> and forwarding' by embedding small programs in the protocol packets
>> that specify which processing is to be done at active nodes.
[...]

Regards,
G.
--
Guillermo Rodriguez Garcia
<[email protected]>