Re: Randomness in NSClick

Hi,

oops, didn't post to the list last message.

i suppose there is no special need for every node (or click as a whole) 
having it's own rng stream, but it could help conserving reproducability 
when only small changes in the scenario setup are done, for example.
Regarding the sim_click_command(RNG) i think it should be transparently 
done in click_random(), so that would make a difference when not 
implemented in other sims.

Regards, Björn


On 01/24/2012 12:24 PM, Sascha Alexander Jopen wrote:
> Hmm, i do not have any ns2 simulations but there should be something 
> like a global RNG and seed manager as well. Someone who has running 
> ns2 simulations could create the necessary simclick_sim_command 
> extension for ns2. As long as there is no one who can provide this 
> extension, the new command would only be an unknown command, which 
> wouldn't break anything for existing simulations.
>
> Should every click node node have its own rng, or one global rng for 
> every click node in ns3?
>
> Regards,
> Sascha
>
> On 01/24/12 12:08, Björn Lichtblau wrote:
>> Hi,
>>
>> Yes, same thoughts here, if we want reproducability (almost a
>> requirement for simulations) we should use ns3's RNGs, which were
>> designed for that. Is this also possible for ns2?
>> Without thinking about ns2 i would have taken exactly that route,
>> replace the random function in click by a simclick_sim_command(..RNG..)
>> for simulations.
>>
>> Regards, Björn
>>
>>
>> On 01/23/2012 04:00 PM, Sascha Alexander Jopen wrote:
>>> Another approach would be another simclick_sim_command enhancement, to
>>> get a random number from RNGs from NS3.
>>> ...
>>> Having an interface to the simulators RNGs in some way is mandatory, i
>>> think.
>>>
>>> Any opinions or solutions?
>>
>>
>