On 2/28/06, Michele Battelli <[EMAIL PROTECTED]> wrote: > Hi Soo-Hyun, thanks for your reply. I have a few comments: > > Soo-Hyun Choi wrote: > 1. A PhD student at Caltech wrote a patch to seep up a usual > ns-2 simulation. You can look it up > at http://netlab.caltech.edu/~weixl/technical/ns2patch/ns2patch.htm. > I will take a look at this > > 2. As far as my understanding, several hundred nodes with ns-2 might not be > a good choice in terms of CPU time, memory consumption, debugging etc etc. > If your simulation needs to be run with that many nodes, people usually > write his/her own simulator.
> Writing your own code is not the best solution for many reasons (check the > paper "MANET Simulations: The Incredibles" by Stuart Kurkowski et al.), > that's why I would like to stick with ns-2. So far I managed to have a 600 > hundred nodes static wireless network with my own agent, running within > 10seconds. I am trying to reduce more the running time by avoiding useless > ns-2 internal information storage. This issue would be very controversial depending on a simulation study that you are running. And people in networking research community have different views on this. Writing your own code may be worse than using an existing tool (e.g. NS-2), but that does not necessarily mean that you are on the right track just because you use NS-2. Conversely, the fact that you use your own simulator for your study does not mean that you are doing a wrong way of research. So I will leave this issue on your own decision. By the way, the intention of the paper is not about using ns-2 or not, but about a self-developed simulator which we don't (or cannot) trust. > > 3. In general, you would normally need the trace-all trace file in order to > understand/analyse your simulation. Anyway, you should be able to get rid of > it from the script with no error. > One of the way I found to speed up simulation is to avoid ns-2 tracing. In > order to retrieve the data I need I have a dynamic structure in the agent > that stores the data I need. When the agent is done with its work it simply > dumps the values in the structure using a simple "cout". > Unfortunately, removing the line > > $ns trace-all $tracefile > > ends up in a error when the nodes are instanced in the tcl file. I can't > figure out why. > Thank you for further suggestions > > Michele > > > > Soo-Hyun On 2/28/06, Michele Battelli <[EMAIL PROTECTED]> wrote: > Hi all, I am facing the problem of speeding up ns-2 simulations. I have my > own agent and I do simulation on wireless networks with up to 600 nodes. > I don't use any routing agent nor I use traffic generators. Assuming that my > agent is internally well optimized, I tried to reduce the number of "things" > that ns-2 does in order to avoid unnecessary computation. The first thing > was to I avoid the use of GOD. Second, I disabled all the tracing from my > Tcl script. Unfortunately I could not disable the creation of the trace > file, i.e., I could not get rid of the following lines: set tracefd [open > wireless-out.tr w] $ns_ trace-all $tracefd without getting an error. I > would like to know 1) if I can totally disable any internal tracing 2) if > there are other things that I can disable to make ns-2 run faster either in > the C part of the simulator or though Tcl script. Thank you in > advance, Regards Michele Battelli > > >
