[ns] Fast simulations

[Debojyoti Dutta <ddutta@ISI.EDU>]

Hi Huan

We are in the process of adding features to ns that can address the issues
that you raised. 

Proceeding from where John left: Suppose you had a very large network with
lots of traffic sources and sinks. You want to run the simulations for
4000 seconds. One reason why you may want to run that long would be to
find the steady state and find out whether that scenario is interesting
for you. That could be done using analytical techniques and would be much
faster as the speed of simulations is of the order of the number of
packets generated (or events). 

We are working on an architecture that will (within NS) do a quick
but approximate analytical simulator that will give you the status of the
network in an user friendly fashion. This work is nto available for
release as yet. 

If you have any further questions, please let me know. I will try to
answer them ....

Regards
Debo

-----------------------------------------------------------



Hi NS users,

I have searched through NS mailing-list archieve and have
found a similar question posted a year ago, but I failed to
find an answer for that question. So I re-post this question
again, hopefully someone can give me a hint.

http://www.isi.edu/nsnam/archive/ns-users/webarch/2000/msg03604.html

http://www.isi.edu/nsnam/archive/ns-users/webarch/2000/msg03609.html

As I understand, because NS implements a packet-level
simulation, the time it takes to run the simulation, and the
computer power required is proportional to the number of
packets generated. If I want to simulate a big network (many
links with large capacity), does it pose a scalability
problem for NS ?

Let consider the actual example:

I have just implemented a small simulation to simulate WEB,
Audio, and FTP traffic going between Workstation and Node
n0, The topology is as following:


# a Lan consists of 20 workstations
# a 10Mbits Switch
# a 10Mbits link to outside
#
#
#           n2
#         \  |
#          \ |
#     ... -- n1(Switch) --------- n0
#          / |
#         /  |
#            n21
#

If the link utilisation for n1-n0 is 50%, i.e. throughput =
5Mbits equivalent to 500 packets /second  (each packet of
about 1Kbytes ~ 10 Kbits). If I run the simulation for 100
seconds, since each packet travel through 2 links, NS has to
handle a total of 2 * 500 * 100 = 10^5 packets .

The actual time it takes to run on my machine (Pentium III,
700MHz, 256M RAM), is 2 minutes

Now if I want to have a network of 50 routers fully meshed,
with links of 10Mbits , and link utilisation = 50% , and run
it for 4000 seconds.

Total number of links ~ 50 ^ 2 = 2500
Total number of packets generated = 500 * 2500 * 4000 = 5 *
10^9 (packets).
Expected time to run the simulations = 2(min.)  * 5 * 10^9
/ (10^5) = 10^5 minutes = 600 (days)

In other words, it is just imposible to run a simulations
that generate many traffic. Does anyone have any idea about
this.  Pls reply to my address, (and cc NS mailing list if
you want to do so). Thank you very much.