615 Project 1

Project 1

First, you will need to acquire and install the ns2 sources. You can do so from the ns2 distribution here.

Here are some reference links on ns2:

• ns2 home page.
• ns2 Manual.
• A tutorial on using ns2.
• A shorter tutorial.

Once you have installed and compiled ns2, it's time to run some simulations through it. In the examples below, replace the ns2 version number (2.1b8a) with the version number that came with the ns2 installation you are using.

• Run the following examples to get a rough idea of how ns and nam work:
  • ns ns-allinone-2.1b8a/ns-2.1b8a/tcl/ex/nam-example.tcl
    This is just a demo of nam's capabilities. What happens at time 4 ?

    ns ns-allinone-2.1b8a/ns-2.1b8a/tcl/ex/tg.tcl

    This showcases the different default traffic generation schemes in ns. What are the traffic patterns being generated in this example ?

    ns ns-allinone-2.1b8a/ns-2.1b8a/tcl/ex/mcast.tcl

    This shows the operation of a wired multicast algorithm. What happens at time 1 ?

    ns ns-allinone-2.1b8a/ns-2.1b8a/tcl/ex/wireless-test.tcl

    This simulates a wireless ad hoc network! The resulting trace is in out-test.tr

• Simulate 10 CBR source-sink pairs communicating over DSR, using a field size of 1000x1000, 250m radius, random waypoint mobility and constant bit rate sources. Measure the simulation time for 10, 30, 50, 100, 150, and 200 nodes. You can use the Unix time command for measurements. How does ns2 scale with increasing numbers of nodes ?

• The DSR paper examined a field size of 1500x300m, with 250m radius. First, duplicate one of their protocol comparison graphs. Next, change the field size to 1500x1500 (while also increasing the number of nodes to keep density constant) so the network is no longer linear and caching is less likely to be useful. Is there a qualitative difference in the results ? Some people have noticed that a field size of 1500x1500 requires 250 nodes, at which point ns2 runs too slowly to collect any measurements (as shown by your measurements from the previous question). So, you can scale the field size down (say, to 1000x1000m or 750x750m), which should make the simulation run much faster and hopefully complete before the deadline!

• Set up a linear, static (i.e. no mobility) network of six nodes, connected together over 802.11b. That is, A can talk to B can talk to C can talk to D can talk to E can talk to F, and there are no other links in the network. Set up a constant-bit-rate (CBR) source on node B, with a corresponding sink on E. Set up another CBR source-sink pair between A and F. Run the system for a sufficiently long time to determine statistically significant usage measurements, and compare the bandwidth achieved by the B-E flow to the bandwidth achieved by the A-F flow. Is there a substantial difference between the bandwidths achieved by the two flows ? Are they efficiently using the maximum available bandwidth available under 802.11b ? Describe why your findings make sense. (Hint: think about the consequences of the RTS/CTS protocol used by the 802.11 MAC layer protocol).

• Set up a simulation, using a field size of 1000x1000, 250m radius, random waypoint mobility, 802.11b wireless physical layer, 30 nodes, using TCP for all communication, and {DSR, AODV and TORA} underneath. The nodes repeatedly engage in TCP transactions (at a reasonable rate, e.g. 10 per second), where one node initiates a TCP connection to another node and sends {1KB, 6KB, 30KB, 60KB} of data. Determine the mean latency and bandwidth as a function of routing protocol and transaction size. Does the interaction of any of the routing protocols with TCP cause any anomalies ?

You should now be a fairly decent ns2 user.