readme

thomson_mesh_modules_Enhanced Wireless Mesh Networking for ns-2 simulator

 Network Simulator (ns-2) Modules for Mesh Netoworking (Vivek Mhatre, mhatre at gmail.com).
 DISCLAIMER: This is very much work in progress! Use at your own risk. The modifications were made with ns-2.30 Trying to compile these
 files with other versions may not work.

 Details of implementation in the ACM Sigcomm CCR July 2007 paper "Enhanced Wireless Mesh Networking for ns-2
 simulator".

 Quick ns-2 installation on the blades.
 The ns-2 installation was performed such that all the modules (tcl, tk, otcl, and tclcl) were locally installed in my home in that sequence.
 All the original tar-balls for each of the packages are located /home/mhatre/ See the following files to know the exact ./configure command:

 /home/mhatre/tcl8.4.14/unix/config.status
 /home/mhatre/tk8.4.14/unix/config.status
 /home/mhatre/otcl-1.12/config.status
 /home/mhatre/tclcl-1.18/config.status

 Follow up the "./configure" with a "make" and then "make install"

 The options for configuring ns after installing the above packages can be found in

 /home/mhatre/ns/ns-2.30/config.status

 Grep for "./configure" Then follow it up with "make". Set the following path variables
 in .bashrc or .cshrc so that the new tcl executables and libraries are visible.

 LD_LIBRARY_PATH=/home/mhatre/tcl/lib
 LD_RUN_PATH=/home/mhatre/tcl/lib

 At this point, you should have a basic ns up and running.

 Installing contributed ns-Agents
 Follow instructions given at the following URL and install the NOAH routing agent. It is a static routing package needed for mesh routing.
 NOAH ROuting Agent

 Install the Fading Agent contributed by CMU.
 CMU Fading agent

 Install the corrected version of the above agent.
 Corrected Fading Agent, Rutgers

 Note that any addition such as NOAH or Fading should be followed up with "make depend" and then a "make" for the changes to take effect.

 Now ns should be running with all the external agents installed.

 Modules Added by Vivek
 All the modified .cc, .h and .tcl files in the source directory of ns have a tag VIVEK as a comment, and should be easy to spot.

 Cumulative Interference Model
 During a packet reception, current ns just compares the strength of a newly arriving packet with the current packet, and if the ration of RSSI is bigger than
 capture, it ignores the new packet. However, the model does not take into account the cumulative effect of multiple packets that might arrive during the reception of
 a given packet. This is a major limitation as pointed by Bagrodia etal. A cumulative interference model has now been implemented. We look at interference plus noise
 power. Note that it is necessary to specify noise power in the ns script.

 Files affected:
 ns-2.30/mac/wireless-phy.cc, h
 ns-2.30/mac/mac-802_11.cc, h
 ns-2.30/common/packet.h
 ns-2.30/common/packet-stamp.h

 Combined Shadow-Fading
 Current Fading model works only with Two Ray Propagation Model. I have modified it so that we can use shadowing and fading at the same time. This is needed, because
 shadow-fading combined with path loss is the most accurate model. See Rappaport, Stuber.

 Files affected:
 ns-2.30/mobile/shadowing.cc,h
 ns-2.30/mobile/prop_ricean.cc,h

 However, note that there is still an incomplete task whereby the randomization of shadowing through seeding has not been implemented. Earlier shadowing module had a
 bug whereby it recomputed the shadowing gains during each packet reception. Shadowing gains are NOT time-varying. So depending on the seed, one should generate
 random shadowing gains between nodes, and these should stay fixed during the course of the simulation. A new module needs to be added for this, and without this
 module, NO REPEATABILITY can be ensured. This is part of ongoing work.

 Multi SINR and Multi Rate Links
 Current ns version forces us to use a common rate for all the wireless links in the network (unless you use multiple interfaces). However, in reality, wireless links
 can use multiple (8 for 11a/g) rates. Multi-SINR multi-rate module has been added. There are 8 rates (6, 9, 12, 18, 24, 36, 48, 54 Mbps) and correspondingly, there
 are 8 SINRs. Each link can now operate independently at its own rate depending on its SINR.

 Files affected:
 ns-2.30/mac/mac-802_11.cc, h

 Auto Rate Fallback
 When the channel is time-varying, and there are multiple rates that can be used, one needs an auto rate algorithm to pick the best rate for each link. The original
 ARF (Auto Rate Fallback) has been implemented for this purpose. I also implemented AARF (Adaptive Auro Rate Fallback), but there is a bug in my implementation of
 AARF. For the time begin, ARF is simple and does the job, and so I did not bother to fix AARF code. One can also specify fixed link rate for each link instead of
 using ARF. The description of ARF and AARF can be found in the following paper.

 INRIA Report on AARF (and ARF)

 Files affected:
 ns-2.30/mac/mac-802_11.cc, h

 Framework for measuring ETX and ETT
 Currently, there are no broadcast MAC layer packets that can be generated for periodic probing. At least I couldn't find an easy way to generate them. So I
 implemented the following, so that ETX and ETT metrics can be measured with periodic broadcast and unicast probes.

 The ping agent has been modified to support an option whereby the receiving node does not reply back with an ICMP echo reply. This uses the newly added send-no-echo
 option. See the pingSend function in example scripts. There is a sniffer running on all the nodes which log the rssi of all the successfully decoded packets using
 printf. To avoid tampering with normal ns, separate ns executables called ns_brdcst_sniff and ns_unicst_sniff have been created as follows. Search for ETT_ETX_HACK
 in mac-802_11.cc, then comment out the printfs at three different places. This printfs ensure that the per packet rssi info of the probes (ping send-no-echo pkts)
 are logged. Do a make, and copy the generated ns executable to ns_brdcst_sniff. For performing ETX measurements of a given topology, instead of using ns, use
 ns_brdcst_sniff, and redirect the output to a file using > pipe. Postprocessing of this file gives info about the transmitted probes, and received probes at all the
 nodes.

 ns_brdcst_sniff skeleton.tcl ETX_probes.tcl > output.tr

 Similarly, for ETT, there are three places where the printf corresponding to the tag ETT_ETX_HACK in mac-802_11.cc has to be uncommented. Do a make, and copy the
 generated ns executable to ns_unicst_sniff. For running unicast probes with ARF, to determine the steady state rate and packet loss probability in ETT metric, use
 the ns_unicst_sniff executable instead of ns, and redirect the output to a file using > pipe. Postprocessing of this script gives info about no. of transmitted and
 received probe pkts for each link.
 
 ns_unicst_sniff skeleton.tcl ETT_probes.tcl > output.tr

 To run TCP traffic
 ns skeleton.tcl traffic.tcl

 Example tcl files are in ex/ directory.

 TODO:
 There seems to be a packet size mismatch issue (somewhere I must not have deducted the MAC hdr size). The traces show rcvd pkts larger than the tx pkts by
 MAC hdr size.. Potentially due to the new MAC header settings that I am using for 802.11g.