Wireless Simulation|NETWORK SIMULATIONS: NS-2
Wireless Simulation
When simulating wireless networks, there are no links; all the configuration work goes into setting up the nodes, the traffic and the wireless behavior itself. Wireless nodes have multiple wireless-specific attributes, such as the antenna type and radio-propagation model. Nodes are also now in charge of packet queuing; before this was the responsibility of the links. Finally, nodes have coordinates for position and, if mobility is introduced, velocities.
For wired links the user must set the bandwidth and delay. For wireless, these are both generally provided by the wireless model. Propagation delay is simply the distance divided by the speed of light. Bandwidth is usually built in to the particular wireless model chosen; for the Mac/802_11 model, it is available in
An Introduction to Computer Networks, Release 2.0.4
attribute dataRate_ (which can be set). To find the current value, one can print [Mac/802_11 set dataRate_]; in ns-2 version 2.35 it is 1mb.
Ad hoc wireless networks must also be configured with a routing protocol, so that paths may be found from one node to another. We looked briefly at DSDV in 13.4.1 DSDV; there are many others.
The maximum range of a node is determined by its power level; this can be set with node-config below (using the txPower attribute), but the default is often used. In the ns-2 source code, in file wireless-phy.cc, the variable Pt_ – for transmitter power – is declared; the default value of 0.28183815 translates to a physical range of 250 meters using the appropriate radio-attenuation model.
We create a simulation here in which one node (mover) moves horizontally above a sequence of fixedposition nodes (stored in the Tcl array rownodes). The leftmost fixed-position node transmits continuously to the mover node; as the mover node progresses, packets must be routed through other fixed-position nodes. The fixed-position nodes here are 200 m apart, and the mover node is 150 m above their line; this means that the mover reaches the edge of the range of the ith rownode when it is directly above the i+1th rownode.
We use Ad hoc On-demand Distance Vector (AODV) as the routing protocol. When the mover moves out of range of one fixed-position node, AODV finds a new route (which will be via the next fixed-position node) quite quickly; we return to this below. DSDV (13.4.1 DSDV) is much slower, which leads to many packet losses until the new route is discovered. Of course, whether a given routing mechanism is fast enough depends very much on the speed of the mover; the simulation here does not perform nearly as well if the time is set to 10 seconds rather than 100 as the mover moves too fast even for AODV to keep up.
Because there are so many configuration parameters, to keep them together we adopt the common convention of making them all attributes of a single Tcl object, named opt.
We list the simulation file itself in pieces, with annotation; the complete file is at wireless.tcl. We begin with the options.
| # ====================================================================== # Define options # ====================================================================== set opt(chan) Channel/WirelessChannel ;# channel type set opt(prop) Propagation/TwoRayGround ;# radio-propagation model set opt(netif) Phy/WirelessPhy ;# network interface type set opt(mac) Mac/802_11 ;# MAC type set opt(ifq) Queue/DropTail/PriQueue ;# interface queue type set opt(ll) LL ;# link layer type set opt(ant) Antenna/OmniAntenna ;# antenna model set opt(ifqlen) 50 ;# max packet in ifq set opt(bottomrow) 5 ;# number of bottom-row nodes set opt(spacing) 200 ;# spacing between bottomãÑrow nodes set opt(mheight) 150 ;# height of moving node ãÑabove bottom-row nodes set opt(brheight) 50 ;# height of bottom-row ãÑnodes from bottom edge set opt(x) [expr ($opt(bottomrow)-1)*$opt(spacing)+1] ;# x ãÑcoordinate of topology set opt(y) 300 ;# y coordinate of topology |
An Introduction to Computer Networks, Release 2.0.4
|
set opt(adhocRouting) AODV ;# routing protocol # the next value is the speed in meters/sec to move across the field |
The Channel/WirelessChannel class represents the physical terrestrial wireless medium; there is also a Channel/Sat class for satellite radio. The Propagation/TwoRayGround is a particular radiopropagation model. The TwoRayGround model takes into account ground reflection; for larger inter-node distances d, the received power level is proportional to 1/d4 . Other models are the free-space model (in which received power at distance d is proportional to 1/d2 ) and the shadowing model, which takes into account other types of interference. Further details can be found in the Radio Propagation Models chapter of the ns-2 manual.
The Phy/WirelessPhy class specifies the standard wireless-node interface to the network; alternatives include Phy/WirelessPhyExt with additional options and a satellite-specific Phy/Sat. The Mac/ 802_11 class specifies IEEE 802.11 (that is, Wi-Fi) behavior; other options cover things like generic CSMA/CA, Aloha, and satellite. The Queue/DropTail/PriQueue class specifies the queuing behavior of each node; the opt(ifqlen) value determines the maximum queue length and so corresponds to the queue-limit value for wired links. The LL class, for Link Layer, defines things like the behavior of ARP on the network.
The Antenna/OmniAntenna class defines a standard omnidirectional antenna. There are many kinds of directional antennas in the real world – eg parabolic dishes and waveguide “cantennas” – and a few have been implemented as ns-2 add-ons.
The next values are specific to our particular layout. The opt(bottomrow) value determines the number of fixed-position nodes in the simulation. The spacing between adjacent bottom-row nodes is opt(spacing) meters. The moving node mover moves at height 150 meters above this fixed row. When mover is directly above a fixed node, it is thus at distance ?(2002 + 1502 ) = 250 from the previous fixed node, at which point the previous node is out of range. The fixed row itself is 50 meters above the bottom of the topology. The opt(x) and opt(y) values are the dimensions of the simulation, in meters; the number of bottom-row nodes and their spacing determine opt(x).
As mentioned earlier, we use the AODV routing mechanism. When the mover node moves out of range of the bottom-row node that it is currently in contact with, AODV receives notice of the failed transmission from the Wi-Fi link layer (ultimately this news originates from the absence of the Wi-Fi link-layer ACK). This triggers an immediate search for a new route, which typically takes less than 50 ms to complete. The earlier DSDV (13.4.1 DSDV) mechanism does not use Wi-Fi link-layer feedback and so does not look for a new route until the next regularly scheduled round of distance-vector announcements, which might be several seconds away. Other routing mechanisms include TORA, PUMA, and OLSR.
The finishing time opt(finish) also represents the time the moving node takes to move across all the bottom-row nodes; the necessary speed is calculated in opt(speed). If the finishing time is reduced, the mover speed increases, and so the routing mechanism has less time to find updated routes.
The next section of Tcl code sets up general bookkeeping:
| # create the simulator object set ns [new Simulator] |
An Introduction to Computer Networks, Release 2.0.4
|
# set up tracing # create and define the topography object and layout $topo load_flatgrid $opt(x) $opt(y) # create an instance of General Operations Director, which keeps track of ãÑnodes and # node-to-node reachability. The parameter is the total number of nodes in ãÑthe simulation. create-god [expr $opt(bottomrow) + 1] |
The use-newtrace option enables a different tracing mechanism, in which each attribute except the first is prefixed by an identifying tag, so that parsing is no longer position-dependent. We look at an example below.
Note the special option namtrace-all-wireless for tracing for nam, and the dimension parameters opt(x) and opt(y). The next step is to create a Topography object to hold the layout (still to be determined). Finally, we create a General Operations Director, which holds information about the layout not necessarily available to any node.
The next step is to call node-config, which passes many of the opt() parameters to ns and which influences future node creation:
|
# general node configuration $ns node-config -adhocRouting $opt(adhocRouting) \ |
Finally we create our nodes. The bottom-row nodes are created within a Tcl for-loop, and are stored in a Tcl array rownode(). For each node we set its coordinates (X_, Y_ and Z_); it is at this point that the
An Introduction to Computer Networks, Release 2.0.4
|
# create the bottom-row nodes as a node array $rownode(), and the moving node ãÑas $mover for {set i 0} {$i < $opt(bottomrow)} {incr i} { } set mover [$ns node] |
We now make the mover node move, using setdest. If the node reaches the destination supplied in setdest, it stops, but it is also possible to change its direction at later times using additional setdest calls, if a zig-zag path is desired. Various external utilities are available to create a file of Tcl commands to create a large number of nodes each with a designated motion; such a file can then be imported into the main Tcl file.
| set moverdestX [expr $opt(x) - 1] $ns at 0 "$mover setdest $moverdestX [$mover set Y_] $opt(speed)" |
Next we create a UDP agent and a CBR (Constant Bit Rate) application, and set up a connection from rownode(0) to mover. CBR traffic does not use sliding windows.
|
# setup UDP connection, using CBR traffic set udp [new Agent/UDP] |
The remainder of the Tcl file includes additional bookkeeping for nam, a finish{} procedure, and the startup of the simulation.
|
# tell nam the initial node position (taken from node attributes) for {set i 0} {$i < $opt(bottomrow)} {incr i} { |
An Introduction to Computer Networks, Release 2.0.4
|
} $ns initial_node_pos $mover 20
proc finish {} { global ns tracefd namtrace } # begin simulation $ns run |
The simulation can be viewed from the nam file, available at wireless.nam. In the simulation, the mover node moves across the topography, over the bottom-row nodes. The CBR traffic reaches mover from rownode(0) first directly, then via rownode(1), then via rownode(1) and rownode(2), etc. The motion of the mover node is best seen by speeding up the animation frame rate using the nam control for this, though doing this means that aliasing effects often make the CBR traffic appear to be moving in the opposite direction.
Above is one frame from the animation, with the mover node is almost (but not quite) directly over rownode(3), and so is close to losing contact with rownode(2). Two CBR packets can be seen en route; one has almost reached rownode(2) and one is about a third of the way from rownode(2) up to the blue mover node. The packets are not shown to scale; see exercise 17.
The tracefile is specific to wireless networking, and even without the use of use-newtrace has a rather different format from the link-based simulations earlier. The newtrace format begins with a letter for
An Introduction to Computer Networks, Release 2.0.4
send/receive/drop/forward; after that, each logged attribute is identified with a prefixed tag rather than by position. Full details can be found in the ns-2 manual. Here is an edited record of the first packet drop (the initial d indicates a drop-event record):
| d -t 22.586212333 -Hs 0 -Hd 5 ... -Nl RTR -Nw CBK ... -Ii 1100 ... -Pn cbr - ãÑPi 1100 ... |
The -t tag indicates the time. The -Hs and -Hd tags indicate the source and destination, respectively. The -Nl tag indicates the “level” (RouTeR) at which the loss was logged, and the -Nw tag indicates the cause: CBK, for “CallBacK”, means that the packet loss was detected at the link layer but the information was passed up to the routing layer. The -Ii tag is the packet’s unique serial number, and the P tags supply information about the constant-bit-rate agent.
We can use the tracefile to find clusters of drops beginning at times 22.586, 47.575, 72.707 and 97.540, corresponding roughly to the times when the route used to reach the $mover node shifts to passing through one more bottom-row node. Between t=72.707 and t=97.540 there are several other somewhat more mysterious clusters of drops; some of these clusters may be related to ordinary queue overflow but others may reflect decreasing reliability of the forwarding mechanism as the path grows longer.
Frequently Asked Questions
Recommended Posts:
- Uses of Computer Networks
- Network Hardware
- Installing and Running ns-3|THE NS-3 NETWORK SIMULATOR
- A Single TCP Sender| The ns-3 Network Simulator
- Wireless|The ns-3 Network Simulator
- IPsec |Public-Key Encryption
- THE THEORETICAL BASIS FOR DATA COMMUNICATION
- GUIDED TRANSMISSION MEDIA
- WIRELESS TRANSMISSION
- COMMUNICATION SATELLITES
- DIGITAL MODULATION AND MULTIPLEXING
- THE PUBLIC SWITCHED TELEPHONE NETWORK
- THE MOBILE TELEPHONE SYSTEM
- CABLE TELEVISION
- THE DATA LINK LAYER