Examples of Performance Analysis using NS

来源：互联网发布：长沙招聘网络推广编辑：程序博客网时间：2024/05/17 03:11

Performance Analysis of Drop Tail Queues(the NS script is here:click here )

Constructing the Simulation Network

The classic performance analysis network:

Link 0-1 is the bottleneck link
N flows share the bottleneck link

NS script to construct this network:

# ##################################################################### Create bottleneck and dest nodes  set n2 [$ns node]  set n3 [$ns node]# ##################################################################### Create bottleneck link (between bottleneck nodes n2-n3)  $ns duplex-link $n2 $n3 0.7Mb 20ms DropTail# ##################################################################### Set Queue Size of bottleneck link (n2-n3)  $ns queue-limit $n2 $n3 100# ##################################################################### Set number of sources  set NumbSrc 3# ##################################################################### Use a for-loop to create $NumbSrc source nodes  for {set j 1} {$j<=$NumbSrc} { incr j } {     set S($j) [$ns node]  }# ##################################################################### This part is used to start the sources at random times# ##################################################################### Create a random generator for starting the ftp and# for bottleneck link delays  set rng [new RNG]  $rng seed 2# ##################################################################### parameters for random variables for begenning of ftp connections  set RVstart [new RandomVariable/Uniform]  $RVstart set min_ 0  $RVstart set max_ 7  $RVstart use-rng $rng# ##################################################################### Define RANDOM starting times for each connection# And set (possible random) delay for connection from sources to Node 0  for {set i 1} {$i<=$NumbSrc} { incr i } {     set startT($i)  [expr [$RVstart value]]     set dly($i) 1     puts "startT($i)  $startT($i) sec"     puts $param "startT($i)  $startT($i) sec"  }# ##################################################################### Create links between source and bottleneck#  1. Select the delay from each source to bottleneck node#  2. Set queue size (#packets)  for {set j 1} {$j<=$NumbSrc} { incr j } {     $ns duplex-link $S($j) $n2 10Mb $dly($j)ms DropTail     $ns queue-limit $S($j) $n2 20  }# ##################################################################### Orient the links  $ns duplex-link-op $n2 $n3 orient right  $ns duplex-link-op $S(1) $n2 orient right-down  $ns duplex-link-op $S(2) $n2 orient right  $ns duplex-link-op $S(3) $n2 orient right-up# ##################################################################### Create TCP Sources  for {set j 1} {$j<=$NumbSrc} { incr j } {     set tcp_src($j) [new Agent/TCP/Reno]     $tcp_src($j) set window_ 8000  }# ##################################################################### Color the packets  $tcp_src(1) set fid_ 1  $ns color 1 red  $tcp_src(2) set fid_ 2  $ns color 2 yellow  $tcp_src(3) set fid_ 3  $ns color 3 blue# ##################################################################### Create TCP Destinations  for {set j 1} {$j<=$NumbSrc} { incr j } {     set tcp_snk($j) [new Agent/TCPSink]  }# ##################################################################### Schedule START events for the FTP agents:  for {set i 1} {$i<=$NumbSrc} { incr i } {     $ns at $startT($i) "$ftp($i) start"     $ns at $SimDuration "$ftp($i) stop"  }

Obtaining TCP CWND Window information (in general)

This is done through the classic "plotWindow" procedure that reschedules itself.

This version of "plotWindow" can be used for multiple TCP sources:

# ##################################################################### plotWindow(tcpSource file k): Write CWND of k tcpSources in file  proc plotWindow {tcpSource file k} {     global ns NumbSrc     set time 0.03     set now [$ns now]     set cwnd [$tcpSource set cwnd_]     if {$k == 1} {        puts -nonewline $file "$now \t $cwnd \t"     } else {     if {$k < $NumbSrc } {        puts -nonewline $file "$cwnd \t"        }     }     if { $k == $NumbSrc } {        puts -nonewline $file "$cwnd \n"     }     if { $k == $NumbSrc } {        puts -nonewline $file "$cwnd \n"     }     $ns at [expr $now+$time] "plotWindow $tcpSource $file $k"  }Usage:# ##################################################################### Start plotWindow() for all tcp sources  for {set j 1} {$j<=$NumbSrc} { incr j } {     $ns at 0.1 "plotWindow $tcp_src($j) $windowVsTime $j"  }

The output format is as follows:

  TIME   Win_flow1   Win_flow2   Win_flow3 ....   Win_flowN

Use the following plot commands in gnuplot toplot the window trends for each TCP flow:

  plot "windowFile" using 1:2 t "Flow 1" w lines 1, \       "windowFile" using 1:3 t "Flow 2" w lines 2, \       "windowFile" using 1:4 t "Flow 3" w lines 3

Obtaining Queue Length information (in general)

(Current) Queue length information can be obtained byqueue monitoring - see: click here

Code in NS script to obtain queue length information:

# ##################################################################### Monitor avg queue length of link ($n2,$n3)  set qfile [$ns monitor-queue $n2 $n3  [open queue.tr w] 0.05]  [$ns link $n2 $n3] queue-sample-timeout;

Process output with:

 plot "queue.tr" using 1:5 t "Queue Length" w lines 1

Obtaining Throughput (Goodput) information from a TCP connection(Perl program: click here)

Drop Tail queues does not provide built-in tracing facilityto recorded average and current queue length information(RED queues do)
The trace information (ns trace-all - click here) containsenough information to determine the TCP (good) throughputof any TCP connection.
Recall the structure of the trace record:
The Seq Number starts with 0 when simulation begins
Therefore, we can determine the (good) throughput of a TCPconnection by looking at theSeq Number received by the destination node of the TCP connection !

Here is a PERL script that can be used to generatea plot of the throughput information of a TCP connection: (you can get the Perl script here:click here)

NOTES:

@x = split(' ') tokenizeseach input line into an array of strings
x[0] is the first element (i.e., event), we look for "r" (receive)events
x[1] is the first element (event)

# ##################################################################### Usage: ##   perl throughput     #                    ## Example: plot the throughput of connection 2.0 - 1.0 every 0.5 sec##perl throughput  out.tr  1 2.0 1.0  0.5# ##################################################################### ############################################################## Get input file (first arg)   $infile=$ARGV[0];# ############################################################## Get node that receives packets (second arg)   $destnode=$ARGV[1];   $fromport=$ARGV[2];   $toport=$ARGV[3];# ############################################################## Get time granularity (time interval width   $granularity=$ARGV[4];# ###################################################################### We compute how many bytes were transmitted during time interval # specified by granularity parameter in seconds   $sum=0;   $grantsum=0;   $clock=0;# ###################################################################### Open input file   open (DATA,"<$infile") || die "Can't open $infile $!";     while ()    {# Tokenize line using space as separator      @x = split(' ');# column 1 is time: Check for next epoch      if ($x[1]-$clock <= $granularity)      {# checking if the event (column 0) corresponds to a reception          if ($x[0] eq 'r')          { # checking if the destination (column 3) corresponds to node in arg 1            if ($x[3] eq $destnode && $x[8] eq $fromport && $x[9] eq $toport)             { #checking if the packet type is TCP               if ($x[4] eq 'tcp')                {                  $sum=$sum+$x[5];  $grantsum=$grantsum+$x[5];               }            }         }      }      else# One interval has passed, compute throughput      {            $throughput=0.000008*$sum/$granularity;         print STDOUT "$clock $throughput\n";         $clock=$clock+$granularity; if ($x[0] eq 'r' && $x[3] eq $destnode && $x[8] eq $fromport      && $x[9] eq $toport && $x[4] eq 'tcp')         {    $sum=$x[5];    $grantsum=$grantsum+$x[5]; } else         {    $sum=0; }         while ($x[1]-$clock > $granularity)         {            print STDOUT "$clock 0.0\n";            $clock=$clock+$granularity;         }      }      }   $throughput=0.000008*$sum/$granularity;   print STDOUT "$clock $throughput\n";   $clock=$clock+$granularity;   print "Avg throughput $fromport - $toport = ", $grantsum/$clock, "\n";   close DATA;   exit(0);

How to perform throughput analysis
- Look at the NAM network and find the destination node(in our example, the TCP sinks are attached to node 1
- Look inside the trace file and find some like with destination = 1.?
  Example:
```
r 0.005749 2 0 tcp 40 ------- 1 2.0 1.0 0 0...r 0.026206 0 1 tcp 40 ------- 1 2.0 1.0 0 0  
```
  - The first packet is in transit at node 0 towards the finaldestination 1 (received on port 0).
  - The received packet is received by node 1 and it is alsothe final destination node - so this packet counts towardsthe (good) throughput)
  - Write down the src-node.src-port and dest-node.dest.port:
```
        2.0    1.0 
```
    You need these as input parameters for the "throughput" script.
- To compute the throughput of the TCP connection (2.0 -> 1.0) over 0.5 secinterval, use:
```
     throughput trace-file  1  2.0  1.0  0.5 > plotdata  
```
- You will get a two column output:
```
     Time   Throughput (in MBytes/sec) 
```
  that you can use gnuplot to plot the trend:

A more space-efficient throughput trace-file

If all you want is to study the throughput of TCP connections,you can save a lot of disk space by filtering the trace recordsinside the NS script (see:click here)

Here is a PERL script that only write the RECEIVED packetsby a given node out:

(The Perl script is here: click here )

# ######################################################################### throughput-filter: PERL utility for reducing the trace-file size ## ## Usage:  ## ##   When you open the trace-file in NS, use this command: ## ##      set file1 [open "| perl thruput-filter destNode > trace-file" w]  ## ######################################################################### ############################################################## Get node that receives packets    $tonode=$ARGV[0];# ######################################################################### <STDIN> is the standard input stream, already opened !     while (<STDIN>) {# Tokenize line using space as separator      $line=$_;      @x = split(' ');# checking if the event (column 0) corresponds to a reception       if ($x[0] eq 'r')      {# checking if the destination corresponds to node in arg         if ($x[3] eq $tonode)          {     print STDOUT $line         }      }   }   exit(0);

To use this filter, open the trace-file in your NS script using:

  set file1 [open "| perl thruput-filter destNode > trace-file" w]

Performance Analysis of RED Queues(the NS script is here:click here )

The DropTail simulation script is modified slightly to performsimulation on RED queues, as follows:

First, change the bottleneck queue (n2,n3) into a RED queue:

# ##################################################################### Create RED links between bottleneck nodes  $ns duplex-link $n2 $n3 0.7Mb 20ms RED

The RED queue component does provide a "trace" feature thatyou can activate to see the average and current queue length:

# ##################################################################### Monitor avg queue length of RED link ($n2,$n3) in file "red-queue.tr"  set traceq [open red-queue.tr w]- open a trace file for RED  set redq [[$ns link $n2 $n3] queue]- get the RED queue  $redq trace curq_    - Tell the RED queue to trace changes       to its variable curq_ (current Q len)  $redq trace ave_    - Tell the RED queue to trace changes       to its variable ave_ (average Q len)  $redq attach $traceq    - Send the trace data to file "traceq"

The output of the "traceq" file will look like this:

      Trace      code  Time       Value (unit = #packets)    --------------------------------------a   0.106137   0.00569799Q   0.106137   1<--- first packet arrivesa   0.148626   0.00556923Q   0.148626   0<--- first packet departeda   0.1491     0.00553749a   0.155392   0.00550594a   0.155865   0.0111726Q   0.155865   1<--- second packet arrives....

NOTE: The default buffer unit in RED queue is #packets

You can change it to operate in "byte" mode: click here

To obtain a plot of average and current queue length of the RED queue:
- Use "grep a REDtrace-file" and "grep Q REDtrace-file"to separate the different inputs lines:
```
   grep a red-queue.tr > ave.tr   grep Q red-queue.tr > cur.tr 
```
- Use the following command in gnuplot to plot columns 2 and 3from the input file:
```
  plot "ave.tr" using 2:3 t "Average Qlength" w lines 1, \       "cur.tr" using 2:3 t "Current Qlength" w lines 2 
```

Plotting Performance Graphs with GNUPLOT

Some GNUPLOT documents
- Quick intro: click here
- Length manual: click here
Basic plot command
```
 plot "data" using 1:2 title "Flow 1" with lines 1 
```
Make a line plot with color 1 using columns 1 and 2 from file "data"
```
 plot "data" using 1:2 title "Flow 1" with lines 1 \       "data" using 1:3 t "Flow 2" w lines 2
```
Make a plot containing 2 line plots. First, a line plot withcolor 1 using columns 1 and 2 from file "data" and second with color 2 using columns 1 and 3 from file "data"
GNUPLOT command files
- You can execute prepared GNUPLOT command files with the "load"command:
```
 load "gnuplot-script"  
```
  Execute the stored commands inside the file "gnuplot-script"
- Example, stored the previous "plot" commands in a file anduse "load" to run them
Postscript Output
```
 set out "plot.ps" set size 0.7, 0.7 set terminal postscript portrait color   replot set terminal x11 set size 1,1 
```
These commands first changes the output to the file "plot.ps. Then changes the size and the terminal to a "color postscript device" and replot the figure (the figure will now be saved as a postscript file).
The last 2 command reset the terminal back to the X-window display and resize to normal.
```
 set out "plot.ps" set size 0.7, 0.7 set terminal postscript portrait mono  replot set terminal x11 set size 1,1
```
Same, now with mono-color
- Save as Postscript GNUPLOT scripts
  - Save as color postscript script: click here
  - Save as mono postscript script: click here
  - Usage:
```
     set out "plot.ps"     load "gp-color-ps"    or load "gp-mono-ps" 
```

Example of TCP Performance Analysis

Drop Tail
- Current queue length for Drop Tail queue:(see: click here )
- TCP Congestion Window: (see: click here )
- Avg throughput (over 0.05 sec) for Drop Tail queue:(see: click here )
  
  Average throughputs:
  - Flow 1 = 0.35 MBytes/sec
  - Flow 2 = 0.17 MBytes/sec
  - Flow 3 = 0.16 MBytes/sec
RED Queue
- Current and average queue length for RED queue:(see: click here )
- TCP Congestion Window: (see: click here )
- Avg throughput (over 0.05 sec) for Drop Tail queue:(see: click here )
  
  Average throughputs:
  - Flow 1 = 0.267 MBytes/sec
  - Flow 2 = 0.183 MBytes/sec
  - Flow 3 = 0.241 MBytes/sec