result.txt~

This is a simulator written in Tcl to simulate a network node carrying GSM and GPRS traffics with Qo

May 14th:
EB settings: equal to the CIR of each type
set cir_VoIP   30000
set cir_video  128000
set cir_http   100000
set cir_ftp    200000

Simulation time: 3600s

Arrived session #: 673
Blocked session #: 507
Core queue: GGSN->SGSN link Pkt Statistics
 
Packets Statistics
=======================================
 CP  TotPkts   TxPkts   ldrops   edrops
 --  -------   ------   ------   ------
All    15198    15178       20        0
 30    15147    15147        0        0
 31       51       31       20        0


Result may 18th:
# settings: 
set cir_VoIP   30000
set cir_video  128000
set cir_http   100000
set cir_ftp    200000

set EB_VoIP  [expr $cir_VoIP / 2.0]
set EB_video [expr $cir_video / 2.0] 
set EB_http  12000  
set EB_ftp   12000 

# traffic model
In this experiment, only Voip, video and ftp was created. total load = 15% of bottleneck link BW.

# Simulation time: 1200s

# simulation output: 
Arrived session #: 377
Blocked session #: 63
Core queue: GGSN->SGSN link Pkt Statistics
 
Packets Statistics
=======================================
 CP  TotPkts   TxPkts   ldrops   edrops
 --  -------   ------   ------   ------
All    87958    87949        9        0
  0     8719     8719        0        0
 20    27550    27550        0        0
 21    29217    29217        0        0
 30    22456    22456        0        0
 31       16        7        9        0

Analysis:
In the output of May 14, all the traffic other than voip was blocked. It's due to 
the BHCA of voip occupied  75% of total session arrival, though its average load is 
just totally 33%. So the EB setting of voip is crucial for the total available BW. 
A less conservative setting of each EB will result in a much less blocking rate as in the 
result from May 18th.  

Next, extract the delay of RT in the same settings, later increase the traffic load.