Mirai

Yesterday, 13th April I had a meeting with my supervisor. I showed what I had done for the last 1 month. We disscused some problem regarding the simulation using the HS-DSCH transport channel.

So at the end of the meeting, he introduced me a new discrete network simulator called mirai developed by Wireless Applications Group on the Next Generation Mobile Network Project, National Institute of Information and Communications Technology (NICT).

So my simulation DiffServ in EURANE core network using HS-DSCH is still pending until I find a solution.

Do you have Path Problems when running Java J2SE 5.0 on PCs. Well click here!

NS-2 Link

Click Here
Click here too

what to do...

Things that I have achieved so far:-
-----------------------------------
1. A simple simulation on DCH channel with DiffServ in the core network. Two UDP sources with constant bit rate traffic are to simulate the VoIP data packets. Throughput at the ssgn which is configure as an edge router are measured in term of Mbps.

Things that I did:-
-------------------
1. Read the Seacorn deliverable D3.2v2.
2. Read thesis on QoS provisioning.
3. Read TS 23.207


Things to do:-
--------------
1.Read 3gpp technical document on End-to-end QoS and HS-DSCH.
2.Find out why simulation on HS-DSCH channel with DiffServ in the core network stop abruptly. It should not have any problem because Diffserv is working on the network layer. Mac-Hs is the important mechanism for HS-DSCH which is on the data link layer.
3. Find some script to analyse the trace file for end-to-end delay and jitter.

Things to do as on 20th March
-----------------------------
1.Read and understand
2.Build some simple simulation scenarios.
Check with 3gpp standard for QoS

Max HSDPA Throughput in EURANE = 1.3Mbps

I think have similar observation of the throughput



Greetings,

I have simulated an HSDPA cell with two UEs but only one active CBR source sending at 4Mbps. The input trace file is the same as the idealtrace but I changed all CQI values to 29 in order to achieve maximum throughput. However, the throughput was saturating at only 1.3Mbps !!
Can someone clarify why? There might be a configuration parameter limiting the number of codes or coding rate??


regards.

AMM

source

TCP Window

The variable "window_" in ns2 corresponds to "awnd" in the literature.
Using this variable you can set an upper bound for the TCP window at the
source. So, if you wish to "set the TCP maximum windows size to 50",
just
set window_ to 50 in your tcl script.
Note that "window_" is binded as "wnd_" in the c++ code
(tcp/tcp.{h,cc}).


source

Had some problem in calculating the throughput from the trace file. Still searching the method to implement flowmon.

I came across posting where it explain that TCP adjust its sending rate when there is a congestion in the link.


TCP adjusts its sending rate using a congestion control algorithm

Take a look at the following paper, it could help you

Van Jacobson, “Congestion Avoidance and Control,” in Proceedings of ACM
Sigcomm’88, Stanford CA, August 1988, 314-329.

Best Regards

Alfredo

---------------------------------------------
Luigi Alfredo Grieco, Ph. D
Assistant Professor
DEE Politecnico di Bari,
Via Orabona 4 - 70125 Bari, Italy
Ph. +39-080-5963301
Fax +39-080-5963410
http://www-ictserv.poliba.it/grieco


-----Messaggio originale-----
Da: ns-users-bounces at ISI.EDU [mailto:ns-users-bounces at ISI.EDU] Per conto
di mehri rajayi
Inviato: martedì 12 aprile 2005 11.11
A: ns ns
Oggetto: [ns] how create tcp connection with spesific rate?



i want to create a tcp connection with ftp traafic sourse, with this
parameter (rate:128Kbps pktsize:1000byte ) but tcp hasn't rate
parameter. I set packetSize_ to 1000 but in trace file PKTsize is equal
40. can you help me.


-source-

Link

Met with the supervisor on 16th March. Wrote a progress report to submit to the supervisor. The report mentioned some literature reviews and preliminary simulation works. Supervisor kind of agreed with the idea and ask to come out with some simulations for the next meeting.

Things to do:
Read and understand.
Build some simple simulation scenarios.
Check with 3gpp standard for QoS


Link

RED again..

I think I solved the problem. Actually it is not a problem.

I think in ns2.28 the paramater "Queue/RED set queue-in-bytes_ true" is set to true by default. In ns2.27, the parameter is set to false by default.

I want the queue size in term of packets. To over ride this just add the following parameter in red.tcl script when using ns228.

"Queue/RED set queue-in-bytes_ false"

I think this is also apply for ns229.

Any reason why we want to monitor the queue in bytes rather than in packest?

Windows reads linux partition...

Click here for the website

Why use simulator ?

Simulation is often based on random events.

Therefore, if you want a real time implementation, then if 2 events are
seperated by minutes, then you would just have to wait. Ns-2 is
event-oriented, and therefore if there is nothing to do between 2
events, then the clock is incremented, and you do not have to wait.

In the other hand, suppose you have a large simulation with a lot of
users. In case a complex treatment has to be done at each user's
equipment, then the computer would have to do it all, with NO real-time
guarantee !

Thirdly, ns-2 is a supported by a whole community, and freely available;
if you can find a real-time simulator please let me know, but I do not
see any.

-source-

VoIP...

Since many VoIP codecs send data at the constant rate, you can use normal UDP/CBR application to simulate VoIP. You have to set appropriate parameters so that the rate and the packet size correspond to the codec you use.

mount hard disk

Click here

RED queue

I have some problem while simulating a simple scenario using 3 TCP sources with RED queue at the bottleneck lick. The tcl script was obtainde for this website. The result that I got is totally different from this document. The queue just grew beyond the limit of 100. The same thing is happening when I rerun the simulation script using UDP protocols under Constant bit rate traffic.

Steps to find the root the problem
1. Recompile ns-2.28 and run the simulation again
2. Run the simulation under ns-2.27
3. Or if this does not work run under ns-2.29
4. Try running the simulation on Fedora

link

link

Finish...

I have completed my exam and waiting for my result to be out. The result will be out in two weeks from now. I hope my cpga will fulfill my graduation requirement. I'm not keen to repeat any paper just to improve my grades.

'akashah on the world map

Haha gotcha, I'm still in Malaysia lah!. I was surprised that when I googled and found that `Akashah is located in Northern Sudan, Afrika. Click `Akashah google map. :P

An update of my work. I'm still in the progress implementing a method to measure the throughput of the packets when the 3g base station is transmitting the packets to mobile node. The packets is actually had been traced in the log file. The problem is that I have still have to do some simple script to calculate the throughtput. It would save some time and memory space if I implements directly into the code.

I have some problem with queue. I still thinking how I'm going to integrate the simulator. Right now, I have no idea. I stumbled upon which explained the modelling of queue. I have got time to read it yet.

Links...

HSDPA validation project
Xavier Dubois's Homepage
Research project
UMTS Forum

Reason

I think why the result of simulation using the UMTS extension did not match the result of simulation using regular ns-2. I set the bandwidth at the core network too big and this did not create any bottleneck.

links...

Click here

real time and non-real time traffic

I confused when the result of two different simulation are not similar. i'm simulating real time traffic and non-real time traffic using UDP contant bit rate (CBR) and exponential respectively.The first simulation is done using the eurane extenstion to ns-2. In this extension, users are allow to create additional nodes (i.e. ue, bs and rnc) that are not defined in the original ns-2. The second simulation is done using the normal nodes defined in the orignal ns-2. For both simulation, I created 2 user equipments, 1 node b, 1 rnc, 1 sgsn, 1 ggsn, 2 nodes, 2 sinks Loss Monitor to measure the UDP bandwindth in Mbit per second. I created a bottle neck between sgsn and ggsn.

The result of the second simulation shows that the the real time traffic will suffer from mixing with non-real time traffics in the link congestion. The result of first simulation does not show any of that. I have set the same bandwidths between each duplex links for both simulation. Is it something do to with sink Loss Monitor or am I not tracing at the correct location?

I'm interested in measuring the throughput at the bottleneck link between sgsn and ggsn.

Anyone has come across any journals or literatures simulating real time traffic and non-real time traffic using UDP?

I got a reply:

The SGSN and GGSN in EURANE are just normal NS-2 nodes. If you placed
the bottleneck there then both simulations would be effected in a
similar way. However the effects of this bottleneck may be negated by
the high throughput possible with the HS-DSCH channel. Without knowing
the data rates of the traffic, and the channel model used for the
HS-DSCH, its difficult to say.

/Neill