Confusion on setting up wifi bandwidth

174 views
Skip to first unread message

Najwan Khambari

unread,
Feb 9, 2015, 11:17:15 AM2/9/15
to ns-3-...@googlegroups.com
Hi,
I have a very basic wifi design consisting of 2 nodes (a sender and a receiver) and 1 AP.  
The resulting parameters are all being captured by Flowmon.

For the Data Mode, I had set it to 6Mbps.  

  std::string myPhy="OfdmRate6Mbps";
  wifi.SetRemoteStationManager ("ns3::ConstantRateWifiManager", 
"DataMode", StringValue (myPhy),
"ControlMode",StringValue(myPhy));


Then for the sender application (OnOff Application), I had tried several values from 1Mbps to 6Mbps with 1000 bytes of payload.

sender.SetConstantRate( DataRate(myDataRate), myPayload);


Since there is only one sender in the network, I am expecting nearly the whole bandwidth can be used by the application.  However, this is not the case.  
The maximum throughput of the sender is only at 2490 kbps at the Data Rate of 6Mbps of the application.

App Data Rate Loss Tput (kbps)
------------- ---- ----------
1 Mbps 0 1004.33
2 Mbps 0 2007.98
3 Mbps 0 2498.00
4 Mbps 855 2491.19
5 Mbps 1729 2490.91
6 Mbps 2605 2490.54

I'm not sure why this is happening or perhaps I have setup the scenario in a wrong manner? 
I just want to see at which point the network becomes saturated but with a bandwidth of 6Mbps, it should not be saturated by just before 4Mbps.  

Any opinion would be very much appreciated.  I attached the script together with this message.

Thank you so much.
test3_clone.cc

Najwan Khambari

unread,
Feb 9, 2015, 1:25:04 PM2/9/15
to ns-3-...@googlegroups.com
Hi Sebastian,

Thank you for your reply, and I appreciate it. 
Yes, I understand about the overheads, contention and ACK (but in this case, I'm using UDP).
I'm just curious because it got saturated at 50%, where I actually expected it to be 60-70%.  The contention in this scenario is very minimal.

However, it's good to know that I have done the right thing to setup the physical transmission rate and it's normal to have an 50% goodput.

Which is the other sender that you said? Is it the receiver who sends the ACK or the time when contention happens?


On Monday, February 9, 2015 at 5:39:47 PM UTC, Sebastien Deronne wrote:
It is wrong, there is more than one sender in your scenario. Results are not at all surprising.

In infrastructure mode, all traffic must go through the access point. In your case, the sender sends packet to the access point, which then sends packets to the receiver. So, the wireless channel is shared and you also experience contention in your network.

Please also note that even though it is a 6 Mbit/s physical transmission, you have to consider overheads and ACK frame transmission to compute your real goodput! 
So, typically, for a .11a/g connection, the maximum achievable goodput is around 50% of the physical bitrate.

I advise you to read first some basic materials about Wi-Fi standard.

Konstantinos

unread,
Feb 9, 2015, 1:31:04 PM2/9/15
to ns-3-...@googlegroups.com
Hi Najwan,

You have one STA sending the traffic and the AP forwarding it to the other STA. So there are your two contending nodes.
The ACKs are from the MAC layer, not TCP-ACKs, for each 'hop'. So you have ACKs from AP to STA_1 and ACKs from STA_2 to AP.

Najwan Khambari

unread,
Feb 9, 2015, 2:26:18 PM2/9/15
to ns-3-...@googlegroups.com
Hi Kostas,
Thank you very much for your reply.
That makes sense.

Sebastien, Kostas, with all the things happening in the world, it's a good reminder for me that there are still a lot of kind-hearted people in this world like you guys.

Many thanks and kind regards.

Sebastien Deronne

unread,
Feb 9, 2015, 12:39:47 PM2/9/15
to ns-3-...@googlegroups.com
Reply all
Reply to author
Forward
0 new messages