MAC-level throughput

256 views
Skip to first unread message

Manuel Cano

unread,
Dec 18, 2014, 7:16:48 AM12/18/14
to ns-3-...@googlegroups.com

Hello everyone,

I’m configuring a onoff traffic source with a 5 Mbps constant rate in downlink, 20MHz (PRB = 100), MCS = 28, TBSize = 9422 bytes and 60 sec. We use UDP sockets:

// Create the On-Off application in the remote node

// This one represents the Video server in the internet

OnOffHelper onoffremoteVideo ("ns3::UdpSocketFactory",

                        Address (InetSocketAddress (ueAddr[0], dlPort)));

onoffremoteVideo.SetAttribute ("OffTime", StringValue ("ns3::ConstantRandomVariable[Constant=0]"));

onoffremoteVideo.SetConstantRate(DataRate ("5Mbps"),500);  //. PacketSize: 2500-> If we want to avoid IP fragmentation (MTU<1472). Back off IP(20)+UDP(8) header.

serverVideoApps.Add(onoffremoteVideo.Install(remote));

 

  I’m running  LTE LENA simulator on ns 3.19 and according to FlowMon  the resulting throughput is 5.03473 Mbps which matches the transmission rate defined. When I calculate throughput using PDCP traces I get 662500 Bps = 5,3 Mbps, as it can be seen in the following trace fragment:

start

end

        nTxPDUs

      TxBytes

      nRxPDUs

      RxBytes

2

2.25

312

165360

313

165890

 

2.25

2.5

313

165890

312

165360

 

2.5

2.75

312

165360

313

165890

 

2.75

3

313

165890

312

165360

 













                                       DlPdcpStats

But when I calculate throughput at MAC level I have 1000 packets/second * 9422 Bytes = 9,422 MBps = 75,376 Mbps, as extracted from the DlMac.

% time            cellId              IMSI               frame              sframe             RNTI               mcsTb1           sizeTb1            mcsTb2           sizeTb2

1.1                   1                      1                      111                  1                      1                      28                    9422                0                      0

1.101               1                      1                      111                  2                      1                      28                    9422                0                      0

1.102               1                      1                      111                  3                      1                      28                    9422                0                      0

1.103               1                      1                      111                  4                      1                      28                    9422                0                      0

1.104               1                      1                      111                  5                      1                      28                    9422                0                      0

1.198               1                      1                      120                  9                      1                      28                    9422                0                      0

1.199               1                      1                      120                  10                    1                      28                    9422                0                      0

1.2                   1                      1                      121                  1                      1                      28                    9422                0                      0

                                                                                      DlMacStats

If I run ns3 with a throughput higher than 4 Mbps I always get 75,376 Mbps because up to 4 Mbps I have 1 transmission per second in MAC level.

Why the achieved throughput at MAC-level (DlMacStats) is so high with respect to IP (FlowMon) and PDCP-level (DlPdcpStats)?

Maybe there is something that we are missing out, any explanation for this?

Nicola Baldo

unread,
Dec 19, 2014, 7:22:18 AM12/19/14
to ns-3-...@googlegroups.com
with the application setting that you have, you'll see approx one new IP packet per 0.8 milliseconds, right? So this explains why the LTE MAC sends one TB per TTI.
The size of the TB depends on the LTE MAC scheduler implementation, but it could be that it's large just because of the very strong signal and the absence of other users, i.e., the TB could be partially empty.
Reply all
Reply to author
Forward
0 new messages