F5/unison
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

updated documentation

Browse Source
This commit is contained in:
Nicola Baldo
2011-05-20 18:42:04 +02:00
parent fb29cfb14d
commit 0fbb1e39d1
3 changed files with 13 additions and 42 deletions
Download Patch File Download Diff File Expand all files Collapse all files

36
src/lte/doc/source/lte-design.rst
View File

@@ -128,16 +128,16 @@ Proportional Fair (PF) Scheduler
Here we describe the PF scheduler that we implement, providing references from the literature.
LTE Spectrum Model
~~~~~~~~~~~~~~~~~~
Here we describe how the LTE spectrum model works and what it allows to do.
Spectrum Model
Physical layer
~~~~~~~~~~~~~~
The usage of the radio spectrum by eNBs and UEs in LTE is described in [TS36.101]_.
In the simulator, we model it as follow. for communications is modeled as follows.
Let :math:`f_c` denote the LTE Absolute Radio Frequency Channel Number, which identifies the carrier frequency on a 100 kHz raster; furthermore, let :math:`B` be the Transmission Bandwidth Configuration in number of Resource Blocks. For every pair :math:`(f_c,B)` used in the simulation we create a corresponding spectrum model using the `ns3::Spectrum`` framework of [Baldo2009]_. All these LTE-specific spectrum models are used to simulate different spectrum usage policies in the same simulation, for example allowing the simulation of cognitive radio / dynamic spectrum access strategies for LTE.
The physical layer model provided in this LTE simulator model supports Frequency Division Duplex (FDD) only.
@@ -146,11 +146,11 @@ Propagation Loss Models
~~~~~~~~~~~~~~~~~~~~~~~
NOTE: this information refers to the GSoC model, which as of this writing is not working anymore.
A proper propagation loss model has been developed for the LTE E-UTRAN interface (see [2]_ and [3]_).
A proper propagation loss model has been developed for the LTE E-UTRAN interface (see [TS25.814]_ and [Piro2010]_).
It is used by the PHY layer to compute the loss due to the propagation.
The LTE propagation loss model is composed by 4 different models (shadowing, multipath,
penetration loss and path loss) [2]_:
penetration loss and path loss) [TS25.814]_:
* Pathloss: :math:`PL = 128.1 + (37.6 * log10 (R))`, where R is the distance between the
UE and the eNB in Km.
@@ -162,25 +162,7 @@ penetration loss and path loss) [2]_:
* Shadowing: log-normal distribution (mean=0dB, standard deviation=8dB)
Every time that the ``LteSpectrumPHY::StartRx ()`` function is called, the
``SpectrumInterferenceModel`` is used to computed the SINR, as proposed in [3]_. Then,
``SpectrumInterferenceModel`` is used to computed the SINR, as proposed in [Piro2010]_. Then,
the network device uses the AMC module to map the SINR to a proper CQI and to send it
to the eNB using the ideal control channel.
References
**********
.. [1] N. Baldo and M. Miozzo, Spectrum-aware Channel and PHY layer modeling for ns3, Proceedings
of ICST NSTools 2009, Pisa, Italy. The framework is designed to simulate only data
transmissions. For the transmission of control messages (such as CQI feedback, PDCCH,
etc..) will be used an ideal control channel).
.. [2] 3GPP TS 25.814 ( http://www.3gpp.org/ftp/specs/html-INFO/25814.htm )
.. [3] Giuseppe Piro, Luigi Alfredo Grieco, Gennaro Boggia, and Pietro Camarda", A Two-level
Scheduling Algorithm for QoS Support in the Downlink of LTE Cellular Networks", Proc. of
European Wireless, EW2010, Lucca, Italy, Apr., 2010 ( draft version is available on
http://telematics.poliba.it/index.php?option=com_jombib&task=showbib&id=330 )
.. [4] 3GPP R1-081483 (available on
http://www.3gpp.org/ftp/tsg_ran/WG1_RL1/TSGR1_52b/Docs/R1-081483.zip )

13
src/lte/doc/source/lte-testing.rst
View File

@@ -145,17 +145,4 @@ Let :math:`Ri` the estimation done by PFS of the throughput of the :math:`i` UE
The test passes if the measured throughput matches with the reference throughput within a relative tolerance of 0.1. The choice of this tolerance has the same motivations already discussed for the Round Robin scheduler test suite.
References
**********
.. [TS36.213] 3GPP TS 36.213 "LTE Physical layer procedures"
.. [Kushner2004] H.J. Kushner and P.A. Whiting, "Convergence of proportional-fair sharing algorithms under general conditions", IEEE Trans. on Wireless Communications, July 2004
.. [Piro2011] G. Piro, N. Baldo. M. Miozzo, "An LTE module for the ns-3 network simulator", Wns3 2011
(in conjunction with SimuTOOLS 2011), March 2011, Barcelona (Spain)
.. [Seo2004] H. Seo, B. G. Lee. "A proportional-fair power allocation scheme for fair and efficient multiuser
OFDM systems", In Proc. of IEEE GLOBECOM, December 2004. Dallas (USA)
.. [R1-081483] 3GPP R1-081483 "Conveying MCS and TB size via PDCCH"

6
src/lte/doc/source/lte-user.rst
View File

@@ -74,7 +74,9 @@ Performance evaluation
Execution time and memory consumption
-------------------------------------
In order to provide an running time and memory consumption estimation a reference simulation script has been developed, ``src/lte/examples/profiling-reference``. The scenario is composed by a set of eNodeBs, each one of them with a constant number of UEs attached. All eNodeBs have the same number of attached UEs. The UEs are all in the same position than its eNodeB and the eNodeBs are distributed in a line, each one 140m away from the previous one. Simulated time is set to 60s.
In order to provide an running time and memory consumption estimation a reference simulation program has been developed, ``src/lte/examples/profiling-reference``. This program simulates a scenario composed by a set of eNodeBs, and a set of UEs attached to each eNB. All eNodeBs have the same number of attached UEs. Communications are performed both in the dowlink and in the uplink using a saturation model (i.e., each RLC instance always has a PDU to transmit). The UEs are all in the same position than its eNodeB and the eNodeBs are distributed in a line, each one 140m away from the previous one. The total simulation time is set to 60s.
With this considerations, the execution time and the memory consumtpion has been obtained for a certain number of UEs attached to each eNodeB and number of eNodeBs. The reference hardware platform is a Intel Core2 Duo E8400 3.00GHz with 512 MB of RAM memory running a Fedora Core 10 with kernel 2.6.27.5. It scenarios considered range the number of eNodeBs between 1 and 15 and the UEs per eNodeB in 1, 5 and 10. The performance figures obtained show follows.
Using this simulation program, we ran a simulation campaign varying the number of eNBs as well as the number of UEs per eNB. For each simulation, we measured the execution time using the ``time`` shell command in linux, and the memory consumtpion by looking at the information in ``/proc/{pid}/statm``. The reference hardware platform is a Intel Core2 Duo E8400 3.00GHz with 512 MB of RAM memory running a Fedora Core 10 with kernel 2.6.27.5. The simulator build used in this experiment was configured with the options ``-d optimized --enable-static``.
The results are reported in the figures below. We note that the memory usage primarily depends on the number of eNBs, and is in general quite low. The execution time depends significantly on both the number of eNBs and the number of UEs per eNB, and in the largest experiment that we ran (12 eNBs and 10 UEs per eNB) it corresponds to 30 times the simulation time.