Introduction Objectives Centralized Radio Access Network Real-time emulations Performance Results Conclusions References Real-time Emulation Methodologies for Centralized Radio Access Networks USE OF OPENAIRINTERFACE IN RESEARCH AND PROTOTYPING Luis Felipe Ariza Vesga Universidad Nacional de Colombia Raymond Knopp EURECOM Nokia Bell Labs, Murray Hill, New Jersey - June 26, 2019 1 / 21
23
Embed
Real-time Emulation Methodologies for Centralized Radio ... · Conclusions References Real-time Emulation Methodologies for Centralized Radio Access Networks USE OF OPENAIRINTERFACE
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
IntroductionObjectives
Centralized Radio Access NetworkReal-time emulationsPerformance Results
ConclusionsReferences
Real-time Emulation Methodologies forCentralized Radio Access Networks
USE OF OPENAIRINTERFACE IN RESEARCH ANDPROTOTYPING
Luis Felipe Ariza VesgaUniversidad Nacional de Colombia
Raymond KnoppEURECOM
Nokia Bell Labs, Murray Hill, New Jersey - June 26, 20191 / 21
IntroductionObjectives
Centralized Radio Access NetworkReal-time emulationsPerformance Results
ConclusionsReferences
Agenda1 Introduction2 Objectives3 Centralized Radio Access Network
Architecture4 Real-time emulations
Frequency domain extension5 Performance Results
Performance MetricsA proof-of-conceptVideo Demo
6 Conclusions7 References
2 / 21
IntroductionObjectives
Centralized Radio Access NetworkReal-time emulationsPerformance Results
ConclusionsReferences
Introduction
There is a trade-off between network simulations, networkemulators and real test-beds :
1 A network simulator has good scalability and reproducibi-lity.
2 A network emulator has good applicability and captures3GPP standard-compliant environments.
3 A test-bed has good applicability but reproducibility issuesin multi-vendor scenarios.
Optimizing software functions and simulating the multipathchannel in terms of a frequency domain representation, wedecrease the signal processing complexity in a software-only environment.
3 / 21
IntroductionObjectives
Centralized Radio Access NetworkReal-time emulationsPerformance Results
ConclusionsReferences
Objectives
Increase the scalability of real-time synthetic networks (Mul-tiple Remote Radio Units and User Ends) in a software-onlyenvironment.Prototype 4G and 5G rapid proof-of-concept designs beforelaunching to the market.Hybridize real-time synthetic network components, and Ra-dio Frequency (RF) hardware for complex scenarios.
4 / 21
IntroductionObjectives
Centralized Radio Access NetworkReal-time emulationsPerformance Results
ConclusionsReferences
Architecture
Architecture
We extracted Primary and Secondary SynchronizationSignals (PSS and SSS) information from the eNB andassigned to the UE (frame_type, cell_id).PBCH is decoded from the rxdataF at the UE(initial_synch_freq() function).
TABLE – Averagecomputation times in timeand frequency domains.C-RAN architecture, 5MHz of bandwidth, 10000frames, AWGN channelmodel, and 5 MB of iperftraffic.
7 / 21
IntroductionObjectives
Centralized Radio Access NetworkReal-time emulationsPerformance Results
ConclusionsReferences
Frequency domain extension
Gaussian random number generators
Gaussian random number generators (rf_rx_simple_freq) areemployed to simulate the noise at the receiver.
-5 -4 -3 -2 -1 0 1 2 3 4 5
Fre
qu
en
cy o
f o
cu
rre
nce
Standard deviation
-4 -3 -2 -1 0 1 2 3 4
Fre
qu
en
cy o
f o
cu
rre
nce
Standard deviation
-4 -3 -2 -1 0 1 2 3 4
Fre
qu
en
cy o
f o
cu
rre
nce
Standard deviation
FIGURE – Zigguratmethod to generateGaussian randomnumbers.
0
5000
10000
15000
20000
25000
30000
35000
40000
45000
-5-4-3-2-1 0 1 2 3 4 5
Fre
qu
en
cy o
f o
cu
rre
nce
Gaussian pseudo-random number
0
50000
100000
150000
200000
250000
300000
350000
-6 -4 -2 0 2 4 6
Fre
qu
en
cy o
f o
cu
rre
nce
Gaussian pseudo-random number
0
100000
200000
300000
400000
500000
600000
700000
-6 -4 -2 0 2 4 6
Fre
qu
en
cy o
f o
cu
rre
nce
Gaussian pseudo-random number
FIGURE – Box-Mullermethod to generateGaussian randomnumbers.
Generator Samples Chi-Square Computation time (ns/samples)Box-Muller 9.99e+05 290
TABLE – Chi-Square andaverage computation timemetrics for Box-Muller andZiggurat methods.
8 / 21
IntroductionObjectives
Centralized Radio Access NetworkReal-time emulationsPerformance Results
ConclusionsReferences
Frequency domain extension
Physical slot structure
In the frequency domain analysis the Cyclic Prefix is not imple-mented. Inter-symbol interference is not avoided. We change thetime_stamp in eNB_trx_read and UE_trx_read functions.
TABLE – Average computation times intime and frequency domains. C-RANarchitecture, 5 MHz of bandwidth, 10000frames, AWGN channel model, and 5 MBof iperf downlink traffic.
15 / 21
IntroductionObjectives
Centralized Radio Access NetworkReal-time emulationsPerformance Results
ConclusionsReferences
Performance MetricsA proof-of-conceptVideo Demo
Average computation time
101
102
103
104
105
1 2 3
Tim
e pe
r sub
fram
e (u
s)
UEs
Time DomainFrequency Domain
FIGURE – Average computation time benchmark.
16 / 21
IntroductionObjectives
Centralized Radio Access NetworkReal-time emulationsPerformance Results
ConclusionsReferences
Performance MetricsA proof-of-conceptVideo Demo
Synthetic network scalability
RCCs CCs/RRUs UEs Time Domain µs Frequency Domain µs B200mini-i µs
Centralized Radio Access NetworkReal-time emulationsPerformance Results
ConclusionsReferences
Conclusions
We successfully implemented affordable real-time emula-tion methodologies in the frequency domain for C-RANs asa prototyping framework to rapid proof-of-concept and time-to-market designs in a software-only environment.We reduced near 10-fold the average computation time ofthe multipath channel in the frequency domain compared tothe time domain. The cost in time we need to pay is relatedto the additional uplink PRACH channel.We improved the applicability and the scalability fro CRANson top of OpenAirInterface.Our proposal allows real-time 3GPP standard-compliant C-RANs emulations for downlink and uplink transmissions.
20 / 21
IntroductionObjectives
Centralized Radio Access NetworkReal-time emulationsPerformance Results
ConclusionsReferences
References
Gitlab branches : large_scale_simulations for RRUs + UEs,and master_large_scale_emulations for the RCC andmultiple CCs.Several videos related to the extensions in the frequencydomain.
Centralized Radio Access NetworkReal-time emulationsPerformance Results
ConclusionsReferences
L. M. P. Larsen, A. Checko, and H. L. Christiansen, “Asurvey of the functional splits proposed for 5g mobilecrosshaul networks,” IEEE Communications SurveysTutorials, vol. 21, no. 1, pp. 146–172, Firstquarter 2019.
LG. (2019) Lte resource grid. [Online]. Available :http://niviuk.free.fr/lte_resource_grid.html
F. Kaltenberger, “Low-complexity real-time signalprocessing for wireless communications,” Ph.D.dissertation, Vienna University of Technology, Vienna,Austria, May 2007.