A simulation-based approach for performance evaluation of sdr baseband architectures Brussels, Belgium 28 th June, 2012
Goal and objectives Trends in the field of radio communication Mobile devices with more and more wireless interfaces, user applications and adaptation capabilities Parallel architectures clustered by application category to implement mobile terminals Goal of our work To facilitate performance evaluation of SDR baseband architectures Objectives of this presentation To propose simulation-based approach to analyze and compare the growing number of potential architectures To illustrate benefits of this approach with a realistic adaptive multi-service system 2
Model requirements Specification design step Definition of the system properties and its performance requirements Executable model to evaluate and compare performances of candidate architectures Fundamental criteria to respect Quick-to-develop and lightweight to decrease modeling effort of the designer Accuracy and simulation speed for different use cases 3
Considered modeling approach Performance evaluation of system architectures Performance model of system architecture TC A2 TC A1 SA / k:=0; A 2 A 1 s 0 Cc A2 =0; TC A11 TC A12 M 3 / k:=k+1; t:=0; k<N AND t=T j A 11 A 12 s 1 Cc A2 =Cc s1 ; k=N AND t = T j M 4 M 1 M 2 M 3 / t:=0; s 2 Cc A2 =Cc s2 ; t = T k / t:=0; s 3 Cc A2 =Cc s3 ; t = T l M 4 Considered system architecture Node F 12 F 2 F 11 P 1 P 2 Mem. 4
Considered case studied Activity diagram of an adaptive multi-standard and multi-application system and its environment Radio communication system System management Service request Service management Service response Application RAT management Active application request RAT discovery Application Application and RAT response RAT information Switching request Network environment RAT control QoS information System data flow User Radio reception and transmission Application processing Packet data voice Voice QoS Voice processing UTRA reception Downlink UTRA Voice frame Packet data web Web processing Web QoS Packet data video Web page Video processing WiFi reception Video QoS Downlink Wi-Fi Video frame Voice processing UTRA transmission Packet data voice Voice frame Uplink UTRA 5
Modeling the system environment Modeling technique based on scenario files Network environment User Radio communication system Application requests Network environment control reading reading Waiting Waiting Application request Max rate radio links Update Rat data rate Application request Send request Max rate radio links WiFi Transmission UTRA Transmission Downlink WiFi Downlink UTRA 6
Modeling the communication interfaces Activity diagram of adaptive radio interfaces Radio reception RAT control t 1 t 2 up to WiFi reception 347µs UTRA reception [WIFI] 1 Stopped t 1 t 2 up to 20 ms 2347 bytes RAT control=stop RAT control=start 1 Downlink WiFi 244 bits Started Packet data voice t 1 t 2 Downlink UTRA 1s DownlinkUTRA 1 t 1 t 2 Demux ≈ 12000 bytes 10ms & Assemb [UTRA] 1 Packet data web up to t 1 t 2 Packet Packet Cond1 Cond3 480 bytes data data 1 s voice video Packet 1 Cond2 Cond4 data web 8000 bytes Packet data video 7
Generation and simulation of the performance model Graphical modeler and ANSI C/C++ code editor to capture the performance model Generation of executable SystemC code from capture model C++ code Specification Graphical modeler and code editor SystemC executable code Functional verification Use cases scenario and performance evaluation Simulation of the executable SystemC program according to complex use case scenarios Evaluation of real time performances Evaluation of the expected ressources 8
Temporal behavior analysis of the system 1s VoiceCall Start 15s WebSession Start 16s VideoStreaming Start 2s WebSession Stop 25s VideoStreaming Stop 8s VoiceCall Stop User Application request Application response Service management Service request Service response Switching request System management RAT discovery RAT control Web page Video frame WiFi reception UTRA reception Voice frame 0 10 20 30 40 50 60 Time (s) Latency (ms) Threshold Web 1s 1000 800 Web Threshold video 600 Video 500ms 400 200 Threshold voice Voice 20ms 0 10 20 30 40 50 60 Time (s) UTRAN 384 kbps 130 kbps WLAN 0 kbps 1500 kbps 0s UTRAN 384 30s WLAN 1500 9s UTRAN 130 9
Performance evaluation of the flexible baseband architecture Studied architecture to perform baseband processing related to activities UTRA and WiFi reception Architecture based on a set of dedicated hardware resources Performance model express computational complexity per time unit each function causes on the resources when executed Performance model of radio reception and transmission architecture Radio reception and transmission TC WiFi WiFi reception Packet data voice TC UTRA Downlink WiFi Packet data web UTRA reception Packet data video Downlink UTRA Studied architecture P 1 UTRA baseband functions WiFi baseband functions 10
Simulation results of the model Evolution in time of the required computational complexity per time unit (in MOPS) for UTRA and WiFi decoding Global computational complexity per time unit (MOPS) 1260 MOPS 1300 1183 MOPS 1200 Observation for studied architecture (3) 1100 and operating scenario considered (2) 1000 900 Maximal computational complexity 800 700 per time unit observed 600 600 Resource utilization of P 1 500 400 300 (1) 200 77 MOPS 100 t(m s) 40040 40060 40080 UTRA decoding Wi-Fi decoding 11
Sum-up and conclusion Sum-up Simulation-based approach and modeling technique to evaluate efficiently performances of candidate SDR baseband architectures Simulate easily multiple complex use cases Study dynamic and non determinist ic effects in the architecture model Further work Validation of estimates providing by simulation Applying the same modeling principle to other non functional properties such as dynamic power consumption 12
A simulation-based approach for performance evaluation of sdr baseband architectures Brussels, Belgium 28 th June, 2012
Recommend
More recommend
