Designing an Efficient and Extensible Mobile TV Testbed Cheng-Hsin Hsu Simon Fraser University, Canada joint work with Mohamed Hefeeda, Yi Liu, and Cong Ly

Mobile TV Service 2  Broadcast mass-market programs to subscribers  Mobile devices have stringent energy budgets  Devices receive a data burst and turn off receiving circuits until the next burst  called time slicing

Base Station Mobile TV Networks 3  Program feeds are IP streams from streaming servers or cameras  Multiple TV programs are multiplexed AND time sliced by a multiplexer into a MPEG-2 TS stream  The MPEG-2 TS stream is modulated, amplified, and broadcast to mobile devices Streaming Server Camera Multiplexer (IP Encapsulator) Modulator/ Amplifier IP Networks ASI Networks Content Providers Network Operators

Problem Statement 4  Design a mobile TV base station for academic prototyping and cost-efficient small- to medium- size deployments  platform to analyze: energy consumption, channel switching delay, no. broadcast programs, and perceived streaming quality  TV channels with a commodity PC or low-end server

Current Solutions (1/2) 5  Commercial Base Stations  expensive, e.g., a single EXPWAY FastESG server costs 75k USD [Sarri09]  a complete base station costs even more Need a more cost-efficient base station!

x20? Current Solutions (2/2) 6  Open-Source Base Station [FATCAPS]  too many disk I/O’s  does not scale well  too many utilities with no admin interface Need a better design! VLC Server MPEG Data Aggregator Pcap Time Slicer TS Tdt Updater Correct PSISI Dtplay Python Compiler Python Sources For PSISI TS Packets Null Packet Replacer w/PSISI

Design Goals 7  [G1] Higher efficiency and scalability  avoid disk I/O’s and memcpys  [G2] Utilization of multi-core processors  pipelined structure to allow parallelism  [G3] Integrated software solution  centralized admin interface  [G4] Better extensibility  future supports for other networks such as MediaFLO, WiMAX, and MBMS

Design Decisions (1/3) 8  [D1] Use Burst as the unit of time slicing, encapsulation, and transmission.  Burst is self-contained with IP payloads and headers/trailers of all protocols  No disk I/O’s for intermediate data  No memcpys for IP payloads

Design Decisions (2/3) 9  [D2] Divide the base station into three indep. Phases, which are connected by two priority queue  pipelined and parallelism Empty Burst Time Slicing Thread With IP Payload Request Queue Encap. Thread With All Headers /Trailers Ready Queue Trans. Thread Encap. Thread Encap. Thread

Design Decisions (3/3) 10  [D3] Implement a centralized Configuration Manager to allow save/restore settings  interface with Web GUI for management  [D4] Modularized design for future extensions  For example, MPE-FEC Burst is a subclass of MPE Burst

Software Architecture 11

High-Level Design 12

Design of Burst Schedulers 13

Design of Burst Readers 14

Design of Bursts 15

Design of Encapsulators 16

Design of Transmitter 17

Design of PSI/SI Tables 18  PAT: program association  PMT: program map  NIT: network information  INT: IP/MAC notification  SDT: service description  TDT: time and date

Design of Configuration Manager 19

Testbed Setup 20

Future Work 21  Web GUI for configuration management [Cong]  MPE-FEC support [Hamed]  PSI/SI table implementation [Farid]  StreamReader classes [Som]  Flute server integration  ESG files implementation

Conclusions 22  Presented layout of general broadcast network  Outlined design goals of a broadcast base station  Described our design decisions and system architecture  Presented the high-level system design and detailed design for each component  Highlighted future work

Thank You, and Questions? 23 More details can be found online at