1 / 22

Best-Case WiBro Performance for a Single Flow

AAF workshop Measurement Session. MICNET 2009. Best-Case WiBro Performance for a Single Flow. Shinae Woo † , Keon Jang † , Sangman Kim † Soohyun Cho * , Jaehwa Lee * , Youngseok Lee ‡ , Sue Moon †. † Computer Science Department, KAIST

lazar
Download Presentation

Best-Case WiBro Performance for a Single Flow

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. AAF workshop Measurement Session MICNET 2009 Best-Case WiBro Performance for a Single Flow Shinae Woo†, Keon Jang†, Sangman Kim† Soohyun Cho*, Jaehwa Lee*, Youngseok Lee ‡, Sue Moon † † Computer Science Department, KAIST ‡ School of Computer Science and Engineering, Chungnam National University * KT Central R&D Laboratory Nov 23, 2009

  2. What is WiBro (Mobile WiMAX) ? • Provides high-bandwidth mobile data service • Served by KT and SKT in Korea • Being deployed in US, India and Japan Han River Population 20,000,000 WiBro users 200,000 Seoul 50km Inchen Korea Airport Service Area

  3. Market Positioning of WiBro (Mobile WiMAX) • Speed : Cellular networks < WiBro < WiFi • Mobility : WiFi < WiBro < Cellular networks Speed WiFi 144Mbps WiBro 35.44Mbps HSPA 14.4Mbps GSM Mobility 120km/h 300km/h

  4. Characteristics of WiBro (Mobile WiMAX) • IP-based packet-switching in the core • Uses 2.3GHz frequency, OFDMA • - Multiple users can access same time • Uses TDMA for up/downlink b/w allocation • 5 QoS classes implemented, only BE used now

  5. Outline Background Motivation Experiment Environment Analysis Basic characteristics of WiBro TCP Performance over WiBro Conclusion

  6. Motivation • Experiments on commercial network • Hard to pinpoint major factors affecting performance • To know the baseline of WiBro performance • Focus on the performance of a single flow • Assume no competition and interaction between flows

  7. KREONET Experiment Set-up GPS GPS WiBro Network 1Gbps link Server Client KT’s IP Network

  8. KREONET Experiment Set-up GPS Link speed = 1Gbps GPS WiBro Network Server Client KT’s IP Network KT – KREONET MRTG graph (26 April, 19:40:14)

  9. RTT by the Hop RTT (ms) Hop number

  10. RTT by the Hop RTT (ms) Wireless link dominates e2e performance Hop number

  11. Experiment Environment Architecture of Our GPS System GPS module U-blox LEA-5 LAN cable NMEA data 1 PPS USB RS232

  12. Experiment Environment Clock Synchronization in Our GPS System GPS signal per second by PPS Packet sent or arrival event CPU clock time GPS UTC time

  13. Analysis Overview Basic Characteristics of WiBro • How different is WiBro from WiFi and cellular networks? • We use UDP to measure min and max one-way delay • - Min one-way delay = best-case w/o queuing • - Max one-way delay = represents queue size TCP Performance over WiBro • Do we get theoretically maximum 35.44 Mbps over WiBro? • - Vary send and recv window sizes

  14. Basic Characteristics of WiBro Minimum one-way delay • Definition : delay without queuing • = transmission delay + propagation delay • + scheduling delay (at ACR ) • Measurement methodology • Send 40 byte packet every 40ms • Measurement outcome: • - Downlink: 14ms • Uplink: 100ms

  15. Bandwidth and Queuing Delay of WiBro • We saturated link with UDP traffic Downlink10 Mbps Uplink2.5 Mbps Large queuing delay

  16. Basic Characteristics of WiBro Queue Size of WiBro links • Queue size decides the queuing delay!! • With full queue, • Queue size = (max. one-way delay) * (bandwidth) • Measured queue size in each direction, • - Downlink 961ms * 9603Kbps = 1126 KB • - Uplink 417ms * 2628Kbps =137 KB • If bandwidth temporarily decreases to 1 Mbps, • queuing delay can increase up to 10 s!!!

  17. TCP performance over WiBro How Does the Buffer Size Affect TCP Bandwidth? • WiBro link has high bandwidth-delay product • TCP need socket buffer size of at least bandwidth-delay product • ex) Window XP’s default TCP socket buffer size, 17KB, is not enough to utilize the WiBro link fully

  18. TCP performance over WiBro How Does the Buffer Size Affect TCP Bandwidth? Bandwidth (Mbps) Window Size(KB) 64KB 5 Mbps with Windows XP’s default socket buffer size Min 128 KB buffer size can fully utilize the link

  19. TCP performance over WiBro Auto-tuned TCP Socket Buffer in Ubuntu TCP get the same bandwidth with UDP  Auto-tuned buffer works well in WiBro

  20. TCP performance over WiBro One-way Delay of Downlink TCP Traffic Increase with queuing Decrease with piggybacking Window Size (KB)

  21. Our Contributions • Developed cheap GPS module with sub-millisecond accuracy • Showed that WiBro is high bandwidth-delay network High Long Big • To fully utilize WiBro network, • min 128KB or more TCP buffer size is needed • - Windows XP’s default size without TCP 1323 extension option is only 64KB

  22. Thank you!! GPS LAN cable USB RS232

More Related