Xiaojun Hei, Chao Liang, Jian Liang, Yong Liu and Keith W. Rossy

Basic:

  • - a dedicated PPLive crawler
  • - viewing behaviors as regular TV users
  • - peer dynamically with a large number of peers
  • - super peers act as video proxy
  • - long start-up delays and playback lags, ranging from several seconds to a couple of minutes.
  • - PPLive: Spring Festival Gala 2006, 200,000 users 400-800 kbps range

Deviation with BT

  • - no reciprocity mechanisms deployed
  • - video chunk scheduling is required
  • - BT size < 1000. need gossip peer search algorithms

1. Meas

  • - improve on [13] to trace peers behind NAT/firewalls.
  • - offset field 4 bytes, 340 kbps, chunk size of 14 Kbytes
  • - get buffer map by TCP, request one or more chunks by TCP.
    • - likely, get chunk with priority, e.g. the earlier chunk, the rarest chunk
  • - continually searches for new partners
  • - port of internal streaming engine HTTP server. 8888
  • - active crawling to obtain user behaviors and global view
  • - passive sniffing to gain insight from the perspective of residential users and campus users
  • - round by round, in each round T, first crawl peer list S second and then sleep.
  • - Ethereal
  • - built our own customized PPLive packet analyzer to analyze the various fields in the various PPLive signaling and content packets.
  • - Data were obtained at different granularities, including byte-level, packet-level and session-level
  • - CCTV3-Campus, 784M Byte trace on 2hr. in it, 360M download video. 4.5G upload

Basic Result

  • - chunk size > 14K bytes (the exact chunk size depends on the bit rate). the maximum payload size of a TCP segment (typically 1460 bytes), so a video tcp chunk should have > 10 tcp segment.
  • - signaling overhead: 5% - 8%
  • - traffic redundancy: small 13.8%. first 10 minutes of the traces are not used.
  • - signaling by UDP/TCP. video only by TCP
  • - Duration of Video
    • TCP Connections: SYN -> FIN, or 2min no data.
    • median 22.5 seconds, mean is 381.1 seconds. 10% of the connections last for over 15 minutes and the longest session lasts for more than 2 hours.
  • - Neighbors. campus: 40, home, cctv3 20, cctv10 5.
    • - top peer contribute 50% in download, but dynamic from 350kbps to 0.
    • top only take 5% in upload. on average: SN
  • - a true mesh overlay
    • - exists bidirectional video traffic exchange between a pair of peers even in a small time scale (< 1 minute).

2. Crawler

  • - To obtain all participating peers and monitor
  • - get from multiple peerlist servers and peers
  • - Steps
    • - Peer Registration
      • - 128 bit channel identifier
      • - IP, TCP and UDP ports
      • - random peer ID
    • - Bootstrap
      • - bootstrap peer list query message to each peer-list root server
      • - 50 peers, IP and ports (TCP, UDP?)
    • - Peer Query
  • - Peer behind NAT no response. > 50%
  • - can find 95% peers in 5s, T = 60s, S = 15s
  • - evaluation:
    • - use a controlled node to test, 33 experiments, arrival lag was 31.6 seconds; departure lag was 104.2 seconds.
    • - sojourn (寄居) time lag: 70s longer. average sojourn times: 800 - 1600 seconds
    • - peer life time lag overestimates number of active peers.
      • Little law, overestimate ratio 70/X, X real average peer sojourn time measured in seconds.
    • - Consequently, the active peer numbers we subsequently report overestimate the real active peer numbers by 5~9%.
  • Arrive and departure
    • - users evolve over time: diurnal trend
    • - a sharp jump from 50K to 200K at 7:00pm for new year festival. scale well
    • - join and leave at a higher rate at peak times
    • - Peer arrival rate ~ Peak number. Max: 150-200/minute
    • - Peer departure rate:
      • - movie channel: batch-departure when program end -> expect lower peer churn rates in the middle of a program. Max: 1200/minute. Although it is big, but program is end, so no problem. The general rate: 100-200/min
      • - tv channel: no periodic batch departure pattern: 150-200/min
    • - 90% of peers for both programs have lifetimes shorter than 1.5 hours.
  • - buffer level and playable content size: 0-10M, 50% with 7M.
    • Hei et al. [9] also observe that peers seem to strive for buffer levels of 7 Mbytes or higher by analyzing actively crawled cache bitmap but also didn’t dive in deeper.
  • - lag up to 140s

 

3. Refer

  • - chunk scheduling:
    • [14] BT: http://bittorrent.com/
    • [4] X. Zhang, J. Liu, B. Li, and T.-S. P. Yum, ¡°DONet/CoolStreaming: A Data-driven Overlay Network for Peer-to-Peer Live Media Streaming,¡± in IEEE INFOCOM, vol. 3, Mar. 2005, pp. 2102 ¨C 2111.
      • Performance evaluation over PlanetLab [4] and showed that mesh-pull live streaming systems achieve significant more continuous media playback than tree based systems.
    • [13] –, "A measurement study of a large-scale P2P IPTV system, Polytechnic University, Tech. Rep., May 26 2006. [Online]. Available:
    • http://cis.poly.edu/"ross/papers/P2PliveStreamingMeasurement.pdf
  • mesh-pull live streaming systems achieve significant more continuous media playback than tree based systems
    • [20] X. Zhang, J. Liu, and B. Li, "On large-scale peer-to-peer live video distribution: Coolstreaming and its preliminary experimental results," in IEEE MMSP’2005, Oct. 2005.
  • another prototype
    • [21] J. Liu, S. G. Rao, B. Li, and H. Zhang, "Opportunities and challenges of peer-to-peer Internet video broadcast," Nov. 2006.
  • broadcasting video over the Internet
    • [22] R. Rejaie, "Anyone can broadcast video over the Internet," Commun. ACM, vol. 49, no. 11, pp. 55-57, 2006.
  • a number of mesh-pull P2P streaming systems
    • [23] V. Pai, K. Kumar, K. Tamilmani, V. Sambamurthy, and A. E. Mohr, "Chainsaw: Eliminating trees from overlay multicast," in IPTPS’05, Feb. 2005.
    • [24] M. Zhang, L. Zhao, Y. Tang, J.-G. Luo, and S.-Q. Yang, "Large-scale live media streaming over peer-to-peer networks through global Internet," in P2PMMS’05, 2005, pp. 21-28.
    • [25] C. Dana, D. Li, D. Harrison, and C. N. Chuah, "BASS: BitTorrent assisted streaming system for video-on-demand," in IEEE MMSP, Oct. 2005.
    • [26] X. Liao, H. Jin, Y. Liu, L. M. Ni, and D. Deng, "AnySee: Peer-to-Peer live streaming," in IEEE INFOCOM, Apr. 2006.
    • [27] F. Pianese, J. Keller, and E. W. Biersack, "PULSE, a flexible P2P live streaming system," in Global Internet, Apr. 2006.
    • [28] A. Vlavianos, M. Iliofotou, and M. Faloutsos, "BiToS: Enhancing BitTorrent for supporting streaming applications," in Global Internet, Apr. 2006.
    • [29] N. Magharei and R. Rejaie, "Understanding mesh-based peer-to-peer streaming," in NOSSDAV ‘06, May 2006.
    • [30] T. Piotrowski, S. Banerjee, S. Bhatnagar, S. Ganguly, and R. Izmailov, "Peer-to-peer streaming of stored media: the indirect approach," in SIGMETRICS ‘06/Performance ‘06, 2006, pp. 371-372.
  • theoretical studies of mesh-pull streaming systems
    • [18] R. Kumar, Y. Liu, and K. W. Ross, "Stochastic fluid theory for P2P streaming systems," in Proceedings of INFOCOM, 2007.
      • - show buffering can significantly improve video streaming quality.
    • [31] S. Tewari and L. Kleinrock, "Analytical model for BitTorrent-based live video streaming," in IEEE NIME 2007 Workshop, Jan. 2007.
  • Gnutella
    • measurement of Napster and Gnutella [35] and detailed characterization of end-user hosts in these two systems.
      • [35] S. Saroiu, K. P. Gummadi, and S. D. Gribble, "Measuring and analyzing the characteristics of Napster and Gnutella hosts," Multimedia Syst., vol. 9, no. 2, pp. 170-184, 2003.
    • monitor of KaZaa traffic [36] and showed locality-aware P2P file-sharing architectures can achieve significant bandwidth savings.
      • [36] K. P. Gummadi, R. J. Dunn, S. Saroiu, S. D. Gribble, H. M. Levy, and J. Zahorjan, "Measurement, Modeling, and Analysis of a Peer-to-Peer File-sharing Workload," in ACM SOSP, 2003, pp. 314-329.
    • Ripeanu et al. crawled the one-tier Gnutella network to extract its overlay topology. For the latest two-tier Gnutella network, Stutzbach et al. provided a detailed characterization of P2P overlay topologies and their dynamics in [37].
      • [37] D. Stutzbach, R. Rejaie, and S. Sen, "Characterizing Unstructured Overlay Topologies in Modern P2P File-Sharing Systems," in ACM IMC, Oct. 2005.
    • Liang et al. deployed active crawling in [38] to reveal in-depth understanding of KaZaa overlay structure and dynamics.
      • [38] J. Liang, R. Kumar, and K. W. Ross, "The FastTrack Overlay: A Measurement Study," Computer Networks, vol. 50, no. 6, pp. 842-858, Apr. 2006.
    • In [39], Liang et al. further demonstrated the existence of content pollution and poisoning in KaZaa using an active crawler.
      • [39] J. Liang, N. Naoumov, and K. Ross, "The Index Poisoning Attack in P2P File-Sharing Systems," in IEEE INFOCOM, Apr. 2006.
  • measurement study of live streaming workload from a large CDN
    • [40] K. Sripanidkulchai, B. Maggs, and H. Zhang, "An analysis of live streaming workloads on the Internet," in ACM IMC, 2004, pp. 41-54.
  • measurement results for BitTorrent content distribution.
    • [41] M. Izal, G. Urvoy-Keller, E. W. Biersack, P. Felber, A. A. Hamra, and L. Garc´es-Erice, "Dissecting bittorrent: Five months in a torrent’s lifetime." in PAM, 2004, pp. 1-11.
    • [42] J. Pouwelse, P. Garbacki, D. Epema, and H. Sips, "The Bittorrent P2P File-sharing System: Measurements and Analysis," in IPTPS’05, Feb. 2005.
  • protocol analysis of Skype and Skype traffic pattern
    • [43] S. A. Baset and H. Schulzrinne, "An Analysis of the Skype Peer-to-Peer Internet Telephony Protocol," in IEEE INFOCOM, Apr. 2006.
    • [15] S. Guha, N. Daswani, and R. Jain, "An Experimental Study of the Skype Peer-to-Peer VoIP System," in IPTPS’06, Feb. 2006.
  • - free MaxMind GeoIP database [16]
    • [16] "Maxmind," http://www.maxmind.com/app/country.
  • - file-sharing monitoring companies (such as Big Champagne [17]