Will Web Video-on-Interest be Beneficial?.

Uploaded on:
We utilize the 9 months MSN follow, which was gathered for a customer ... Driven by 9 months of MSN Video follow. 2 features. Gold stream: the most well known feature, mainstream for a couple of days. ...
Slide 1

Could Internet Video-on-Demand be Profitable? Cheng Huang, Jin Li (Microsoft Research Redmond), Keith W. Ross (Polytechnic University) ACM SIGCOMM 2007

Slide 2

Outlines Motivation Trace – User request & conduct Peer-helped VoD Theory Real-follow driven reenactment Cross ISP activity issue Conclusion

Slide 3

Motivation Saving cash for tremendous substance suppliers, for example, MSN Video, Youtube, Yahoo Video, Google Video, … Video quality is simply satisfactory User BW ++++++ User BW + User BW +++ User request +++ Traffic ++ Traffic + Traffic +++ Traffic ++++++++ ISP Charge + ISP Charge +++++++ ISP Charge ++ ISP Charge +++ P2P Client Server Video quality +++ Video quality +++ Video quality + Video quality +++++++

Slide 4

P2P Architecture Peers will help each other and won " t devour the server BW. Every associate have commitment to the entire framework. Toss the ball back to the ISPs The activity does not vanish, it moved to elsewhere.

Slide 5

Outlines Motivation Trace – User request & conduct Peer-helped VoD Theory Real-follow driven reproduction Cross ISP movement issue Conclusion

Slide 6

Trace Analysis Using a follow contains 520M spilling solicitations and more than 59,000+ recordings from Microsoft MSN Video . http://video.msn.com/From April to December, 2006 .(9 Months)

Slide 8

The prevalence appropriations are entirely comparable. There is without a doubt a high-level of territory. The circulation is more skewed than a Zipf dissemination. Video Popularity The more skewed, the greatly improved.

Slide 9

Download transfer speed Use ISP download/transfer evaluating table Downlink dispersion to create transfer BW circulation

Slide 10

Demand v.s. Support Available transfer data transfer capacity at customers far surpasses client request.

Slide 11

Users for the most part view expansive portion of short recordings. Be that as it may, under 20% of the clients see more than 60% of recordings bigger than 30 minutes. Client conduct - Churn

Slide 12

User Behavior Fraction of sessions that begin toward the start of a video and have no intelligence is vital in the accomplishment of an associate helped VoD. For < 30 min recordings, 80% of the session does not have intelligence.

Slide 13

Content quality upset The interest and the bitrates for VoD increment quickly.

Slide 14

They trust it is likely that bitrates will increment speedier than customer transfer B/W. Activity Evolution 1.23 2.27 Quality Growth: half User Growth: 33% Traffic Growth: 78.5 %

Slide 15

Contributions of this paper The principal estimation investigation of an on-interest video gushing framework in a huge scale. We introduce a basic hypothesis for associate helped VoD. This hypothesis distinguishes 3 essential working methods of associate helped VoD framework. The surplus mode, the adjusted mode, and the shortfall mode. For the single-video methodology, we depict 3 regular prefetching arrangements for misusing surplus companion transfer limit. No-prefetching, water-leveling, and voracious approach. We utilize the 9 months MSN follow, which was gathered for a customer server organization, to drive reenactments for companion helped arrangements. We investigate the effect of companion helped VoD on ISPs.

Slide 16

Outlines Motivation Trace – User request & conduct Peer-helped VoD Theory Real-follow driven recreation Cross ISP activity issue Conclusion

Slide 17

Peer-helped VoD Peer-helped VoD Users viewing the video will help with the re-dissemination of the video to different clients. There is still a server (or server ranch) which stores the greater part of the distributer\'s recordings. Ensures that clients playback the video at the playback rate with no quality debasement. The server is just dynamic when the associates alone can\'t fulfill the interest. 2 outline ways to deal with companion helped VoD Single video: an associate just redistributes the video it is at present viewing. (This paper utilize!!) Multiple video: an associate can redistribute a video that it already saw however is right now not seeing.

Slide 18

Modeling – Single Video The time client stays online to see the video is T The bitrate of the video is r Users touch base at the framework with Poisson conveyance rate l M is the quantity of client sort, where a sort m client has transfer join BW w m p m : the likelihood that an entry is a sort m client System " s normal transfer B/W of an arriving client is m = S p m w m Expected number of sort m clients is p m l T In unfaltering state , the normal aggregate Demand is r S p m l T = r l T The normal Supply is S p m l T w m = m l T If Supply > Demand Surplus mode , little server load If Supply < Demand Deficit mode , VERY expansive server load If Supply ≈ Demand Balanced mode , medium server load

Slide 19

Prefetch Policy Let each companion get more video information than interest (if conceivable) in surplus mode. What\'s more, in this manner can hold over deficiency mode. Companions can conceivably prefetch video from each other utilizing the associates\' surplus data transmission. 3 prefetching arrangements No-prefetching Each client downloads content at the playback rate r and does not prefetch content for future needs. At any given moment of time, the client might download from different associates and also from the server. Accept that every client sees the video without crevices. Water-leveling prefetching Peer just prefetches from associates touch base before it and have adequate transfer data transmission, and interest is rely on upon the client support level. Make every one of the associates to have the same cushion levels of prefetched substance. Eager prefetching Each client just devotes its outstanding transfer BW to the following client directly after itself. For every client i, give it " s transfer BW and collected BW to client i+1 [4] C. Huang, J. Li, and K. W. Ross, "Peer-Assisted VoD: Making Internet Video Distribution Cheap," IPTPS , Bellevue, WA, Feb. 2007.

Slide 20

Outlines Motivation Trace – User request & conduct Peer-helped VoD Theory Real-follow driven recreation Cross ISP movement issue Conclusion

Slide 21

Methodology Discrete-occasion test system. Driven by 9 months of MSN Video follow. 2 recordings Gold stream : the most prevalent video, well known for a couple days. Silver stream : the second most mainstream video, well known for a month. Concentrate on the adjusted mode . Insatiable prefetching . no P2P : the assets utilized by the unadulterated customer server arrangement.

Slide 22

Simulation: Non-early-takeoff Trace P2P sending at the present quality level, ordinarily no server assets are required. A few assets required when couple of simultaneous clients.

Slide 23

Simulation: Early takeoff (No collaboration) When video length > 30mins, 80%+ clients wear " t complete the entire video. Table 4: Server rates (in Mbps) under various framework modes with early takeoffs. April 2006. Indeed, even with early takeoffs peer-help can give a decent change. Prefetching keeps on giving enhancements over non-prefetching.

Slide 24

Simulation: Full Trace How to manage support gaps ? As client may skip part of the video. 2 techniques Conservative: accept that client transfer BW=0 after the primary connection. Hopeful : overlook all cooperations, there is no gap in the client\'s cushion.

Slide 25

Results of full follow recreation (1/2) Due to intelligence, a client may have openings in its cushion. The genuine execution will lie between these two limits.

Slide 26

Results of full follow reproduction (2/2)

Slide 27

Cost Reduction With companion helped VoD, server BW decrease from 2.2Gbps to 79.4Mbps on Dec. 2006.

Slide 28

Outlines Motivation Trace – User request & conduct Peer-helped VoD Theory Real-follow driven reenactment Cross ISP movement issue Conclusion

Slide 29

ISP-disagreeable P2P VoD ISPs, in view of business relations, will shape monetary substances. 3 connections between ISPs: 1) travel relationship (likewise called client supplier) one ISP acquiring Internet access from another ISP and paying for the data transfer capacity utilization. 2) kin relationship the interconnection among a few ISPs having a place with the same association. 3) peering relationship ISPs matching with each other. Peering ISPs can trade activity specifically, which would some way or another need to experience their suppliers. Movement don\'t go through the limit won " t be charged. ISP-unpleasant P2P will bring about substantial measure of movement.

Slide 30

Simulation consequences of antagonistic P2P Most P2P VoD crosses ISP limits.

Slide 31

Simulation aftereffects of well disposed P2P Peers lies in various financial substances don\'t help each other. Table 8: Server transfer speed (in Mbps) in an ISP-advanced situation. Silver stream single video, 5000 particular video circulations. Main 10 more prevalent recordings among the 12000 in follows. At the point when an element contains few companions, the sharing turns out to be more troublesome too, and the server transmission capacity is expanded appropriately.

Slide 32

Conclusion Peer-helped VoD, with the best possible prefetching strategy, can significantly decrease server data transmission costs. Peer-helped VoD can be both server and ISP inviting.

Slide 33

References [3] B. Cheng, X. Liu, Z. Zhang, and H. Jin, "A Measurement Study of a Peer-to-Peer Video-on-Demand System," IPTPS , Bellevue, WA, Feb. 2007. [4] C. Huang, J. Li, and K. W. Ross, "Peer-Assisted VoD: Making Internet Video Distribution Cheap," IPTPS , Bellevue, WA, Feb. 2007. [19] A. Al Hamra, E. W. Biersack, and G. Urvoy-Keller, "A Pull-based Approach for a VoD Service in P2P Networks," IEEE HSNMC , Toulouse, France, Jul. 2004. [20] Y. Cui, B. Li, and K. Nahrstedt, "oStream: Asynchronous Streaming Multicast in Application-Layer Overlay Networks," IEEE JSAC , 22(1), 2004. [21] J. Li, Y. Cui, and B. Chang, "PeerStreaming: Design and Implementation of an On-Demand Distributed Streaming System with DRM Capabilities," Multimedia Systems Journal , 2007. [22] S. Annapureddy, C. Gkantsidis, P. R. Rodriguez, and L. Massoulie, "Giving Video-on-Demand Using Peer-to-Peer Networks,

View more...