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. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.
Peer-to-Peer Networks (3) - IPTV Hongli Luo CEIT, IPFW
Internet Video Broadcasting • References: • “Opportunities and Challenges of Peer-to-Peer Internet Video Broadcast” by Liu et al. • “Insights into PPLive: A Measurement Study of a Large-Scale P2P IPTV System” by Hei et al.
Background • Large-scale video broadcast over Internet • Real-time video streaming • Applications: • Internet TV • Broadcast of sports events • Online games • Distance education • Need to support large numbers of viewers • AOL Live 8 broadcast peaked at 175,000 (July 2005) • CBS NCAA broadcast peaked at 268,000 (March 2006) • Very high data rate • TV quality video encoded with MPEG-4 would require 1.5 Tbps aggregate capacity for 100 million viewers • NFL Superbowl 2007 had 93 million viewers in the U.S. (Nielsen Media Research)
Possible Solutions • Broadcasting is possible in air, cable networks, or local area networks • Possible solutions for broadcasting over Internet • Single server - unicast • IP multicast • Multicast overlay networks • Content delivery networks (CDNs) • Application end points (pure P2P)
Single Server • Application-layer solution • Single media server unicasts to all clients • Needs very high capacity to serve large number of clients • CPU • Main memory • Bandwidth • Impractical for millions of simultaneous viewers
IP Multicast • Network-layer solution • Routers responsible for multicasting • Efficient bandwidth usage • Requires per-group state in routers • High complexity • Scalability concern • Violates end-to-end design principle
IP Multicast Multicast group • Unicast via Multicast Unicast Multicast Clients Clients C C Server Server S S C C C C
IP Multicast • End-to-end design principle: a functionality should be • Pushed to higher layers if possible, unless • Implementing it at the lower layer can achieve significant performance befits that outweigh the cost of additional complexity • Slow deployment • IP multicast is often disabled in routers • Difficult to support higher layer functionality • Error control, flow control, and congestion control • Needs changes at the infrastructure level
IP Multicast Gatech Stanford Source: Purdue Berkeley Per-group Router State “Smart Network” Source: Sanjay Rao’s lecture from Purdue
Multicast Overlay Network • Consists of user hosts and possibly dedicated servers scattered through the Internet • Hosts, servers and logical links between them form an overlay network, which multicasts traffic from the source to users • Originally in IP multimcast, router responsible for forwarding packets, application run on the end systems. • New applications can now make their own forwarding decisions. • A logical network implemented on top of a physical network. • Consists of application-layer links • Application-layer link is logical link consisting of one or more links in underlying network • Each node in the overlay processes and forwards packets in an application-specific way • Used by both CDNs and pure P2P systems
MBone • The multicast backbone (MBone) is an overlay network that implements IP multicast. • Mbone was an experimental backbone for IP multicast traffic across the Internet • It connects multicast-capable networks over the existing Internet infrastructure • One of the popular applications run on top of the MBone is Vic • Vic • Supports multiparty videoconferencing • Broadcast both seminars and meetings across the Internet, e.g. IETF meetings
Content replication challenging to stream large files (e.g., video) from single origin server in real time solution: replicate content at hundreds of servers throughout Internet content downloaded to CDN servers ahead of time placing content “close” to user avoids impairments (loss, delay) of sending content over long paths CDN server typically in edge/access network Content distribution networks (CDNs) origin server in North America CDN distribution node CDN server in S. America CDN server in Asia CDN server in Europe
Content replication CDN (e.g., Akamai) customer is the content provider (e.g., CNN) CDN places CND servers close to ISP access networks and the clients CDN replicates customers’ content in CDN servers. when provider updates content, CDN updates servers Content distribution networks (CDNs) origin server in North America CDN distribution node CDN server in S. America CDN server in Asia CDN server in Europe
Content distribution networks (CDNs) • When a client requests content, the content is provided by the CDN server that can best deliver the content to the specific • The closest CDN server to the client • CDN server with a congestion-free path to the client • A CDN server typically contains objects from many content providers
origin server (www.foo.com) distributes HTML replaces: http://www.foo.com/sports.ruth.gif withhttp://www.cdn.com/www.foo.com/sports/ruth.gif CDN example HTTP request for www.foo.com/sports/sports.html origin server 1 DNS query for www.cdn.com 2 CDN’s authoritative DNS server client 3 HTTP request for www.cdn.com/www.foo.com/sports/ruth.gif CDN server near client CDN company (cdn.com) • distributes gif files • uses its authoritative DNS server to route redirect requests
routing requests CDNs make use of DNS redirection in order to guide browsers to the correct server. The browser does a DNS lookup on www.cdn.com, which is forwarded to authoritative DNS server. CDN’s DNS server returns the IP address of the CDN server that is likely the best for the requesting browser. when query arrives at authoritative DNS server: server determines ISP from which query originates uses “map” to determine best CDN server CDN nodes create application-layer overlay network CDN: bring content closer to clients More about CDNs
Why P2P? • Previous problems • Sparse deployment of IP multicast • High cost of bandwidth requirement for server-based unicast and CDNs. • Limit video broadcasting to only a subset of Internet content publishers
Why P2P? • Every node is both a server and client • Easier to deploy applications at endpoints • No need to build and maintain expensive routers and expensive infrastructure • Potential for both performance improvement and additional robustness • Additional clients create additional servers for scalability • Performance penalty • Can not completely prevent multiple overlay edges from traversing the same physical link • Redundant traffic on physical links • Increasing latency
Peer-to-peer Video Broadcasting • Characteristics of video broadcasting • Large scale, corresponding to tens of thousands of users simultaneously participating the broadcast. • Performance-demanding, • involving bandwidth requirements of hundreds of kilobits per-second and even more. • Real-time constraints, requiring timely and continuously streaming delivery. • While interactivity may not be critical and minor delays can be tolerated through buffering, it is critical to get video uninterrupted. • Gracefully degradable quality, • enabling adaptive and flexible delivery that accommodates bandwidth heterogeneity and dynamics.
Peer-to-peer Video Broadcasting • Stringent real-time performance requirement • Bandwidth and latency • On-demand streaming – users are asynchronous • Audio/video conferencing – interactive, latency more critical • File downloading • No time constraint, segments of contents can arrive out of order • Needs efficient indexing and search • P2p video broadcasting • Simultaneously support a large number of participants, • Dynamic changes to participant membership, • High bandwidth requirement of the video • Needs efficient data communication.
Overlay construction • Criterion: • Overlay efficiency • Scalability and load balancing • Self-organizing • Honor per-node bandwidth constraint • System considerations • Approaches • Tree-based • Data-driven randomized
P2P Overlay • Tree-based • Peers are organized into trees for delivering data • Each packet disseminated using the same structure • Parent-child relationships • Push-based • When a node receives a data packet, it forwards copies of the packet to each of its children. • Failure of nodes result in poor transient performance. • Uplink bandwidth not utilized at leaf nodes • Data can be divided and disseminated along multiple trees (e.g., SplitStream) • The structure should be optimized to offer good performance to all receivers. • Must be repaired and maintained to avoid interruptions • Example: End System Multicast (ESM)
P2P Overlay • Data-driven • Do not construct and maintain an explicit structure for delivering data • Use gossip algorithms to distribute data • A node sends a newly generated message to a set of randomly selected nodes. • These nodes do similarly in the next round, and so do other nodes until the message is spread to all. • Pull-based • Nodes maintain a set of partners • Periodically exchange data availability with random partners and retrieve new data • Redundancy is avoided since node pulls data only it does not have it. • Similar to BitTorrent, but must consider real-time constraints • Scheduling algorithm schedules the segments that must be downloaded to meet the playback deadlines • Example: CoolStreaming, PPLive
Tree-based vs. Data driven • Data driven • Simple • Suffer from a latency-overhead trade-off • Tree-based • No latency-overhead trade-off • Instability • Bandwidth under-utilization • A combination of both
P2P live streaming system • CoolStreaming • X.Zhang, J. Liu, B. Li, and T. S. P. Yum. Coolstreaming/donet: A data-driven overlay network for efficient live media streaming. In Proceedings of IEEE INFOCOM’05, March 2005. • PPLive • http://www.pplive.com • PPStream • http://www.ppstream.com • UUSee • http://www.uusee.com • AnySee • X. Liao, H. Jin, Y. Liu, L. M. Ni, and D. Deng. Anysee: Peer-to-peer live streaming. In Proceedings of IEEE INFOCOM’06, April 2006. • Joost • http://www.joost.com/
Case Study: PPLive • PPLive: free P2P-based IPTV • As of January 2006, the PPLive network provided • 200+ channels • with 400,000 daily users on average. • Typically over 100,000 simultaneously users • It now covers over 120 TV Chinese stations, 300 live channels and 20,000 VOD (video-on-demand) channels (from Wiki) • The company claimed that they have more than 200 million user installations and 105 million active monthly user base (as of Dec 2010). (from Wiki) • The bit rates of video programs mainly range from 250 Kbps to 400 Kbps with a few channels as high as 800 Kbps. • The channels are encoded in two video formats: Window Media Video (WMV) or Real Video (RMVB). • The encoded video content is divided into chunks and distributed to users through the PPLive P2P network. • Employs proprietary signaling and video delivery protocols
Case Study: PPLive • BitTorrent is not a feasible video delivery architecture • Does not account for the real-time needs of IPTV • Bearing strong similarities to BitTorrent • Video chunk has playback deadline • No reciprocity mechanism deployed to encourage sharing between peers • Two major application level protocols • A gossip-based protocol • Peer management • Channel discovery • P2P-based video distribution protocol • High quality video streaming • Data-driven p2p streaming
Case Study: PPLive • User starts PPlive software and becomes a peer node. • Contact channel server for list of available channels • Select a channel • Sends query to root server to retrieve an online peer list for this channel • Find active peers on channel to share video chunks Channel and peer discovery from “Insights into PPLive: A Measurement Study of a Large-Scale P2P IPTV System” by Hei et al.
TV Engine • Download video chunks from PPLive network • Stream the downloaded video to a local video player • Streaming process traverses two buffers • PPLive TV engine buffer • Media player buffer • Cached contents can be uploaded to other peers watching the same channel. • This peer may also upload cached video chunks to multiple peers. • Peer may also download media content from multiple active peers • Received video chunks are reassembled in order and buffered in queue of PPLive TV engine, forming local streaming file in memory.
When the streaming file length crosses a predefined threshold, the PPLive TV engine launches media player, which downloads video content from local HTTP streaming server. • After the buffer of the media player fills up to required level, the actual video playback starts. • When PPLive starts, the PPLive TV engine downloads media content from peers aggressively to minimize playback start-up delay. • PPLive uses TCP for both signaling and video streaming • When the media player receives enough content and starts to play the media, streaming process gradually stabilizes. • The PPLive TV engine streams data to the media player at media playback rate.
Measurement setup • One residential and one campus PC “watched” channel CCTV3 • The other residential and campus PC “watched” channel CCTV10 • Each of these four traces lasted about 2 hours. • From the PPLive web site, CCTV3 is a popular channel with a 5-star popularity grade and CCTV10 is less popular with a 3-star popularity grade.
Start-up delays • A peer search for peers and download data from active peers. • Two types of start-up delay: • the delay from when one channel is selected until the streaming player pops up; • the delay from when the player pops up until the playback actually starts. • The player pop-up delay is in general 10-15 seconds • The player buffering delay is around 10-15 seconds. • Therefore, the total start-up delay is around 20 - 30 seconds. • Nevertheless, some less popular channels have a total start-up delays of up to 2 minutes. • Overall, PPLive exhibits reasonably good start-up user experiences
Video Traffic Redundancy • It is possible to download same video blocks more than once • Transmission of redundant video is a waste of network bandwidth • Define redundancy ratio as ratio between redundant traffic and estimated media segment size. • The traffic redundancy in PPLive is limited • Partially due to the long buffer time period • Peers have enough time to locate peers in the same channel and exchange content availability information.
Video Buffering • Estimation of size of media player buffer: • at least 5.37 MBytes • Estimation of size of PPLive engine buffer: • 7.8 MBytes to 17.1 Mbytes • The total buffer size in PPLive streaming • 10 - 30 Mbytes • A commodity PC can easily meet this buffer requirement
PPLive Peering Statistics • A campus peer has many more active video peer neighbors than a residential peer due to its high-bandwidth access network. • A campus peer maintains a steady number of active TCP connections for video traffic exchanges. • Peers with less popular channel have difficulty in finding enough peers for streaming the media • If the number of active video peers drops, the peer searches for new peers for additional video download. • Peer constantly changes its upload and download neighbors