Resilient Peer-to-Peer Streaming Paper by: Venkata N. Padmanabhan Helen J. Wang Philip A. Chou Discussion Leader: Manfred Georg Presented by: Christoph.

Resilient Peer-to-Peer Streaming

Paper by:Venkata N. Padmanabhan Helen J. Wang Philip A. Chou

Discussion Leader: Manfred Georg

Presented by: Christoph Jechlitschek

The Problem

Distribution of live streaming media to a potentially large and highly dynamic

population of hosts

Motivation

Flash crowds Often due an event of widespread interest … … but not always (e.g. Webcast of a birthday party) Can effect relatively obscure sites (e.g. www.cricket.org)

Site becomes unreachable precisely when it is popular!

Streaming server can quickly be overwhelmed Network bandwidth is the bottleneck

Other solutions

Content distribution networksIn-house server farmsIncrease bandwidthPeer-to-Peer networksIP multicast

Why other solutions do not work

Not cost effectiveTo expensiveDo not scaleForce user to dedicate bandwidthCan not handle high churn rateNot widely supported

CoopNet

Cooperative NetworkingClients help server to distribute contentIn return the overall quality of the content

increasesPlacing only minimum demands on peers

Challenges

Unreliable peers can disconnect/crash without notice

Constrained and asymmetric bandwidth Last hop is often bottleneck Median broadband bandwidth 900 Kbps/212

Kbps Congestion due to competing applications

Reluctant users Some ISPs charge based on usage

Design Decisions

P2P for scalabilityPeers forward data only if tuned inNo more upload than downloadRedundancy in network pathsRedundancy in data

Redundancy in network paths

A single distribution tree is vulnerable to node failure

Create multiple distribution treesSplit data and send it with or without

redundancy over multiple trees

Example – Before…

Example – After…

Tree Construction

Nodes inform the server when they join Also send delay coordinates

Server constructs and repairs the treeNodes report losses to server

Aggregate reports to avoid overloading server

Design Decisions

Short trees Minimizes chance of disruption

Tree diversity vs. efficiency Diversity: minimizes chance of disruption Efficiency: matches underlying network topology

Quick join and leave Number of round-trips should be small

Scalability Preferable an algorithm that uses O(1) round-trips

Tree Construction algorithms

Randomized Search tree top down for a node with enough

capacity and appoint it as parent for new node Alternatively search also 1 or 2 levels below

and make better decision

Deterministic Each node is fertile in only one tree and sterile

in all other trees Allows shorter and more diverse trees

Random vs. Deterministic

Repairing a tree

If a peer leaves it stops forwarding packets to its children

Peers do not need to notify the rootThe root has to find and replace those

nodesA 1 second repair interval gives good

results

Redundancy in data

Multiple Description Coding Old idea, dates back to the 1970

Voice splitting work at Bell Labs No ordering of the descriptions Any subset of descriptions is decodable The more descriptions received the better the

image quality Increases video stream size

Multiple Description Coding

Order bits by their importance

Split bits in the range [Ri-1, Ri) into i blocks

Send 1 block from each range

If no more blocks to send then send error correction instead

Scalable Client Feedback

Individual feedback from each client can overwhelm server

Instead each peer reports to its parentThe parent combines that report with its

own and passes it to its parent periodicallyNot all trees have to carry feedbackFeedback is used to increase or decrease

redundancy in data

Evaluation

Simulate client joins and leaves based on a 911 flash crowd trace

Needed to substitute original video clipRoot bandwidth 20 Mbps

Peer bandwidth 160 Kbps

Stream bandwidth 160 Kbps

Packet size 1250 bytes

GOF duration 1 second

# descriptions 16

# trees 1, 2, 4, 8, 16

Reporting interval 1 second

Repair interval 1, 5, 10 seconds

Flash Crowd Trace

Multiple Trees

Multiple Trees

Randomized vs. Deterministic Tree Construction

MDC versus FEC

Conclusions

P2P streaming is attractive because it is self scaling

Resilience is provided by multiple distribution trees and MDC

Experiments show promising results

Questions?

References

http://www.research.microsoft.com/projects/CoopNet/papers/icnp2003.pdf

Figures were taken from http://www.research.microsoft.com/~padmanab/talks/resilientP2Pstreaming-mar03.pdf