You are on page 1of 40

QoE in Pull Based P2P-TV Systems: Overlay Topology Design Tradeoffs

R. Fortuna, E. Leonardi, M. Mellia, M. Meo, S. Traverso
IEEE International Conference on Peer-to-Peer Computing

August 2010

S. Traverso (Politecnico di Torino)

Overlay Topology Design Tradeoffs

August 2010

Outline

Outline

S. Traverso (Politecnico di Torino)

Overlay Topology Design Tradeoffs

August 2010

1 / 29

Outline

Outline
1

Introduction Motivations Definitions Overlay Topology Design The Strategy Chunk Signaling and Scheduling Performance Evaluation Model
Network Scenario Video Parameters

2

Results
3

Video-Aware Schedulers The Strategy Results Conclusions Q&A
Overlay Topology Design Tradeoffs August 2010 2 / 29

4 5

S. Traverso (Politecnico di Torino)

Introduction

Introduction

S. Traverso (Politecnico di Torino)

Overlay Topology Design Tradeoffs

August 2010

3 / 29

i. the network built by peers at application layer. The “Network-Aware P2P-TV Application over Wise Networks” FP7 project aims at developing an application to broadcast high definition video. but many issues about how to improve the user perceived video quality are still open. while offering a world-wide service. S. Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 4 / 29 . We focused our attention on i) the design of the overlay topology.Introduction Motivations Why this work? P2P-TV is a promising technology to reduce the cost of streaming content over the Internet.e. ... ii) exploiting properties of encoded videostream to improve users’ QoE...

S... The “Network-Aware P2P-TV Application over Wise Networks” FP7 project aims at developing an application to broadcast high definition video. while offering a world-wide service. i. Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 4 / 29 .e. the network built by peers at application layer. but many issues about how to improve the user perceived video quality are still open.. .. We focused our attention on i) the design of the overlay topology.Introduction Motivations Why this work? P2P-TV is a promising technology to reduce the cost of streaming content over the Internet. ii) exploiting properties of encoded videostream to improve users’ QoE.

differently from file sharing. P2P streaming systems. Peers start exchanging chunks according to some scheduling scheme. Chunks are exchanged among peers that are neighbors of each others. Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 5 / 29 .Introduction Definitions Basic concepts A source peer splits the video streams in small chunks that are injected inside the overlay. S. have to face with the real-time constraint! Every chunk must be received within a deadline Dmax also called playout delay.

We neglect the effect of churning since scheduling dynamics are much faster. Peers exchange signaling information to trade chunks. Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 6 / 29 .Introduction Definitions Assumptions Peers are Internet nodes and tipically their upload bandwidth is much lower than the downlink one. Peers upload bandwidth and latencies between peers are supposed to be known somehow. S.

Peers exchange signaling information to trade chunks. S.Introduction Definitions Assumptions Peers are Internet nodes and tipically their upload bandwidth is much lower than the downlink one. We neglect the effect of churning since scheduling dynamics are much faster. Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 6 / 29 . Peers upload bandwidth and latencies between peers are supposed to be known somehow.

S. Peers exchange signaling information to trade chunks. Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 6 / 29 . We neglect the effect of churning since scheduling dynamics are much faster. Peers upload bandwidth and latencies between peers are supposed to be known somehow.Introduction Definitions Assumptions Peers are Internet nodes and tipically their upload bandwidth is much lower than the downlink one.

Introduction Definitions Assumptions Peers are Internet nodes and tipically their upload bandwidth is much lower than the downlink one. Peers exchange signaling information to trade chunks. S. Peers upload bandwidth and latencies between peers are supposed to be known somehow. We neglect the effect of churning since scheduling dynamics are much faster. Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 6 / 29 .

Overlay Topology Design Overlay Topology Design S. Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 7 / 29 .

Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 8 / 29 . S. high bandwidth peers should be well connected to each other (to the source). The number of neighbors of a peer have to choose.Overlay Topology Design The Strategy Intuition To avoid long trading phases. To speed up chunk replication. should be related to its upload bandwidth. peers with short end-to-end latencies should be connected to each other.

Overlay Topology Design The Strategy Intuition To avoid long trading phases. S. should be related to its upload bandwidth. The number of neighbors of a peer have to choose. Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 8 / 29 . To speed up chunk replication. peers with short end-to-end latencies should be connected to each other. high bandwidth peers should be well connected to each other (to the source).

peers with short end-to-end latencies should be connected to each other.Overlay Topology Design The Strategy Intuition To avoid long trading phases. To speed up chunk replication. high bandwidth peers should be well connected to each other (to the source). should be related to its upload bandwidth. S. Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 8 / 29 . The number of neighbors of a peer have to choose.

Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 9 / 29 . we set Kp = max (3. S. all peers q such that Bq > 1/2Bp and lpq < 1/2E [lpq ] are marked as desired peers. Given the size of Kp . How to choose neighbors? Given peer p . The neighborood size The resulting size of the neighborood of a peer is Cp > Kp (edges are bidirectional). αKp peers are randomly selected and (1 − α)Kp peers are selected within a set of desired peers of p.Overlay Topology Design The Strategy The Strategy How many neighbors? Being Kp the number of neighbors of a peer p chooses. 10 Bp /rs ) where Bp is the upload bandwidth of peer and rs is the average video rate of the stream.

P5 P2 OFFERS SELECTS P1 P2 Np P7 OFFERS P4 SELECTS Np Chunk #1 to Peer 2 time Chunk #1 to Peer 5 Chunk #2 to Peer 1 Chunk #2 to Peer 2 Chunk #2 to Peer 7 Chunk #3 to Peer 4 S. Once the select message is received. the chunk is then transmitted.Overlay Topology Design Chunk Signaling and Scheduling A pull mechanism Every peer periodically generates offer messages to publish the list of its useful chunks. When the chunk is received. Neighbors reply to each offer with a select message in which they specify the chunk they need. an acknoledgement message is sent back to the trasmitter. Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 10 / 29 .

This is commonly known as Random Peer. S. neighbors to contact with offer messages are chosen uniformly at random.Overlay Topology Design Chunk Signaling and Scheduling Peer and Chunk Selections The Peer Selection Np neighbors are contacted by a peer in every offer session. Random Useful Chunk scheduling policy. Np is a fraction of the total neighborood Cp . Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 11 / 29 . The Chunk Selection Chunks requested in select messages are randomly chosen over the the list of useful ones proposed in offer messages. In our tests.

7 58.64Mb/s ± 10%.8 4 20 20 20 20 August 2010 12 / 29 Overlay Topology Design Tradeoffs .2 51.8 13.10 H = 0.5 35. Traverso (Politecnico di Torino) 1 1 5 10 15 2 76.5 34. Class H = 0.Overlay Topology Design Performance Evaluation Model Nework Scenario Our simulation involved 2000 peers partitioned in four classes according to their upload capacity: Class 1: 5. with increasing heterogeneity. The average upload bandwidth is E [Bp ] = 1.3 16. We consider 4 scenarios.3Mb/s in all cases. Class 3: 0.0Mb/s ± 10%. Class 2: 1.05 H = 0. Class 4: 0Mb/s.15 S.01 H = 0.6Mb/s ± 10%.2 3 2.

Peers are distributed over the Earth surface and scattered over domains representing continents.Overlay Topology Design Performance Evaluation Model The Transport Network The transport network introduces a latency lpq to all the datagrams sent from peer p to q .6 0. Then. August 2010 13 / 29 Overlay Topology Design Tradeoffs .2 Latency [s] 0.9 0.5 0.3 0.8 0. S. latencies are proportional to the geodetical distance between peers.2 0.05 0.1 0 0 0. E [lpq ] = 96ms.4 0.25 0. Traverso (Politecnico di Torino) Figure: Latency frequencies.35 Figure: Peers distribution over Earth surface.15 0.7 0.3 0.1 0. 80 60 Latitude [Deg] 40 0 -20 -40 -60 -80 -150 -100 -50 0 50 Longitude [Deg] 100 150 PDF 20 1 0.

P. 352 × 240 352 × 288 352 × 288 Frame/sec 25 30 25 Pink of the Aerosmith Paris Foreman The videos consists of 1000 frames ≈ 40s of visualization.3s 40s Spatial Res. B or b. S. Hierarchical structure of GOP: frames can be IDR.Overlay Topology Design Performance Evaluation Model Real video sequences Three well-known video sequences have been considered as benchmarks.264/AVC codec has been adopted for encoding sequences. Length 40s 33.B or b) carry differential information (and are much smaller). Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 14 / 29 . inter frames (P. Intra frames (IDR) carry valuable information (bigger). H.

4 0.6 0.2 0.9 α Average PSNR for different values of α for ρ = 0. Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 15 / 29 .8 0.9 (ρ = E [Bp ]/rs ).7 0.3 0.1 0.Overlay Topology Design Results The Impact of α 45 40 PSNR [dB] 35 30 25 EVQ Dmax=6s Dmax=5s Dmax=4s Dmax=3s 1 0. S.5 0.

0 S. Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 16 / 29 .Overlay Topology Design Results The Impact of α II 80 60 Latitude [Deg] Latitude [Deg] 40 20 0 -20 -40 -60 -80 -150 -100 -50 0 50 Longitude [Deg] 100 150 80 60 20 0 -20 -40 -60 -80 -150 -100 -50 0 50 Longitude [Deg] 100 150 Latitude [Deg] 40 80 60 40 20 0 -20 -40 -60 -80 -150 -100 -50 0 50 Longitude [Deg] 100 150 Figure: α = 0.0 Figure: α = 0.1 Figure: α = 1.

Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 16 / 29 .1 Figure: α = 1.0 Dmax=5s α=0.0 can lead to disconnected topologies! S.0 46 44 PSNR [dB] 42 40 38 36 Figure: α = 0.0 EVQ α=0.1 Dmax=5s α=1.0 Dmax=5s 1 6 11 16 21 26 Peers ID 31 36 41 α = 0.Overlay Topology Design Results The Impact of α II 80 60 Latitude [Deg] Latitude [Deg] 40 20 0 -20 -40 -60 -80 -150 -100 -50 0 50 Longitude [Deg] 100 150 80 60 20 0 -20 -40 -60 -80 -150 -100 -50 0 50 Longitude [Deg] 100 150 Latitude [Deg] 40 80 60 40 20 0 -20 -40 -60 -80 -150 -100 -50 0 50 Longitude [Deg] 100 150 Figure: α = 0.

Overlay Topology Design Results Adapting Kp to the upload capacity 45 40 PSNR [dB] 35 30 25 EVQ Dmax=6s Dmax=5s PSNR [dB] Dmax=4s Dmax=3s 1 45 40 35 30 25 EVQ Dmax=6s Dmax=5s Dmax=4s Dmax=3s 1 0.9 α 0.6 0.8 0.9 α Fixed Kp = 20 (left) and variable Kp = max (3.2 0.4 0.2 0. 10 Bp /rs ) (right) for ρ = 0.9.5 0.8 0. S.3 0. Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 17 / 29 .7 0.6 0.1 0.4 0.5 0.7 0.1 0.3 0.

4 α Dmax=3s Dmax=4s Dmax=5s Dmax=6s 0.8 1 Figure: The network stress index.07 0.02 0 0. Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 18 / 29 .2 0.05 0.06 0.08 0. i.1 0.e.04 0.Overlay Topology Design Results Network stress 0. S.9).03 0. the average distance covered by chunks expressed it terms of the corresponding latency (ρ = 0.6 0.09 Network stress [ms] 0.

9 α 1 0.8 0.9.5 41 0.3 0.5 43 42.15 EVQ 45 40 35 30 25 Paris Foreman Pink Paris EVQ Foreman EVQ Pink EVQ 0. Dmax = 5s and ρ = 0.10 and variable Kp .22 0.5 45 PSNR [dB] 44. Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 19 / 29 .61 0.9.805 1 Figure: Average PSNR versus α for Figure: Average PSNR versus α for different degrees of heterogeneity H with different video sequences with ρ = 0. S. H = 0.7 0.6 0.5 44 43.05 H=0.01 H=0.Overlay Topology Design Results Robustness 46 45.025 PSNR [dB] H=0.2 0.10 H=0.5 42 41.4 0.415 α 0.5 0.1 0.

Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 20 / 29 .Video-Aware Schedulers Video-Aware Schedulers S.

bottom). top) and ρ = 1. Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 21 / 29 .0 (rs = 1290kb/s.Video-Aware Schedulers Different frame types and losses 55 50 45 40 35 30 25 20 15 PSNR [dB] 0 5 10 15 20 25 Time [s] 30 35 40 55 50 45 40 35 30 25 20 15 PSNR [dB] 0 5 10 15 20 25 Time [s] 30 35 40 Figure: PSNR variation of a random peer versus time.6 (rs = 780kb/s. S. ρ = 0.

Video-Aware Schedulers The Strategy Improving chunk selection Different frame types have different importance in a video stream. We can assign priority to chunks transporting precious frames. With ω > 0 we assign a larger weight to more important frames. missing a b frame impairs only the decoding of one single frame. A frame loss may cause very different levels of degradation of the reconstructed video quality: missing an IDR frame impairs the video decoding until the next IDR. A weight q ω is given to every chunk encapsulating a frame. Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 22 / 29 . To avoid chopping frames into several chunks. Priority is assigned to frames based on the amount of degradation they might induce if lost: q is the number of subsequent frames that would be affected by the lost of that frame. S. we set one frame per chunk.

Video-Aware Schedulers The Strategy Improving chunk selection Different frame types have different importance in a video stream. Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 22 / 29 . With ω > 0 we assign a larger weight to more important frames. A weight q ω is given to every chunk encapsulating a frame. To avoid chopping frames into several chunks. A frame loss may cause very different levels of degradation of the reconstructed video quality: missing an IDR frame impairs the video decoding until the next IDR. we set one frame per chunk. S. missing a b frame impairs only the decoding of one single frame. Priority is assigned to frames based on the amount of degradation they might induce if lost: q is the number of subsequent frames that would be affected by the lost of that frame. We can assign priority to chunks transporting precious frames.

Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 23 / 29 .0 ω=0.10. Dmax = 5s and H = 0.0 30 35 40 45 Figure: PSNR for different peers and values of ω with ρ = 1.1. S. Pink by Aerosmith video sequence.Video-Aware Schedulers Results Slight Improvements 42 Pink 40 PSNR [dB] 38 36 34 32 5 10 15 20 25 PeerID ω=0.0 ω=2.5 ω=1.

5 ω=1. Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 24 / 29 . Dmax = 5s and H = 0.0 30 35 40 45 Paris Figure: PSNR for different peers and values of ω with ρ = 1.10.0 ω=0. Paris video sequence. S.Video-Aware Schedulers Results Slight Improvements II 40 38 36 PSNR [dB] 34 32 30 28 26 5 10 15 20 25 PeerID ω=0.1.0 ω=2.

0 ω=0. S.1.Video-Aware Schedulers Results Slight Improvements III 38 36 PSNR [dB] 34 32 30 28 26 5 10 15 20 25 PeerID ω=0. Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 25 / 29 .5 ω=1. Foreman video sequence. Dmax = 5s and H = 0.0 30 35 40 45 Foreman Figure: PSNR for different peers and values of ω with ρ = 1.10.0 ω=2.

Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 26 / 29 .Conclusions Conclusions S.

By carefully designing the overlay topology we can partially localize the traffic and improve the user QoE. S. Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 27 / 29 . system performance can be slightly improved in overloaded conditions. By prioritizing chunks that encapsulate valuable pieces of information at the scheduler level.Conclusions Conclusions We provided guidelines for the design of the overlay topology and the chunk scheduling algorithm.

By prioritizing chunks that encapsulate valuable pieces of information at the scheduler level. S. system performance can be slightly improved in overloaded conditions. By carefully designing the overlay topology we can partially localize the traffic and improve the user QoE. Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 27 / 29 .Conclusions Conclusions We provided guidelines for the design of the overlay topology and the chunk scheduling algorithm.

By prioritizing chunks that encapsulate valuable pieces of information at the scheduler level. S.Conclusions Conclusions We provided guidelines for the design of the overlay topology and the chunk scheduling algorithm. system performance can be slightly improved in overloaded conditions. By carefully designing the overlay topology we can partially localize the traffic and improve the user QoE. Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 27 / 29 .

Q&A Q&A S. Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 28 / 29 .

Traverso (Politecnico di Torino) Overlay Topology Design Tradeoffs August 2010 29 / 29 .Q&A Thank you for your attention! S.