IEG5270 Advanced Topics in P2P Networking P2P streaming protocols and analysis

1
IEG5270 Advanced Topics in P2P NetworkingP2P
streaming protocols and analysis

• Dah Ming Chiu
• Chinese University of Hong Kong

2
Coolstreaming

• XY Zhang, JC Liu, B Li, P Yum, Coolstreaming/DONe
  t a data-driven overlay network for efficient
  live media streaming, Infocom 2005
• DONet Data-driven Overlay Netwrok
• Coolstreaming Cooperative Overlay Streaming
• Background
• XY Zhang was a CUHK M.Phil student
• Built a prototype p2p streaming system, using
  Python (2000 lines)
• Experimented on PlanetLab
• Deployed in May 30, 2004 during European soccer
  finals
• Total 30000 downloads, 4000 simultaneous users
• First large scale p2p streaming experiment
  showed feasibility

3
System diagram of Coolstreaming software
4
Low overhead
5
Continuity playback performance
6
Performance for different streaming rates
7
Compare to tree-based experiment
8
Drastic improvement over tree-based
9
Peer dynamics
10
2-3 years later

• P2P streaming of live video (e.g. TV) is becoming
  established services (at least in China)
• PPlive, PPstream, UUSee
• PPlive claims to have 75m installed base, and 20m
  active users
• See http//www.sigcomm.org/sigcomm2007/p2p-tv/
• More programs are legitimate content (purchased)
• Major issues
• Managing service quality, QoS integration with
  CDN?
• Delivering advertisement
• Managing relationships with ISPs

11
BiToS BitTorent Streaming

• A Vlavianos, M Iliofotou, M Faloutsos, Bitos
  Enhancing bit-torrent for supporting streaming
  applications IEEE Infocom 2006
• They ask can BT (with small change to its chunk
  selection strategy) be used for streaming?
• The modified BT by adapting chunk selection
• Select chunks in a sliding window
• Still use rarest first inside the window
• Also create a priority set of chunks to get
• Their work is based on simulation
• Their work inspired YPs work

12
BiToS design
13
BiToS result
14
A simple p2p streaming model

• YP Zhou, DM Chiu and J Lui, A simple model for
  p2p streaming, to appear in ICNP 2007
• One of very few performance models of p2p
  streaming system
• P2p streaming is different from p2p file
  downloading
• Rate (throughput) is given
• Need good continuity
• Few freezes
• Few jumps
• Need low start-up delay if possible

15
Simple peer model

• M homogeneous peers
• Each has a playback buffer
• In each time slot, the server uploads one peer
  with probability 1/M
• Without P2P network
• continuity p(n)
• 1/M

server
playback
1/M
1/M
M peers
1/M
16
Sliding window

• Each peers buffer is a sliding window
• In each time slot, each peer downloads from a
  random neighbor
• q(i) the probability Bufi gets filled

p(1)1/M p(n)?
timet
sliding window
t1
p(1)1/M p(i1) p(i) q(i)

• q(i) w(i)h(i)s(i)
• w(i) peer wants to fill Bufi
• h(i) the selected peer has the content for
  Bufi
• s(i) Bufi determined by chunk selection
  strategy

17
Chunk selection strategies

• Greedy
• try to fill the empty buffer closest to playback
• Rarest First
• try to fill the empty buffer for the newest
  chunk
• since p(i) is an increasing function, this means
  Rarest First

18
Chunk selection strategy - cont

• Greedy
• p(i1) p(i) p(i) (1- p(i)) (1- p(1) - p(n)
  p(i1))
• Rarest first
• p(i1) p(i) p(i) (1- p(i))2
• Also studied
• continuous forms for these difference equations
• simulation

19
Which strategy is better?

• What do you mean by better?
• Playback continuity p(n) as large as possible
• Start-up latency ? p(i) /R as small as possible,
  for given streaming rate R
• Given buffer size (n) and large peer population
  (M)
• Rarest first is better in continuity!
• Greedy is better in start-up latency

20
Numerical result

• M1000
• N40
• In simulation,
• neighbors60
• Uploads at most 2 in each time slot

21
Mixed strategy

• Partition the buffer into 1,m and m1,n
• Use RF for 1,m first
• If no chunks available for download by RF, use
  Greedy for m1,n
• Difference equations become
• p(1) 1/M
• p(i1) p(i) p(i) (1- p(i))2
  for i 1,,m-1
• p(i1) p(i) p(i) (1- p(i))(1- p(m)- p(n)
  p(i1)) for i m, n-1

22
Comparison

• For different buffer sizes
• Mixed achieves better continuity than both RF
  and Greedy
• Mixed has better start-up latency than RF

23
Closer look with simulation

• Simulate 2000 time slots
• Continuity is the average of all peers
• Continuity for Mixed is more consistent, as well
  highest

24
Adapting m to unknown population

• Adjust m so that p(m) achieves a target
  probability (e.g. 0.3)
• In simulation study, 100 new peers arrive every
  100 slots
• m adapts to a larger value as population increases