oStream: Asynchronous Streaming Multicast in ApplicationLayer Overlay Networks

1
oStream Asynchronous Streaming Multicast in
Application-Layer Overlay Networks

• Yi Cui, Baochun Li, and Klara Nahrstedt

IEEE Journal on Selected Areas in Communications,
vol. 22 (1), Jan, 2004
Presented by Yuk
2
Outline

• Introduction
• Temporal Dependency Model
• Algorithms
• Analysis Scalability and Efficiency
• Performance Evaluation
• Conclusion and Comment
• Q A

3
Introduction

• Fundamental challenge of on-demand media
  distribution is unpredictability
• Asynchrony
• Nonsequentiality
• Burstiness
• Previous IP-Multicast-based solutions
• Repeat the same media content on different
  channels over time
• Clients are synchronized at the price of service
  delay
• Proposed solution Asynchronous Multicast

4
Introduction
5
Temporal Dependency Model
6
Hierarchical Stream Merging (HSM)
D. Eager, M. Vernon, and J. Zahorjan, Minimizing
bandwidth requirements for on-demand data
delivery, IEEE Trans. Knowl. Data Eng., vol. 13,
pp. 742757, Sept.Oct. 2001.
7
Asynchronous Multicast (AM)
8
Qualitative Comparisons

• Asynchronous group
• Purely end-host based
• Sequentialized sources

9
Problem Formulation
Given a MDG, the optimal solution for MDT, i.e.,
to minimize the overall transmission cost of
media distribution, is to find the minimal
spanning tree (MST) on MDG.
10
Algorithms MDT-Delete
11
Algorithms MDT-Insert
12
Some Theorems
Y. Cui, B. Li, and K. Nahrstedt, oStream
Asynchronous streaming multicast in
application-layer overlay networks, Dept.
Comput. Sci., Univ. Illinois at Urbana-Champaign,
Urbana, IL, Tech. Rep. UIUCDCS-2002-2289/UILU-ENG-
2002-1733, 2003.
13
Practical Issues

• Content Discovery Service
• MDT algorithms require knowledge of all its
  predecessors and successors
• Degree Constrained MDT
• Constrain the outbound degree
• Modified MDT problem
• NP-complete!

14
Practical Issues (contd) Simplified Session
Switching
15
Scalability Server Bandwidth Savings
16
Scalability (HSM)
17
Scalability (AM)
18
Scalability Plots
19
Efficiency Link Bandwidth Reduction
20
Efficiency Plots
21
Performance Evaluation

• A single CBR video distribution
• Video length, T 1hr
• Simulation time 12 hrs
• Topology
• k-ary tree
• Router-level (transit-stub)
• AS-level (power-law)

22
Server bandwidth consumption
23
Link cost
24
(No Transcript)
25
(No Transcript)
26
Operation Complexity
27
Conclusion

• Concept of AM
• Take advantage of the strong buffering
  capabilities of end hosts
• Scalability
• Required server bandwidth lt the theoretical lower
  bound of traditional IP-multicast
• Efficiency
• The benefit overshadows the topological
  inefficiency w.r.t link cost

28
Comment

• Simple solution provide good results
• (simply buffering at end-hosts)
• In-depth analysis and Extensive evaluation
• The main drawback
• Outbound gt inbound, not realistic

29
Q A

• Thank you.