A LocatingFirst Approach for Scalable Overlay Multicast - PowerPoint PPT Presentation

1 / 28
About This Presentation
Title:

A LocatingFirst Approach for Scalable Overlay Multicast

Description:

M.A. Kaafar, IWQoS 2006. Application Layer Multicast (ALM) Relies ... M.A. Kaafar, IWQoS 2006 ... about Intermittent connectivity networks (DTN) ... – PowerPoint PPT presentation

Number of Views:19
Avg rating:3.0/5.0
Slides: 29
Provided by: cmra1
Category:

less

Transcript and Presenter's Notes

Title: A LocatingFirst Approach for Scalable Overlay Multicast


1
A Locating-First Approach for Scalable Overlay
Multicast
  • Presented by Mohamed Ali Kaafar.
  • Co-authors T. Turletti, W.Dabbous.

2
Outline
  • Background and motivations
  • LCC overview
  • The locating process
  • Clustering
  • Experiments results
  • Conclusions

3
Application Layer Multicast (ALM)
  • Relies on end systems only
  • Pros
  • Ease of deployment
  • Cons
  • Lower bandwidth efficiency and higher end-to-end
    delay
  • Heavy control overhead
  • Challenges for large groups

4
Why is self-organization hard?
  • Dynamic changes in group membership
  • Members join and leave dynamically, fail
  • Limited knowledge of network conditions
  • Members do not know delay to each other when they
    join
  • Overlay must self-improve as more information
    available
  • Dynamic changes in network conditions
  • Delay between members may vary over time due to
    congestion

5
What is LCC?
  • Locating-First Underlying Topology-aware
  • Practical No landmarks, no virtual coordinates
    systems,
  • Clustering hierarchical approach

6
LCC, QoS aspects enhancements?
  • Scalability Less refinement procedures,
  • Reduce management overhead and delivery tree
    changes.
  • Efficiency Shorter convergence time.
  • Tradeoff Locating Overhead messages/time.

7
The locating process
d(L,n)
d(L,k)
L
n
k
Level 1
Level 2
Level 3
  • A set of non overlapping and exponentially
    increasing levels, represented by intervals
  • ri, ri1, where ri Alpha ei-1 for i gt 1 and
    r0 0

8
Locating scenario, Request
Newcomer
d(L,n)
Boot node
Queried nodes
Level 1
Level 2
Level 3
9
Locating scenario (contd)
  • Queried nodes answers are filtered
  • Only queried nodes whose distances to n are less
    than d(L,n) are selected.
  • List of candidates
  • The first ranked node is the next requested node
  • Recursively repeat the request procedure.
  • The list is updated at each request.

10
Selective Locating process
Level 1
Level 2
Level 3
S.A 5
S.A 4
S.A 3
S.A 2
S.A 1
S.A i /
Newcomer
Requested node
Representative node
11
Clustering process
  • If locating ends with no cluster leader in scope
    (Rmax),
  • create new cluster,
  • Else join one or a few others.
  • Clusters define the top-level links,
  • If cluster leader
  • Or Edge node !

12
LCC architecture
Top-level Delivery Tree
Rmax
Cluster Leader
Edge Node
Top-level Link
Intra-Cluster Link
Cluster member
13
Clustering
  • Define Priority vectors (PV) used to maintain a
    nodes rank.
  • PV lt fmax 1/DL T 1/CLMigrated gt
  • Shared local cache, to construct a proactive
    rescue plan. Eligible nodes in a cluster are
    sorted in the local cache.

14
Clustering
  • Network or membership changes recovery
  • Recursive Recovering algorithm to detect and
    repair migration.
  • A potential leader node asks subsequent nodes in
    the cache to join its cluster,
  • And triggers a recovering procedure for migrated
    nodes that are not in its scope.

15
Performance Evaluation
  • Linux distribution, wrappers for vic, rat, vlc
    (under test).
  • http//www-sop.inria.fr/planete/software
  • Particular real-world experimentations
    PlanetLab.
  • 212 nodes.
  • Simulations Scalability
  • 2000.. 10000 overlay nodes.
  • Comparison
  • OMNI (Overlay routers)
  • ZigZag (Hierarchical)
  • Flat MeshTree

16
Metrics
  • Scalability,
  • Control overhead during data distribution
  • Control overhead during overlay joining
  • Link Adjustment Rate
  • Efficiency,
  • Convergence Time

17
(Sim) Overhead
18
(PL) Locating Overhead
19
(Sim) Adjustment Rate
20
ARDP
  • Average Relative Delay Penalty
  • 1/(N-1) . ?(i1, N-1) d(s,i) / d(s,i)
  • Incurred delay due to overlay

21
(Sim-PL) Convergence Time
22
Conclusions
  • Practical Scalable overlay Multicast scheme.
  • Not too much overhead while locating first.
  • Better adaptability to underlying nw changes.
  • ALM is not only for small groups ?
  • LCC is very suitable for large scale applications.

23
What next?
  • Automatically tune nodes scopes, stop criterion,
  • More real life tests.
  • What about Intermittent connectivity networks
    (DTN),?

24
Annexes
25
(Sim) ARDP
Zoom
26
(Sim) Protocol Robustness
27
(PL-Sim) Locating Accuracy
28
(PL-Sim) Locating Accuracy
Write a Comment
User Comments (0)
About PowerShow.com