Socially-Aware Distributed Hash Tables for Decentralized Online Social Networks

Preview:

Citation preview

Socially-aware DHTs for DOSNs 1

Socially-Aware Distributed Hash Tables for Decentralized Online Social Networks

Muhammad Anis Uddin Nasir, Sarunas Girdzijauskas, Nicolas KourtellisKTH Royal Institute of Technology

Telefonica Research

Socially-aware DHTs for DOSNs 2

Social Networks

Socially-aware DHTs for DOSNs 3

Social Networks

Socially-aware DHTs for DOSNs 4

Social Network Services

• Search

• Information Dissemination

• Storage

• Profile Management

• Third-Party Application

Socially-aware DHTs for DOSNs 5

Privacy and Scalability

Socially-aware DHTs for DOSNs 6

Decentralized Online Social Networks

Socially-aware DHTs for DOSNs 7

Research Question

“Can we design an overlay for decentralized online social networks that can support light

weight, low-cost devices such as mobile phones, web browsers"

Socially-aware DHTs for DOSNs 8

Decentralized Online Social Networks

• Low Management Overhead

• Scalability

• Bounded Degree

• Reliability and Security

Socially-aware DHTs for DOSNs 9

Distributed Hash Tables

Socially-aware DHTs for DOSNs 10

Distributed Hash Tables

0

0.25

0.5

0.75

Socially-aware DHTs for DOSNs 11

Social Networks on DHT

0

0.25

0.5

0.75

Socially-aware DHTs for DOSNs 12

Socially-aware Distributed Hash Tables

0

0.25

0.5

0.75

Socially-aware DHTs for DOSNs 13

Problem Formulation

• – where V is the set of vertices and E is the set of edges,

connecting the vertices.

• A– Neighbors of node i are all the nodes that share an edge

with node i

Socially-aware DHTs for DOSNs 14

Problem Formulation

• Social Ties

• Overlay Distance

• Optimization Problem

Socially-aware DHTs for DOSNs 15

Intuition

• Socially-Aware Distributed Hash Tables

• Gossip-based Algorithm– Initialization Phase– Refinement Phase

Socially-aware DHTs for DOSNs 16

Gossip-based Algorithm

• Initialization 0

0.25

0.5

0.75

Socially-aware DHTs for DOSNs 17

Neighbor Selection

• Direct: A node selects one of his friends in the social graph uniformly at random.

• Greedy: A node selects its friend with strongest tie.

• Smart: A node selects a node m uniformly at random from its top k strongest friends.

• Random: A node i selects a random other node.

Socially-aware DHTs for DOSNs 18

Gossip-based Algorithm

• Refinement 0

0.25

0.5

0.75

Socially-aware DHTs for DOSNs 19

Gossip-based Algorithm

0

0.25

0.5

0.75

• Convergence

Socially-aware DHTs for DOSNs 20

Gossip-based Algorithm

• Reduces Communication Overhead

• Improves Security

• Increases Reliability

• Improves Performance

Socially-aware DHTs for DOSNs 21

Experiments

• What is the tuning cost of the algorithm?

• How does node ordering impact algorithm convergence?

• What are the performance gains with respect to lookup latency and reliability?

Socially-aware DHTs for DOSNs 22

Evaluation

• Comparison with Symphony overlay

Socially-aware DHTs for DOSNs 23

Metrics

• Algorithm Tuning– Peer Selection Schemes

• Social Networks– Lookup Latency– Reliability

Socially-aware DHTs for DOSNs 24

Q1: Peer Selection Scheme

• Direct, Random and Smart Neighbor selection schemes perform similar

Socially-aware DHTs for DOSNs 25

Q2: Execution Ordering• Convergence is fast with ordered execution of the

algorithm

Socially-aware DHTs for DOSNs 26

Q3: Performance

• SD has less improvement due to lack of clustering in the social graph

Socially-aware DHTs for DOSNs 27

Q3: Reliability

• Reliability improves significantly in terms of finger table

Socially-aware DHTs for DOSNs 28

Q3: Reliability

• Significant improvement in the reliability in terms of connections

Socially-aware DHTs for DOSNs 29

Conclusion

• Socially-aware distributed hash tables improves the performance for decentralized online social network

• We propose a gossip-based algorithm that considers social ties to achieve social-awareness in DHT

• We show that our approach reduces the lookup latency by almost 30% and improves the reliability of the communication by nearly 10% via trusted contacts.

Socially-aware DHTs for DOSNs 30

Socially-Aware Distributed Hash Tables for Decentralized Online Social Networks

Muhammad Anis Uddin Nasir, Sarunas Girdzijauskas, Nicolas KourtellisKTH Royal Institute of Technology

Telefonica Research

Socially-aware DHTs for DOSNs 31

Further Read

• http://arxiv.org/pdf/1508.05591v2.pdf

Socially-aware DHTs for DOSNs 32

Dealing with Failures

• Social graphs evolve in a linear fashion

• Mobility or inactivity of peers, and peer or network failure cause instability

Socially-aware DHTs for DOSNs 33

Q1: Migration Cost

Recommended