Peer Clustering and Firework Query Model

1
Peer Clustering and Firework Query Model
Cheuk-Hang Ng and Ka-Cheung Sia Department of
Computer Science Engineering The Chinese
University of Hong Kong chng,kcsia_at_cse.cuhk.edu.
hk
Multimedia Information Processing
Labhttp//www.cse.cuhk.edu.hk/miplab
This poster presents two new strategies for
information retrieval over the Peer-to-Peer (P2P)
network, called Peer Clustering and Firework
Query Model. The Peer Clustering technique
organizes peers sharing similar properties
together, thus, data inside the P2P network are
arranged in a fashion like Yellow Pages. In
Firework Query Model, the query first walks
around the network from peer to peer randomly,
once it reaches the target cluster, the query
message is broadcasted to all peers inside the
cluster much like an exploding firework.
Key Problems In brute force search algorithm,
broadcasting the query across the network causes
Peer Clustering
Cluster peers with similar characteristic
together On top of the original network, we form
deliberate attractive links to group peers
together (Figure 1) Organize data inside the
network in a Yellow Page fashion which makes
query more systematic

• Heavy network traffic
• Waste resources in handling irrelevant queries

• Contributions

Proposed the Peer Clustering and Firework Query
Model improve query efficiency
Firework Query Model

Use our clustered network to route the query
intelligently rather than by brute force
search Query first randomly walks around the
network, once it reaches the target cluster, the
query is broadcasted as a firework explosion
(Figure 2) Avoid unnecessary traffic while fully
utilize each query message

• Reducing the network traffic
• Increasing the searching performance

Experiment Investigates the effect of query
efficiency subject to

• Increasing the network size
• Increasing the Time To Live (TTL) of query
  message

Number of query messages against number of peers
Number of query messages against TTL of query
message
attractive links
Query efficiency against number of peers
Recall against TTL of query message
Figure 1 Peer Clustering
Figure 2 Firework Query