2005/11/09
Continuous Queries in P2P Networks
Motivation
Motivation - Cardinality
How many people are currently listening POP
music?
Rock
Classic
Classic
POPPOP
Result =2
Motivation – Top-K
“Because of you” by Kelly Clarkson
What is the Top-2 Songs?
“Because of you” by Kelly Clarkson
“Wake Me Up When September Ends” by Green Day
“Wake Me Up When September Ends” by Green Day
“Shake It Off”by Mariah Carey
Motivation – Top-K
“Because of you” by Kelly Clarkson
What is the Top-2 Songs?
“Because of you” by Kelly Clarkson
“Wake Me Up When September Ends” by Green Day
“Wake Me Up When September Ends” by Green Day
“Shake It Off”by Mariah Carey
Motivation – Social Network
Rock
Classic
Classic
POPPOP
I want to make friends who have similar
interests as I have
Motivation – Social Network
Rock
Classic
Classic
POPPOP
I want to make friends who have similar
interests as I have
Motivation – Social Network
Rock
Classic
Classic
POPPOP
I want to make friends who have similar
interests as I have
Motivation – Social Network
Rock
Classic
Classic
POPPOP
I want to make friends who have similar
interests as I have
Motivation - Ontology
Search “Shake Your Bon Bon” by Ricky Mart
in
Motivation - Ontology
Search “Shake Your Bon Bon” by Ricky Mart
in
Sorry, I have none
I have “Shake Your Bon Bon”
Dude, get away from me
I do have this song but its
name is “Martin’s Ass”
Motivation - Ontology
Search “Shake Your Bon Bon” by Ricky Mart
in
The result of exact matching = 1By we want to get the actual result 2
Ontology
Continuous Queries
Cardinality Top-K Social network Ontology
Cardinality
The state-of-the-art Aggregation in P2P system
Montresor et al. DSN'04 Epidemic, adaptive
Aggregation with streaming data Das et al. VLDB04
Global knowledge of frequent items
Cardinality (Cont’d.)
As far as we know, there is no study focus on this issue in P2P environment with streaming data
Progress after our summer presentation Use statistics distribution to estimate changes
Top-K
The state-of-the-art Traditional Top-K problem
Combine information for database systems [Fagin]
Approximation on data streams Proposed for data streams under guaranteed
tolerance, but can’t be deployed to P2P.
Super peer based Top-k in P2P Iteratively query
Locality was mentioned, but the method is straight-forward
Top-K (Cont’d.)
Assumptions Based on superpeer-structured P2P networks du
e to the heterogeneity of peers Each super-peer would maintain a routing table
and some metadata for Top-k query
P
P
P
P
P
PP
PSP2
SP1
SP3
SP4
Top-K (Cont’d.)
Problem Query whom? →Locality
space Interest
Query routing (routing table) Reduce the size of table Minimize the communication
peer’s update occurs load balance
T2SP4
T1SP3
T1SP2
TypeNode
Social network + Ontology
The state-of-the-art Similarity computing
Compute the similarity between two nodes/peers, and fix-point scores will be assigned
Social network + Ontology
Our thoughts Using some hierarchical domain structures Ontology / classification
Avril LavigneA: a1, a2B: a3, a4C: a1, b1, b2
Rock
Bon Jovi
a1 a2 a3 a4 b1 b2 b3
Thank You!
What is ontology
A formal, explicit specification of a shared conceptualization
Object
attribute
Objectrelation
Class
Back
Example
Example
Ballet Swim Kung Fu
Example
Ballet Swim Kung Fu
Back