Continuous Monitorin..

Kyriakos Mouratidis, Spiridon Bakiras, Dimitris
Papadias
SIGMOD 2006
1

Motivation

Preliminaries

Method
◦ TMA(Top-k Monitoring Algorithm)
◦ SMA(Skyband Monitoring Algorithm)

Experimental evaluation

Conclusion
2


The existing methods are inapplicable to
highly dynamic environments involving
numerous longrunning queries.
This paper studies continuous monitoring of
top-k queries over a fixed-size window W of
the most recent data.
3

f(X1,X2)=X1+2*X2
(0.2,1)
(0.6,0.8)
(1,0.7)
4
K-skyband :Returns those objects that are
dominated by at most K-1 other objects.
5
6


f(X1,X2)=X1+2*X2
Top-1
7


F(X1,X2)=X1-X2
Top-2
8
P1,P2 expire
 P3,P4 arrive
Search influence list-> P3 has maxscore
P3 become the result of top-1 query

9


P3 expire
P5 arrive
invokes the top-k computation module
10

SMA applies the reduction from top-k to
k-skyband queries in order to avoid
computation from scratch when some results
expire.
11

DC(dominance counter)
2-skyband
When DC reach 2,
then delete.

12

P9 arrive
13

P9 arrive
(2)
(1)
(2)
14

P9 arrive
15

SMA is expected to be faster than TMA, since
it involves less frequent calls to the top-k
computation module.
16
17

TMA re-computes the result from scratch,
whereas SMA maintains a superset of the
current answer in the form of a k-skyband, in
order to avoid frequent recomputations.
18