Processing RangeMonitoring Queries in Mobile Databases

1
Processing Range-Monitoring Queries in Mobile
Databases
CS587x Lecture Department of Computer
Science Iowa State University
2
What is a mobile database?

• A mobile database is a set of mobile devices
• Decentralized these devices form a network by
  themselves
• Mobile ad hoc network (MANET)
• Mesh networks
• Centralized there exists a central server with
  which each mobile device can communicate

3
Characteristics

• Large number of mobile objects
• Continuous movement of mobile objects
• Limited battery power
• Limited communication bandwidth

4
Range-monitoring Query

• Retrieve mobile objects in a spatial region, and
• continuously monitor the population in the area

5
Examples of Range-Monitoring Queries
Q2
a
d
e
Q1
c
b
f
6
Examples of Range-Monitoring Queries
Q2
a
d
Q1
e
b
c
f
7
Some Potential Applications

• Tourist guiding
• Traffic control
• Digital battlefield vehicle tracking
• Wild animal tracking

8
Research Issues

• How to minimize location updates?
• Each update involves mobile communication cost
  and server processing cost
• How to minimize monitoring update cost?
• Query results keep changing

9
Solution I

• A simple solution
• Index the locations of mobile objects, and
• Update the index each time a mobile object moves
• Problems
• Excessive location updates
• Overwhelming server processing

10
Solution II

• Location estimation and trajectory indexing
• The object movement is estimated using segments
• The segments are indexed at server side
• Problems
• No real-time monitoring updates
• No accurate query results

11
Solution III

• Safe Region Approach
• Query rectangles are indexed at server side
• Safe regions are used to reduce location updates
• Problems
• Determining a safe region is computation-intensive
  
• Adding a new query requires to re-compute safe
  regions for all mobile objects

12
Safe Regions
Rectangular Safe Region
Q1
Q2
Q5
a
Q3
Q4
Circular Safe Region
13
Challenge

• How to provide
• accurate query results, and
• real-time updates?

14
Monitoring Query Management (MQM)
Q2
Q3
R1
R21
R22
Q1
R31
Q4
R42
R41
15
Highlights of MQM

• Domain is decomposed into many subdomains
• Each query is handled as one or more monitoring
  regions
• D-tree is used to index the domain and query
  decomposition
• Each object is assigned a resident domain
• An object reports server in the following two
  cases
• Crosses any query boundary
• Moves out of its resident domain

16
D-tree Domain Tree
D
D
Domain Node
Data Node
Domain D
17
D-tree Domain Tree
D1
D2
Domain Node
D1
D2
Data Node
D is split into D1 and D2
18
D-tree Domain Tree
D11
D11
D2
D12
Domain Node
D2
D12
Data Node
D is split into D1 and D2 D1 is split into D11
and D12
19
D-tree Domain Tree
D11
D11
D2
D121
Domain Node
D122
D2
D122
D121
Data Node
D is split into D1 and D2 D1 is split into D11
and D12 D12 is split into D121 and D122
20
D-tree Domain Tree
D11
D1
D2
Domain Node
D2
D121
D122
D11
D121
D122
D2
D is split into D1 and D2 D1 is split into D11
and D12 D12 is split into D121 and D122
Data Node
21
D-tree Domain Tree
D11
D21
D1
D2
Domain Node
D121
D122
D22
D11
D121
D122
D21
D22
D is split into D1 and D2 D1 is split into D11
and D12 D12 is split into D121 and D122 D2 is
split into D21 and D22
Data Node
22
D-tree operations

• Search
• Search the subdomain that contains an objects
  location
• Insert
• Insert a new query rectangle, which is handled as
  one or more monitoring regions
• Delete
• Delete an existing query rectangle

23
Server-Client Communication

• Messages from mobile objects
• updateQueryResult ( oid, r, p )
• requestResidentDomain ( oid, p )

24
Server-Client Communication

• Messages from server
• setResidentDomain ( oid, domain, MRs )
• addMonitoringRegion ( r )
• deleteMonitoringRegion ( r )
• splitDomain ( d, d1, d2 )
• mergeDomain ( d1, d2, MRs )

25
Handling Client Heterogeneity
Q1
Q6
Q3
Q2
a
Q5
Q7
Q4
Resident Domain
26
Computing a Resident Domain

• Given an objects position P and its processing
  capability N, its resident domain should

• contain position P, and

• be as large as possible, but

• contain no more than N queries

27
Domain and Query Decomposition
Q2
Q3
R1
R21
R22
Q1
R31
Q4
R42
R41
28
Domain and Query Decomposition
Q2
Q3
R1
R21
R22
Q1
R31
a
Q4
R42
R41
29
Domain and Query Decomposition
Q2
Q3
R1
R21
R22
Q1
R31
a
Q4
R42
R41
30
Binary Partitioning Tree (BP-tree)
D
domain node
D
data node
31
Binary Partitioning Tree (BP-tree)
D
d1
d2
d1
d2
32
Binary Partitioning Tree (BP-tree)
D
d21
d1
d2
d1
d21
d22
d22
33
Mobile Communication Cost
30
25
20
Safe Region
15
Number of messages sent by mobile objects
(millions)
MQM
10
5
0
10
20
30
40
50
60
70
80
90
100
Number of monitoring queries (thousands)
34
Server Processing Cost
1000
100
Safe Region
10
MQM
Number of index nodes accessed (millions)
1
0.1
10
20
30
40
50
60
70
80
90
100
Number of monitoring queries (thousands)
35
Concluding Remarks

• Power Conservation
• Mobile devices do not report their positions
  constantly
• Scalability
• Server does not track mobile objects and query
  results
• Reliability
• MQM does not use location estimation
• Adaptive
• More capable objects can save power by loading
  more queries at one time