Uniting the Real World and the Virtual World

Mobility-Supporting
Data Management for Location-Based
Mobile Systems (LBMS)
Manageability, Scalability, and Proximity
Generation Performance
of Server-Based Operations for
Location Information Management
Jim Wyse
Wireless Communications and Mobile Computing Research Centre (WCMCRC) Seminar
Series, Faculty of Engineering and Applied Science, Memorial University
Web-Based LBMS
Mobile Commerce (m-Commerce)
•
transactions through communication
channels that permit a high degree of
mobility by at least one of the
transactional parties.
Location-Based m-Commerce
•
m-business with location-referent transactions:
transactions in which the geographical
proximity of the transactional parties is a
material transactional consideration.
•
Critical technological capability: location
awareness.
“Location-Awareness”
The capability to obtain and use the geo-positions
of the transactional parties to perform one or more
of the CRUD (create, retrieve, update, delete)
functions of data management.
The Data Management Problem
•
Location-referent transactions are supported by
proximity queries: What is my proximity to a
goods-providing (or service-offering) location in
a specified category?
•
A proximity query bears criteria that reference
static attributes (e.g., hospital) and dynamic
attributes (e.g., nearest).
•
Proximity queries are burdensome to servers
using conventional query resolution approaches
Proximity Generation – An Example
The Client-Based i-DAR Prototype
(Architecture: Client-Based Functionality, Server-Based Locations Repository)
Web-Based i-Prox Prototype
(Architecture: Functionality and Locations Repository are both Server-Based)
Selected i-Prox Implementations
1: Small Craft Harbours (Marine Services)
2: Smart Bay (Real-time Weather Conditions, etc.)
3: Public Libraries (Free Wireless Internet)
4: Municipalities (Information, Services)
5: Town of Placentia
Other Proximity Generators
Weblocal
Yellow Pages
foursquare
GEOS IERCC
WiGLE
Under the Hood
. . . meanwhile, back at the ‘server’
Locations Server and Repository
Conventional ‘Enumerative’ Methods
A.
Select locations in targeted business category.
B.
Calculate user-relative distances to selected locations.
C.
Sort selected locations by user-relative distance.
D.
Populate the user’s proximity with the ‘k’ nearest locations.
Variations: (1) B, C, D, and then A; (2) Range-based selection
Methods from Computational Geometry: Chevaz et al. (2001), Gaede
and Guther (1998).
The Problem (. . . and a Solution?)


Method-(The Problem)
and Method-
(The Solution)
4,500
3,500
3,000
2,500
2,000
1,500
1,000
500
Repository Size (thousands of locations)
49
47
45
43
41
39
37
35
33
31
29
27
25
23
21
19
17
15
13
11
9
7
5
3
0
1
Query Resolution Time (ms)
4,000
Linkcell Transformation
Geographical Space  Relational Space
Location-Aware Linkcell Method
•
Transforms mu’s position (47.523° N, 119.137° W) into a
linkcell (N47W119).
•
Initiates a search spiral pivoting clockwise around mu’s
linkcell: {N48W119, N48W118, N47W118, N46W118,
N46W119, N46W120, N47W120, N48W120, …}
•
Permits large numbers of locations to be excluded as
proximity portal candidates.
•
Requires an appropriate linkcell ‘size’ (S) to give superior
performance.
Linkcell Construction
Location Li appears in relational table named for X  ‘N’[SL + 3*S]‘W’[EL + 2*S]
For SL of 20°N, EL of 050°W, and S of 1°, we get:
Relational Table for Li: N[20+3*1]W[50+2*1] = N23W052
Proximity Generation: Performance
Re s ults e t Comple tion Time s
Single Category 100,000-Location Repository
Time
Query
QueryResolution
(ms)(ms)
Resolution Time
1,600
1,200
800
400
0
0.000
0.002
0.004
Linkcell
Size
Linkcell Size
(S)
0.006
0.008
Linkcell Performance Analyzer (LPA)
S for Optimal Performance?
200
Query Resolution Time (milli-seconds)
180
160
140
120
100
80
60
40
20
SL
SP
SM
SU
0
0.000 0.002 0.004 0.006 0.008 0.010 0.012 0.014 0.016 0.018 0.020 0.022 0.024 0.026 0.028 0.030
Linkcell Size, S (°'s)
Optimal Linkcell Size, S
‘Brute Force’ or Solve ….
PTC(S) = 1 – (1 –
n
N/CS  0.6
/N)
TC
...
(A)
. . . . for relational table name increments:
‘N’[SL + 3*S]‘W’[EL + 2*S] = (for ex. N23W052)
nTC
is the number of locations in category, TC,
N is total number of locations, and
CS is the number of linkcells of size, S, created
from the N locations.
MCRs and SCRs
• Multiple Category Repositories (MCRs)
• Single Category Repositories (SCRs)
• Equation (A) applies to MCRs but not to SCRs
• For SCRs, nTC = N  PTC(S) = 1, for all S.
Single Category Repositories
For SCRs, previous research hypothesized that
optimal values are given by:
P (S) = 1 – (1 – S2/4A)N  0.6 . . . (B)
where A is the total geographical coverage,
S is the linkcell size, and
N is the number of locations.
Linkcell Method (SCR) Reformulation
 Linkcell Construct
. . . from:
. . . to:
 Linkcell Optimization
. . . from:
P (S) = 1– (1 – nTC/N)N/CS
. . . to:
P (S) = 1– (1 – S2/4A)N
Figure 6
S = 0.0001
S = 0.001
S = 0.01
S = 0.1
S for P = 60%
100,000
(milliseconds)
Resultset Competion Time (Log Scale)
1,000,000
10,000
1,000
100
5
10
15
20
25
30
35
40
45
50
55
60
Repository Size
(thousands of locations)
65
70
75
80
85
90
95
100
Locations Repository: Scenario A
1000
950
900
Query Resolution Time (milli-seconds)
850
800
750
700
650
600
550
500
450
400
350
300
250
200
150
100
50
SM
SU
SL SP
0
0.000 0.002 0.004 0.006 0.008 0.010 0.012 0.014 0.016 0.018 0.020 0.022 0.024 0.026 0.028 0.030
Linkcell Size, S (°'s)
Locations Repository: Scenario B
200
Query Resolution Time (milli-seconds)
180
160
140
120
100
80
60
40
20
SL
SP
SM
SU
0
0.000 0.002 0.004 0.006 0.008 0.010 0.012 0.014 0.016 0.018 0.020 0.022 0.024 0.026 0.028 0.030
Linkcell Size, S (°'s)
Candidate ‘S’ Determination Methods
Interval of Optimality (SL, SU)
Four ‘S’ Candidates
SP:
P (S) = 1– (1 – S2/4A)N
(Probabilistic)
SL:
S = (A/N)1/2
(Equi-Areal)
SU:
S = 3 (A/N)1/2
(Spiral Avoidance)
SM:
S = 2 (A/N)1/2
(Optimality Interval Median)
Proximity Generation Performance
Scenario B: 50,000-Location Repository
Linkcell
Size
Proximity
Generation
Performance
(milliseconds)
SL: Equi-Areal
0.00447
50
SP: Probabilistic
0.00484
48
SM: Opt. Interval Median
0.00894
46
SU: Spiral Avoidance
0.01341
66
Linkcell Determination
Method
Unconstrained Enumerative Method: 121,500 ms (approx. 2 minutes or 2600X)
Proximity Generation
Repository Size Variations
Proximity Generation
Areal Size Variations for 50,000-Location Repository
Proximity Generation
Areal Size Variations for 100,000-Location Repository
Conclusion
SM: Optimality Interval Median
• Flattest proximity generation profile (scalability)
• Lowest proximity generation profile (performance)
• Easily determined (manageability)
Research Outputs
Articles – Professional/Academic Press
Mobile Computing: Concepts, Methods, Tools, and Applications (2009)
Advanced Principles for Improving Database Design, Systems Modeling,
and Software Development (2009)
Handbook of Research on Innovations in Database Technologies and
Applications: Current and Future Trends (2009)
Journal Articles
International Journal of Wireless and Mobile Computing (2009)
Journal of Database Management (2006)
International Journal of Mobile Communications (2003)
Patents
Canada 2010 - Optimization
United States 2004 - Linkcells
Mobility-Supporting
Data Management for
Location-Based Mobile Systems
Jim Wyse
www.busi.mun.ca/jwyse
Thank you!!
Wireless Communications and Mobile Computing Research Centre (WCMCRC) Seminar
Series, Faculty of Engineering and Applied Science, Memorial University, February 2011