QUERY MELTING

A New Paradigm for GIS Multiple Query Optimization

Haifa Elsidani Elariss, Souheil Khaddaj and Darrel Greenhill

Faculty of Computing, Information Systems and Mathematics, Kingston University London

Kingston upon Thames, U.K.

Keywords: Query Optimization, Query Melting, Proximity Analysis, Dynamic Complex Queries, Large-scale GIS

Servers, Visual Query Languages and Mobile GIS.

Abstract: Recently, non-expert mobile-user applications have been developed to query Geographic Information

Systems (GIS) particularly Location Based Services where users ask questions related to their position

whether they are moving (dynamic) or not (static). A new Iconic Visual Query Language (IVQL) has been

developed to handle proximity analysis queries that find k-nearest-neighbours and objects within a buffer

area. Each operator in IVQL queries corresponds to an execution plan to be evaluated by the GIS server.

Since commonalities exist between the execution plans, the same operations are executed many times

leading to slow results. Hence, the need arises to develop a multi-user dynamic complex query optimizer

that handles commonalities and processes the queries faster especially with the large-scale of mobile-users.

We present a new query processor, a generic optimization framework for GIS and a middleware, which

employs the new Query Melting paradigm (QM) that is based on the sharing paradigm and push-down

optimization strategy. QM is implemented through a new Melting-Ruler strategy that works at the low-

level, melts repetitions in plans to share spatial areas, temporal intervals, objects, intermediate results, maps,

user locations, and functions, then re-orders them to get time-cost effective results, and is illustrated using a

sample tourist GIS system.

1 INTRODUCTION

During the last decades, computer applications have

been deployed to manage spatial data with

Geographic Information Systems. GIS are computer-

based tools mostly used to handle the geo-features of

the real world. They provide the capability to input,

store, manipulate, analyze, retrieve, transform, and

display geographical data related to earth surface

and its events as well as to measure aspects of

geographic phenomena and processes. With the

emergence of mobile technologies, the use of GIS

Geo-data by mobile devices becomes very common.

Mobile GIS is the combination of systems which

include mobile devices, Global Positioning Systems

(GPS), wireless communication, and GIS software.

In tracking systems, they provide the facility to track

employees and children in the backyards (Ladd,

2005; Kim, 2007). In road networks, they provide

the ability to organize public transportation

(Repenning, 2006), query moving objects (Guting,

2006), guide tourists (Beeharee, 2007), and exploit

photos (Beeharee, 2006). In data mining, they play a

major role in trajectory pattern recognition of

moving users (Gianotti, 2007; Andrienko, 2007).

Mobile GIS are typically used in tourist and

navigation systems for Proximity Analysis which

includes querying the k Nearest Neighbor (kNN) and

finding the facilities within a buffer area. The

existing Mobile GIS applications have textual or

menu-driven environment, do not provide a user

friendly environment, and in some applications they

are aimed at expert users. Moreover they allow the

user to formulate one query at a time only. Thus, a

new iconic visual query language (IVQL) has been

developed in (Elsidani Elariss, 2006a, 2006b) to

provide the mobile user with the facility to formulate

a visual query using expressive icons and send

complex queries, therefore reducing the overheads

significantly.

Query Optimization includes a list of tasks that

are executed in a particular order for processing a

simple query, which form an execution plan. Having

a complex query made up of a number of simple

Elsidani Elariss H., Khaddaj S. and Greenhill D. (2009).

QUERY MELTING - A New Paradigm for GIS Multiple Query Optimization.

In Proceedings of the 11th International Conference on Enterprise Information Systems - Databases and Information Systems Integration, pages 82-90

DOI: 10.5220/0001960700820090

 SciTePress

queries requires the Query Processor to execute a

number of corresponding plans, combine their result

in one map, and send the result map to the mobile

user. If one or more of the simple queries are

dynamic, the processor repeats the same steps for

every new time instance that corresponds to the

current user location. These repetitions raise the

need to develop a Query Processor for Dynamic

Complex Queries which aims to eliminate all

repetitions, share intermediate results, facilities,

spatial areas, time intervals, and generate one Global

Execution Plan for each dynamic complex query.

2 RELATED WORK

To deal with the different types of queries, different

query optimization strategies have been considered.

In this section we briefly describe some of them

showing some of their advantages and limitations.

2.1 The Sharing Paradigm

The existing query optimization strategies that use

the sharing paradigm are concerned with processing

multiple queries in data analysis applications. An

example is the Virtual Microscope Processor

(Andrade, 2001, 2002c, 2002d, 2006; Afework,

1998). The runtime system was designed for shared-

memory multi-processors. It tackles queries that are

user-defined, allows the input data to be shared, and

the query results to be reused by other queries. It

aimed at optimizing query processing by (1)

maintaining intermediate data structures generated

by queries for intermediate results, (2) caching input

data in memory, and (3) providing support for multi-

threaded execution where each query is executed as

a thread. Caching methods are used to store query

results in the memory in the aim to speed up the

execution of queries (Andrade, 2002a, 2002b, 2002e,

2003). These results along with input data can be

used by other queries to produce new results. When

caching is implemented at the server side, multiple

clients can share the query results.

The experiment was conducted on an 8-processor

Symmetric Multiprocessing machine, running Linux

Kernel version 2.4.3. Two scenarios were considered

in order to examine the performance of the runtime

system and for each scenario two executions were

done; each considered as a case making a total of 4

cases. In the first scenario, 16 clients were emulated.

The calculated value of the overlap index was large

70% reflecting a high overlap among queries. In the

second scenario, 8 clients were emulated making a

relatively small overlap index of 59%. Two

executions were done for each scenario. In the first

one, the Data Store Manager was ON, hence

maintained intermediate results. In the second one, it

was OFF, hence did not maintain intermediate

results. The execution time in seconds of each case

was recorded. An evaluation of the performance was

conducted where results showed that a better

performance was obtained when maintaining

intermediate results. In the high overlap index case,

the execution time decreased by about 30% to 40%.

In the low one, it decreased by about 18%. Results

also reported that the query execution time

decreased as the number of threads increased and

that the query evaluation time decreased as the size

of the data store manager cache increased. The

execution time decreased by 38%.

2.2 The Push-down Strategy

The existing query optimization strategies that use

the push-down approach are concerned with

processing execution plans where the order of

execution of the operators affects the execution time

without affecting the output result of the query such

as having both the Selection and the Join operators

in the same execution plan. In some cases and

depending on the data records, the execution of one

the Selection operator before the Join operator

produces faster results than the execution of the Join

operator before the Selection operator, and vice

versa. The push-down strategy is based on swapping

operators in an execution plan in order to get faster

results. An example of the push-down paradigm is

described in (Elmongui, 2005, 2006) which deals

with the optimization of multiple predicate spatio-

temporal queries for applications such as to find in

which region of a continuously monitored city the

number of suspects is greater than the number of

police officers. The proposed idea is building a

uniform adaptive query optimization framework that

includes: selectivity estimation, cost estimation,

adaptive query optimization model, and an extension

of query optimization to cover multiple multi-

predicate spatio-temporal queries. The Spatio-

Temporal Histogram (ST-Histogram) was used to

estimate the selectivity of a continuous spatio-

temporal query operator that is sent from the query

executer to the histogram manager periodically in

form of statistics. An extension was proposed in

order to accommodate multiple feedbacks especially

the spatial relationship between queries. A data-to-

plan grid was proposed as a framework to prevent

executing the same query plan on all location

QUERY MELTING - A New Paradigm for GIS Multiple Query Optimization

updates. The data-to-plan grid directs each location

update to a relevant plan. A dynamic plan formation

mechanism is used to provide the facility to add,

remove, or reshuffle operators in the evaluation plan.

Also, an extension of query optimization to handle

multiple multi-predicate spatio-temporal queries was

proposed based on sharing sub-plans between

multiple multi-predicate queries.

2.3 Both Sharing and Push-down

The existing query optimization strategies that use

the sharing paradigm and the push-down strategy are

concerned with processing scalable incremental

spatio-temporal queries. An example is in (Mokbel,

2003, 2004b, 2004c, 2005a, 2005b) which deals with

location-aware services. The proposed idea is

sharing the underlying space, operator, and objects.

The first type is sharing the underlying space by the

queries that find whether one or more objects, such

as cars, are located inside a spatial area. All the

queries have the same object but different areas. A

plan was illustrated as a decision tree cost model that

reflects the expected number of comparisons,

calculated as the product of the number of

comparisons and the probability of finding an object

in the region (area). Sharing the query operator is the

second type of sharing. The queries continuously

inform the user about a group of objects that are

located in different areas. All the queries share the

same table, but they search various regions. In order

to optimize the query execution, a new Shared

Global Plan was implemented and evaluated.

Sharing the objects of the same interest is the third

type of sharing. The queries continuously inform the

user about the number of objects that are located in

an area. Another new Shared Global Plan was

implemented and evaluated and the push-down

approach was applied by pushing the Selection

operator pushed below the Join operator. Three

different joining policies introduced in (Xiong,

2004) namely Clock-triggered Join Policy (CJP),

Incremental Join Policy (IJP), and Hot Join Policy

(HJP), were used in (Mokbel, 2004a). In the CJP, the

spatial Join is reevaluated every T seconds. The IJP

does not execute the spatial Join for the objects and

queries that did not change their location since the

last T interval of time. The HJP evaluates hot

objects, if their movement affects the result, at each

evaluation time. The results of the evaluation of the

policies showed that for “moving queries on

stationary objects” and “stationary queries on

moving objects”, the CJP policy had a constant

number of I/O regardless of the percentage of

moving queries. The IJP policy had a number of I/O

that was significantly smaller than CJP. The same

applied to the HJP policy when compared to IJP.

The IJP policy had a much lower CPU cost than CJP

and the HJP policy had a lower CPU cost than IJP.

2.4 Comparison of Strategies

The above described query optimization strategies

have demonstrated a considerable improvement in

the field of query optimization. However, the

Sharing Paradigm approaches the multiple single

predicate queries but not multiple queries with

multiple predicates. It discusses sharing of the

underlying space and object of interest for the type

of queries that deal with finding if an object is

located inside an area but does not consider the

queries that deal with proximity analysis and it alters

the cache memory size. The push-down strategy

approaches re-ordering the operators of a plan based

on the Selectivity and executes the updates in

queries instead of re-executing the whole queries

several times. However it alters the database schema

which is not feasible in real-life applications. In

order to address the limitations of the existing

strategies a new query optimization framework is

needed for multiple dynamic complex queries which

does not change the cache memory size or the

database schema. The proposed query optimization

framework employs the new Query Melting

paradigm (QM) that includes common sub-

expression elimination, sharing objects of interest,

spatial areas, time intervals, underlying space, and

intermediate results. QM is implemented by using

the new sliding ruler strategy.

3 QUERY MELTING

3.1 Commonality in GIS

A thorough examination of different GIS and

Location Based Services (LBS) applications, where

users ask questions related to their position whether

they are moving (dynamic) or not (static), shows that

they have in common some functionalities,

operations, and objects in execution plans. A

predicate is the operator used by the user while

formulating a query such as Find Within a buffer and

Find the k Nearest Facilities. There are two types of

operators namely the static and the dynamic. Each

operator is decomposed as per its own query

evaluation plan which is made up of a list of

functions / operations that are to be executed in the

ICEIS 2009 - International Conference on Enterprise Information Systems

Figure 1: The Global Evaluation Plan of Operator1.

Figure 2: The Global Evaluation Plan of Operator2.

Figure 3: The Global Evaluation Plan of Both.

proper sequence. When multiple dynamic complex

queries are formulated with multiple operators the

global evaluation plan of each operator might

include functions that can be reused by other global

evaluation plans. Figures, 1 and 2 show the two

plans of the two different operators Operator1 and

Operator2. The dots marked with arrows in Figure 2

are the same as in Operator1 plan so they can be

melted. The other dots of Operator2 are appended to

those of Operator1’s because they can not be melted

as shown in Figure 3. In this global execution plan

there are 3 categories of operators: (a) Those that are

executed once at the beginning only, (b) those that

are executed at every new time instance, and (c)

those that are executed for each query.

3.2 Query Melting Paradigm

The Query Melting Paradigm (QMP) is a generic

optimized framework for GIS. It aims at building the

optimal global evaluation plan for GIS spatio-

temporal dynamic complex queries. Its main

objective is to produce the most cost effective query

processing in terms of execution time and memory

storage, thus, minimizing the queries execution time

cost. It is based on a combination of the Sharing

paradigm, Query Optimization, and Push-down

strategy. The Query Optimization that was applied

by (Kang, 1994; Andrade, 2001; Elmongui, 2006;

Mokbel, 2005a) is here extended to include “multi-

user spatio-temporal multi-predicate dynamic

complex queries”.

4 QUERY MELTING

PROCESSOR

Query Melting is implemented using the Query

Melting Processor QMP which is a middleware

software system located on the server. Its major

function is to input user queries, optimize them

based on the Query Melting Paradigm, generate

global execution plan and execute it, produce the

resulting maps, and send the output to the user.

4.1 QMP Components

The Query Melting Processor consists of a number

of components as shown in Figure 4. The

architecture shows the three major steps that

dynamic complex queries have to pass through in

order to produce a query global evaluation plan. The

Preprocessor takes as an input the dynamic complex

query, parses it into multiple simple queries, groups

them by category, and sorts them. Each simple query

is decomposed according to its operator template.

Time

…

OPERATOR 1 and OPERATOR 2

PERAT

R 1

1…

Time

… Time

OPERATOR 2

Time

… Time

QUERY MELTING - A New Paradigm for GIS Multiple Query Optimization

Two-dimensional arrays are used to store the

decompositions. The idea of using a Sliding ruler is

applied here to melt the repetitions that exist in

multiple plans. First, the Query Melting Ruler 1

melts the templates functions that are shared among

multiple simple queries in the aim to implement

common sub-expression elimination, sharing sub-

plans, and sharing the underlying space (map). The

Query Melting Ruler 1 works on a two-dimensional

plane, and its output is the Initial Evaluation Plan for

Time 0 of the whole dynamic complex query.

Second, the Query Melting Ruler 2 is responsible for

implementing sharing the space and areas, sharing

the time intervals, and sharing the object of interest.

Sharing the space and areas occurs when multiple

queries share the same spatial area or when an area

is included in another. The Query Melting Ruler 2

draws the buffer of an area only once and uses it

multiple times to clip or find the objects of other

multiple simple queries. It allows sharing an interval

of time if it is included or equal to the interval of

other multiple queries. Finally, sharing the object of

interest allows sharing the object of interest between

multiple simple queries by reading the object table

once and using it for many queries. The Query

Melting Ruler 2 works on the same two-dimensional

plane as the Query Melting Ruler 1, and its output is

the Final Global Evaluation Plan for Times 1…n.

Figure 4: The Components of the QMP.

4.2 System Mechanism

Before describing the mechanisms of the execution

plans, in the next few sections, we are describing

some abbreviation for the representation of the icons

of the IVQL which represent basically operators,

values, and objects. Also we are showing some

examples of simple as well as complex visual

queries. The smiley icons that represent operators

are listed in Table 1 along with their abbreviation

and explanation. The first six operators are used for

static queries, i.e. they ask about objects at the

current Time 0 only. The next six operators are used

for dynamic queries, i.e. they launch a continuous

query that lives n minutes. Every m minutes, they

update the user with the new results of the queries

based on his new location that is received through

the GPS and satellite systems. Table 2 shows some

of the icons that represent objects and facilities such

as restaurant, hospital, and gymnasium.

Table 1: The Icons Used as Operators to Formulate

Dynamic Complex Queries.

Icon

Abbre

viation

Explanation

Find the facilities that are

within a certain buffer

SN Find the k nearest facilities

SP Find the shortest path

SNP

Find the k nearest facilities

and their shortest paths

STL

Find the time left to reach

the k nearest facilities

SDL

Find the distance left to

reach the k nearest facilities

Continuously find the

facilities that are within a

certain buffer

Continuously find the k

nearest facilities

Continuously find the

shortest path on my way

DNP

Continuously find the k

nearest facilities and paths

DTL

For the next n minutes,

update me every m minutes

with the time left to reach

the k nearest facilities

DDL

For the next n minutes,

update me every m minutes

with the distance left to

reach the k nearest facilities

Dynamic

Complex Query

Operators

Templates and

PreProcessor

• Parsing, Grouping & Sorting Queries

• TCOP: Redirect Queries to GEP of Similar Ones

• Decom

osition of

ueries and Fillin

Arra

Melting Ruler 1

• Melting Two-Dim Array Operator Templates

• Building Initial Evaluation Plan for Time 0

Initial Evaluation

Plan for Time 0

Melting Ruler 2

• Melting Queries, Values, Objects, Methods

• Appending Values and Objects to Methods

• Building Final Evaluation Plan for Times

Final Global Evaluation

Plan for Times 1…n

ICEIS 2009 - International Conference on Enterprise Information Systems

Table 2: The Icons Used as Objects.

Icon Abbreviation Explanation

R Restaurant

M Motel

H Hospital

U Tube / Metro

G Gymnasium

The AND operator is used to combine

multiple simple queries into a complex one. Some

examples of simple queries are:

[SN 1 M]: find nearest 1 motel

[SNP 5 M]: find 5 nearest motels with their paths

[SW 500 R]: find all restaurants that are within 500

meters from my location

[SW 1000 R]: find all restaurants that are within

1000 meters from my location

[DTL 60 10G]: continuously, find the Time Left to

reach the nearest gymnasium for the next 60

minutes. Keep on supplying me with an updated

result every 10 minutes. The result includes the

shortest path.

An example of a dynamic complex query is

shown in Table 3 summarizing a number of queries:

Find the nearest theatre and its shortest path. While I

am on my way, keep on supplying me with the 3

nearest motels and their shortest paths that are within

200 meters until I reach the theatre or 60 minutes

overlap. Update my map every 5 minutes.

Table 3: A Dynamic Complex Query with 3 Predicates.

[DTL 60 5 T] [ DNP 3 M ] [DW 200 M]

Each operator is decomposed into a set of steps

according to the operator’s template. The templates

of some of the static operators are shown in Table 4

where each static operator is performed once only

for Time 0. The steps of each operator are executed

in the order in which they appear in the template

following the top-down direction. The MakeLayer

function creates a new closest facility layer that is

used to find the nearest objects. The ReadXY

function reads the XY location of the user. The

AddXYIncident function adds the user’s location as

an incident to the layer. The Path is set to YES so as

to produce the shortest path and set to No if no path

is required. The NFacilities is the number of

facilities needed to look for. The Impedance is set to

Meters, Pedestrian-Time, or Drive-Time. The

AddFacilities function specifies the object database

table such as restaurants, motels, etc. The Solve

function finds the nearest facilities based on the

previous parameters. The AddToMap and SendMap

functions add the results to the map and send it to

the user. The DrawBuffer draws a circle around the

XY location of the user and the ClipBufFac finds the

facilities that are located in this buffer. The

templates of the dynamic operators are shown in

Table 5. Each of the dynamic operators has two

templates. The first is applied for Time 0 when the

query is launched and the second is applied for each

of the consequent time instances Times 1…n. The

NewWatcher waits for new XY Locations to arrive

then launches triggers to read the data and continue

execution accordingly. The NewTimer launches the

query life time which triggers a timer that lasts as

long as required by the user. The KillQueryIfX is

used to terminate the query if it is expired, the user

reaches his destination, the user issues a cancellation

order, or he gets disconnected. The rest of the

functions operate the same as the static operators.

The Query Melting process is performed by the

query melting rulers. If an element of the template

ends with a star *, the corresponding elements in the

arrays are filled with objects, such as AddFacilities*

leads to AddRestaurant, AddHospital, etc., as shown

in Step 9 of Table 6. If an element ends with two

starts **, the corresponding elements are

increasingly numbered starting with 1, such as

Solve** leads to Solve1, Solve2, and so on. If an

element ends with ***, the corresponding elements

are increased by 1 for every new time instance, such

as ReadXY*** leads ReadXY1 for Time 0 and

ReadXY2 for Times 1…n. Using ** at the end of an

element means that this particular element is

repeated and executed for each simple query

whether static or dynamic, whether for Time 0 or

Times 1…n. The Melting Ruler follows the top down

direction. It eliminates all the repetitions of an

element each row at a time. The result of the Melting

Ruler is a list of functions that are to be executed,

the Global Evaluation Plan.

QUERY MELTING - A New Paradigm for GIS Multiple Query Optimization

Table 4: The Templates of Static Operators.

Table 5: The Templates of Dynamic Operators.

Table 6: Multiple Queries with One Static Operator during Query Melting.

Figure 5: The User Interfa

onds in the box called Cost of

MP. During execution, the queries are displayed in

their relative boxes.

ce before Melting.

4.3 Implementation

The QMP is implemented using the Microsoft

Visual Basic.NET. The underlying experimental

environment consists of a 1.8 GHz Intel Centrino

Duo PC with 1 GB RAM running Microsoft

Windows XP Home Edition. The User Interface of

QMP is divided into sections as shown in Figures 5

and 6. Each Operator such as “Time Left” and

“Distance Left” has its own section that is made up

of two boxes, the first for the dynamic queries and

the second for static. Each box contains six columns,

the first for the template and the next for queries.

Each row corresponds to a function. The two long

list boxes are the global evaluation plans. After

processing a complex query, the execution time is

isplayed in nanosec

ICEIS 2009 - International Conference on Enterprise Information Systems

Figure 6: The User Interface after Melting.

5 EVALUATION

For the evaluation of the Query Melting Processor

(QMP) the ‘time’ quality characteristic is quantified

and cost is estimated using the Big-Oh notation

which reflects how the cost of computation grows as

a function the input size by considering the

computation cost to be the running time in terms of

the size, thus indicating an approximation to the

number of steps taken by a process. Therefore, the

computational cost of Melting the Templates is O(Ns

× (Nt-1)) where Ns is the number of steps, Nt is the

number of templates and the steps correspond to the

row index numbers of each template. Since the

computational cost of processing Ns steps is O(Ns)

where for each step the rest Nt-1 templates that are

numbered from two to the last one are processed in

reverse order and since the computational cost of

processing Nt-1 templates is O(Nt-1), thus the

computational cost of both these dimensions is O(Ns

× (Nt-1)). The computational cost of Melting the

Values is O(Nq-1) where all the queries are

processed starting from the first one until the last

one, Melting the Objects is O(Nq-1) same as Values,

Filling and Appending the Methods with Values and

Objects is O(Ns × Nq) where all the queries are

processed for all the steps, and Generating the

Global Execution Plans is O(Nq × Ns) where all the

multi-dimensional array elements are traversed.

Hence, it can be concluded that the total time cost of

the Query Melting Processor is the sum of the time

costs of all its processes and is calculated as follows:

O(QMP) = O(Ns × (Nt-1))+ O(Nt × Nqti × )+

O(Nq-1)+ O(Nq-1)+ O(Ns × Nq) + O(Ns × Nq)

∑

−

The second quality characteristic that is

quantified in order to evaluate the efficiency of the

Query Melting Processor is the memory space. This

is done by evaluating the array that stores the

templates of the operators and the functions of the

queries. Since the name of each function does not

exceed 20 characters and each character occupies 2

Bytes in RAM (some systems 1 Byte) the name

occupies maximum 40 Bytes. The total space

occupied in the RAM is 40B/Function/Operator. In

other words, it is equal to the product of the number

of operators, the number of functions of each

operator, and 40 Bytes, which is quantified as Space

Cost = 40 Bytes × Number of Functions × Number

of Operators. Hence, the space cost is considered

minimal.

6 CONCLUSIONS

In this paper a query melting approach to processing

dynamic complex queries was introduced. The

approach aims at designing and implementing QMP,

a query melting processor that is a generic

optimization framework for GIS. A critical

evaluation of the existing query optimization

strategies proposed in the literature has been carried

out in order to build the presented approach upon the

current strategies. An investigation has been

conducted to examine the commonalities that exist

between the execution plans of different operators

by studying the static operators alone, the dynamic

ones alone, and a combination of both. The proposed

QMP has been introduced, its components

explained, and its mechanism described. The

mechanism is based on the new query melting ruler

strategy that is responsible for implementing the

query melting paradigm through sharing spatial

areas, time intervals, objects, functions, underlying

space (map), user locations, and intermediate results.

The proposed QMP implementation and user

interface have been described using a tourist GIS

system for proximity analysis. Finally, a theoretical

evaluation has been carried out in order to quantify

the time cost effectiveness of the approach.

QUERY MELTING - A New Paradigm for GIS Multiple Query Optimization

REFERENCES

Afework A., Beynon M. D., Bustamante F., Demarzo A.,

Ferreira R., Miller R., Silberman M., Saltz J., Sussman

A., and Tsang H., 1998. Digital Dynamic

Telepathology – The Virtual Microscope, In AMIA 98,

American Medical Informatics Association, 1998.

Andrade H., Kurc T., Sussman A., and Saltz J., 2001.

Efficient Execution of Multiple Query Workloads in

Data Analysis Applications. Proceedings of SC2001,

Denver, USA.

Andrade H., Kurc T., Sussman A., Borovikov E., and Saltz

J., 2002a. On Cache Replacement Policies for

Servicing Mixed Data Intensive Query Workload.

Proceedings of the 2

Workshop on Caching,

Coherence, and Consistency, 2002.

Andrade H., Kurc T., Sussman A., and Saltz J., 2002b,

Scheduling Multiple Data Visualization Query

Workloads on a Shared Memory Machine,

Proceedings of the 2002 International Parallel and

Distributed Processing, 2002.

Andrade H., Kurc T., Sussman A., and Saltz J., 2002c,

Multiple Query Optimization for Data Analysis

Applications on Clusters of SMPs, In Proceedings of

the 2

International Symposium on Cluster

Computing and the Grid, 2002.

Andrade H., Kurc T., Sussman A., and Saltz J., 2002d,

Active Proxy-G: Optimizing the Query Execution

Process in the Grid, Proceedings of the 2002

ACM/IEEE Supercomputing Conference, 2002.

Andrade H., Kurc T., Sussman A., and Saltz J., 2002e,

Processing Large-Scale Multi-dimensional Data in

Parallel and Distributed Environments, Parallel

Computing, 28(5), 827-859, 2002.

Andrade H., Aryangat S., Kurc T., Slatz J., and Sussman

A., 2003, Efficient Execution of Multi-Query Data

Analysis Batches Using Compiler Optimization

Strategies, Proceedings of the 16

International

Workshop on Compilers for Computing, LCPC, 2003.

Andrade H., Kurc T., Sussman A., and Beomseok N.,

2006, Data Management and Query – Multiple Range

Query Optimization with Distributed Cache Indexing,

SIGMOD Conference, 2006.

Andrienko G., Andrienko N., and Wrobel S., 2007, Visual

Analytics tools for Analysis of Movement Data, In

Proceedings of ACM SIGKDD Explorations

Newsletter, Vol. (9) 2, ACM.

Beeharee A., and Steed A., 2006, A Natural Wayfinding

Exploiting Photos in Pedestrian Navigation Systems,

In Proceedings on Human-computer Interaction with

Mobile Devices and Services, MobileHCI’06.

Beeharee A., and Steed A., 2007, Exploiting Real World

Knowledge in Ubiquitous Applications, Personal and

Ubiquitous Computing.

Elmongui H., Mokbel M., and Aref W., 2005, Spatio-

temporal Histograms, Proceedings of SSTD, 2005.

Elmongui H., Ouzzani M., and Aref W., 2006, Challenges

in Spatio-temporal Stream Query Optimization.

Proceedings of MobiDE 2006, Chicago.

Elsidani Elariss H., Khaddaj S., and Haraty R., 2006a,

Towards a New Visual Query Language for GIS.

IASTED Databases and Applications 2006, 195-202,

Austria 2006.

Elsidani Elariss H., Khaddaj S., and Haraty R., 2006b, An

Evaluation of a Visual Query Language for

Information Systems. ICEIS (5) 2006, 51-58, Paphos.

Gianotti F., Nanni M., Pinelli F., and Pedreschi D., 2007,

Trajectory Pattern Mining, In Proceedings of the 13

ACM SIGKDD International Conference on

Knowledge Discovery and Data Mining, KDD’07.

Guting H., De Almeida T., and Ding Z., 2006, Modeling

and Querying Moving Objects in Networks, VLDB

Journal – The International Journal on Very Large

Databases, Vol. (15) 2.

Kang M., Dietz H., and Bhargava B., 1994, Multiple-

Query Optimization at Algorithm-Level. Proceedings

of SSDI 1994, Data Engineering 14, 57-75.

Kim S., Diverdi S., Chang J., Kang T., Iltis R., and

Hollerer T., 2007, Implicit 3D Modeling and Tracking

for Anywhere Augmentation, VRST 2007, Newport

Beach, California, November 5-7, 2007.

Ladd A., Bekris K., Rudys A., Kavraki L., and Wallach

D., 2005, Robotics-Based Location Sensing Using

Wireless Ethernet, Proceedings of the 8

ACM

International Conference on Mobile Computing and

Networking, MOBICOM, Sep. 2002, Atlanta, GA.

Mokbel M.F., Aref W.G., Hambrush S.E., and Prabhakar

S., 2003, Towards Scalable Location-Aware Services:

Requirements and Reseach Issues. In Proceedings of

the ACM Symposium on Advances in Geographical

Information Systems, ACM GIS.

Mokbel M.F., Xiong X., Aref W.G., Hambrush S.E., and

Prabhakar S., Hammad M., , 2004a, PLACE: A Query

Processor for Handling Real-Time Spatio-temporal

Data Streams. In Proceedings of the VLDB.

Mokbel M.F., Xiong X., and Aref W.G., 2004b, SINA:

Scalable Incremental Processing of Continuous

Queries In Spatio-temporal Databases. In Proceedings

of the SIGMOD.

Mokbel M.F., 2004c, Continuous Query Processing in

Spatio-temporal Databases. In Proceedings of the

ICDE/EDBT PhD Workshop.

Mokbel M.F., Xiong X., Hammad M., and Aref W.G.,

2005a, Continuous Query Processing of Spatio-

temporal Data Streams in PLACE. GeoInformatica,

9(4), 343-365, 2005.

Mokbel M.F., and Aref W.G., 2005b, GPAC: Generic and

Progressive Processing of Mobile Queries over Mobile

Data. In Proceedings of the MDM, Aya Napa, Cyprus.

Repenning A., and Ioannidou A., 2006, Mobility Agents:

Guiding and Tracking Public Transportation Users,

Proceedings of the Working Conference on Advanced

Visual Interfaces, AVI’06.

Xiong X., Mokbel M.F., Aref W.G., and Prabhakar S.,

2004, Scalable Spatio-temporal Continuous Query

Processing for Location Aware Services. In

Proceedings of the International Conference on

Scientific and Statistical Database Management.

ICEIS 2009 - International Conference on Enterprise Information Systems