From Document Warehouse to Column-Oriented NoSQL

Document Warehouse

Ines Ben Messaoud

, Refka Ben Ali

and Jamel Feki

Laboratory Mir@cl, University of Sfax, Sfax, Tunisia

Institute of Management of Gabès, University of Gabès, Gabès, Tunisia

University of Jeddah, Jeddah, Saudi Arabia

Keywords: NoSQL Document Warehouse, Galaxy Model, Column-oriented Model, Hierarchical Transformation.

Abstract: NoSQL (Not only SQL) gathers recent solutions that differ from the SQL model by a different logic of data

representation. It is characterized by its performance and its ability to handle a large amount of data. Due to

the absence of a clear approach to implement a Document Warehouse (DocW) under NoSQL model, we

propose, in this paper, a set of rules to transform the multidimensional galaxy model of the DocW into the

column-oriented NoSQL model. We suggest two types of transformations namely Simple and Hierarchical.

In order to validate our proposed transformation rules, we have used Cassandra as a Column-oriented NoSQL

system to implement a DocW for each type of transformation. We used Talend Data Integration tool to load

data in the implemented DocWs. We evaluate these two DocWs with two metrics WRL (Write Request

Latency) and RRL (Read Request Latency) using a medical collection.

1 INTRODUCTION

Document Warehouse provides an environment to

store unstructured data; it is a repository of

documents collected from external and internal data

sources. It is designed according to the

multidimensional model as either a star model (Tseng

et al., 2006) (Ben Mefteh et al., 2016) or a galaxy

model (Ben Messaoud et al., 2015) (Pujolle et al.,

2011). It organizes data for OLAP (On Line

Analytical Processing) analysis in order to enable

successful business intelligence (Tseng et al., 2006).

Its main concerns are: (i) the uniform storage of data,

and (ii) the restoration of fragments of texts

considered as relevant by the user (Tournier, 2007).

However, the volume of data to analyze reaches

critical sizes (Jacobs, 2009) and then becomes

difficult to manage with the classical available tools;

hence, the need of appropriate storage and processing

techniques is raised (Agrawal et al., 2011). In this

emerging context of “Big Data”, NoSQL

environment appears as an efficient alternative that

can provide scalability while maintaining flexibility

for an OLAP system. In fact, it allows considering

new approaches to implementing a warehouse,

particularly multidimensional implementation

(Chevalier et al., 2015a).

In order to benefit from the NoSQL technology,

we propose in this paper an approach to implementing

a NoSQL Document Warehouse (DocW). More

precisely, we are interested in the column-oriented

NoSQL model. It is the most appropriate model for

the warehouse and the multidimensional data

structure (Dehdouh et al., 2015). In addition, it allows

deploying the warehouse in the cloud and offers high

performance.

In this paper, we define a set of transformation

rules to convert the multidimensional model of the

document warehouse into a column-oriented NoSQL

model. We implement these rules using Cassandra

NoSQL database; we evaluate the results in terms of

the two metrics namely Write Request Latency and

Read Request Latency on a medical collection.

This paper is organized as follows: Section 2

discusses related works that address NoSQL Data

Warehouses. Section 3 presents the DocW specifics

and its relying multidimensional galaxy model. Then,

the column-oriented NoSQL model and our proposed

transformation rules to obtain a NoSQL DocW are

described in Section 4. Section 5 is dedicated to

experiment and evaluation. Finally, Section 6

concludes the paper and addresses future works.

Messaoud, I., Ali, R. and Feki, J.

From Document Warehouse to Column-Oriented NoSQL Document Warehouse.

DOI: 10.5220/0006423500850094

In Proceedings of the 12th International Conference on Software Technologies (ICSOFT 2017), pages 85-94

ISBN: 978-989-758-262-2

2 RELATED WORKS

A DocW is a dedicated technology to store

documents issued from internal and external data

sources to the organization. These documents are

organized for effective analyses in order to enable

distilled and fruitful business intelligence (Tseng et

al., 2006). A DocW can be modelled in different

ways, as a star model (McCabe et al., 2000) (Tseng et

al., 2006), or as a galaxy model (Pujolle et al., 2011)

(Ben Messaoud et al., 2015). Furthermore, with the

increasingly volumes of data, it is becoming

impractical or even impossible to load unstructured

data into a warehouse. To alleviate this problem, new

technologies are currently studied in research

laboratories at universities and by software providers.

Recently, NoSQL (Not only SQL) emerges as a

promising technology to deal with huge volumes in

conventional databases (Stonebraker, 2012) and in

Big Data applications. Furthermore, (Chandawni,

2016) advocate to use NoSQL in the warehousing

domain. In fact, to the best of our knowledge, few

literature works treat the data warehouse using

NoSQL. Furthermore, there is no literature works

those treat the Document Warehouse relying their

solutions on NoSQL.

In this section, we study pertinent works where

researchers have proposed approaches or rules to

implement Data Warehouse using NoSQL, precisely

column-oriented and document-oriented models.

Among the well-known works, we cite the paper

of (Li, 2010) where the author proposes an approach

to transform a relational database into a column-

oriented NoSQL database. The author presents a set

of rules to conduct this transformation. Yet, the

proposed approach does not consider the conceptual

level of the Data Warehouse (DW); it considers only

the logical level to transform the relational model into

a column-oriented model.

In (Dehdouh et al., 2014), the authors have

presented a benchmark for columnar NoSQL DW.

This benchmark allows generating synthetic data and

queries set in order to evaluate the performance of

systems and the impacts of different technical

choices. However, the authors do not give the

formalization for the modeling process. Later, this

work was extended in (Dehdouh et al., 2015) by

proposing three approaches to implement the DW

using a column-oriented NoSQL model. Their

approaches are called NLA (Normalized Logical

Approach), DLA (Denormalized Logical Approach)

and DLA-CF (Denormalized Logical Approach by

using Column Family). The first approach (i.e., NLA)

uses different tables to store facts and dimensions,

and uses a simple attribute for measures and

dimension attributes. The second approach (i.e.,

DLA) proposes to store the fact and dimensions in the

same table, and uses a simple attribute to map

measures and dimension attributes. The third

approach DLA-CF stores the fact and dimensions in

the same table, and uses the compound attribute to

represent measures and dimension attributes.

Nevertheless, the NLA approach is quite inefficient

when performing queries with joins. The DLA-CF

approach is more efficient than DLA approach only

when the query handles attributes belonging to the

same dimension.

In (Chevalier et al., 2015a) (Chevalier et al.,

2015b) the authors propose a set of rules to map the

star multidimensional model into the two NoSQL

model: column-oriented and document-oriented. To

present the star model into a column-oriented NoSQL

model, the authors in (Chevalier et al., 2015c) use the

concepts of the column-oriented NoSQL model.

Indeed, the star model is transformed into a single

table. Since the star model is composed of fact and

dimensions, the fact is transformed into a column-

family where each measure is presented as a column.

And each dimension of the star model is transformed

into a column-family where the dimension attributes

(parameters and weak attributes) are transformed into

a column of the column-family. In the purpose to

transform the star model into a NoSQL document-

oriented model, the authors of (Chevalier et al.,

2015d) define a set of rules in order to transform each

star model into a collection of documents. The fact is

translated into a composite attribute and each measure

is transformed into a simple attribute. As the fact is

surrounded by a set of dimensions, each dimension is

transformed into a composed attribute (a nested

document) and each parameter and weak attribute is

converted to a simple attribute. However, the

proposed rules do not allow a hierarchical

transformation from the star model into the two

NoSQL models: column-oriented and document-

oriented. In other words, the defined rules do not

highlight the hierarchy concept of the star model,

which is fundamental for the DrillDown and RollUp

OLAP operations.

Due to the absence of a clear approach that allows

the implementation of a NoSQL DW, (Yangui et al.,

2016) propose to implement a NoSQL DW. To do so,

they propose a set of four rules to transform a star

multidimensional model into either column-oriented

NoSQL and document-oriented NoSQL models. For

each model, they distinguish two types of

transformations: simple and

hierarchical.

ICSOFT 2017 - 12th International Conference on Software Technologies

The first transformation is the mapping to NoSQL

model without detailing the hierarchy concept; all

hierarchy parameters of a dimension are presented in

a single structure into the NoSQL model. It makes a

distinction between measures and dimensions;

dimensions and facts are stored separately on

different column-family/collection. To ensure the

links between dimensions and fact, the dimension

identifier is duplicated in the column-

family/collection representing the fact.

The second transformation (i.e., hierarchical)

aims to transform the star model of the data

warehouse into a NoSQL model while describing the

hierarchy concept. It uses the concepts super-

column/document to present dimensions and

hierarchies of the multidimensional model (A super-

column is composed of a set of columns and a

document is defined as a collection of attributes).

Table 1 compares the literature works according

to the following criteria. Each criterion Ci means that

the approach

C1: Transforms the multidimensional model into a

column-oriented NoSQL.

C2: Transforms the multidimensional model into a

document-oriented NoSQL.

C3: Transforms the star multidimensional model into

a NoSQL.

C4: Transforms the galaxy multidimensional model

into NoSQL.

C5: Proposes a set of rules to do the transformation

into NoSQL.

Table 1: Comparison of works for NoSQL DW.

Approach C1 C2 C3 C4 C5

(Li, 2010) - - - -



(Dehdouh et al., 2015)



(Chevalier et al., 2015a)

  



(Yangui et al., 2016)

  



In this section, we have presented relevant works

related to NoSQL DW. We note that some of these

works transform the multidimensional model of the

DW into a column-oriented NoSQL model whereas

some others use the document-oriented NoSQL

model. Also, we stress that the proposed approaches

were interested in the star model; nevertheless, there

is no attempt to transform the galaxy model of

document warehouses into NoSQL. This represents a

lack we alleviate for the Document Warehouse

context using NoSQL.

The remaining of this paper presents our proposed

rules to transform a DocW into NoSQL.

3 XML DOCUMENT

WAREHOUSE

Documents contain interesting textual data; therefore,

they represent a source of elements useful for

decisional analyses. Frequently, these documents

have XML format and heterogeneous structures even

though they share the same domain. The DocW

collects and organizes documents to be used by

OLAP analyses dedicated for decisional purposes.

Due to their heterogeneous structures, the DocW user

(i.e., decision-maker) is constrained to write several

queries and then manage/merge their results to build

the final response. This requires competence and skill

to manage this hard task. To alleviate this problem,

(Feki et al., 2013), (Pujolle et al., 2011) and

(Tournier, 2007) propose approaches to build the

schema of the XML DocW. They use the galaxy

multidimensional model to describe the schema of the

DocW.

The galaxy model can be seen as a network of

entities (i.e., dimensions) connected by nodes. Each

node links compatible entities. Compatible entities

are entities that could be used together in OLAP

queries. In a galaxy, each entity can play a double role

either an analysis subject (i.e., fact) or an analysis axis

(i.e., dimension). The basic concepts of the galaxy

model are nodes, dimensions and hierarchies. We

formalize these concepts as follows:

Galaxy Model. A Galaxy model is the

generalization of the constellation multidimensional

model. This model is a grouping of dimensions

connected by nodes. The Fact concept is hidden.

Formally, a galaxy model N is defined by (GM

, GM

) where:

 GM

: is the name of the galaxy model.

 GM

= {D

… D

}: is a non-empty set of

dimensions.

 GM

= {N

… N

}: is a non-empty set of nodes.

Dimension. A dimension models an analysis axis.

It has a set of attributes called parameters organized

in hierarchies. Formally, a dimension D is a triplet

, D

) where:

 D

: is the name of the dimension D.

 D

: is a set of at least one strong attribute

(called parameter) and may have additional

weak attributes those label the parameters.

 D

: is a non-empty set of hierarchies.

Hierarchy. A hierarchy organizes parameters in

several levels. It is defined by (H

, H

, P

) where:

 H

: is the name of the hierarchy.

 H

= (P

,…, P

): is a set of parameters of H

 P

: is a function that associates each

parameter to its weak attributes.

From Document Warehouse to Column-Oriented NoSQL Document Warehouse

Figure 1: An example of a galaxy model (Tournier, 2007).

Node. A node connects compatible dimensions;

i.e., dimension those could be used semantically

together within a same multidimensional query.

Formally, a node is defined by (N

, N

) where:

 N

: is the name of the node.

 N

: is a function that links each dimension to

the set of its nodes. Naturally, a dimension can

be connected to more than one node.

Figure 1 depicts an example of a galaxy

multidimensional model, which is composed of six

dimensions called Conferences, Articles, Authors,

Dates, Rapports and Institutes connected via two

nodes. The first node links the four dimensions

Conferences, Dates, Articles, and Authors. It

describes articles published in a conference and

written by authors on a given date. While, the second

node connects the four dimensions: Dates, Rapports,

Authors, and Institutes. It translates reports of projects

led by institutes and supervised by scientific

personnel (authors) on a specific date. This model

allows decision-makers, for instance, to analyze

research papers authored by authors and published in

conferences.

4 TRANSFORMATION OF A

GALAXY MODEL INTO A

NOSQL COLUMN-ORIENTED

MODEL

In this paper, we assume that a DocW is

multidimensional modelled as a galaxy. Because the

volume of documents is increasing permanently,

looking for new technological solutions that support

huge volume of documents is more and more felt both

by warehouse users and designers communities. The

objective of designers is to better serve decision-

makers whose needs are beyond the classical needs of

DW users. It is mainly to improving performance of

the DocW in terms of response time and queries. In

parallel, NoSQL is becoming rapidly an efficient

alternate to RDBMSs (Relational Data Base

Management Systems) for storing, managing and

querying big volumes of data (Chandawni, 2016).

This motivated us to explore the usage of this new

technology in document warehousing. To do so, we

suggest, in this work, dedicated rules for transforming

a galaxy multidimensional model of XML documents

warehouse into NoSQL. More accurately, we have

elected the column-oriented NoSQL model for this

work. In fact, this model is the richest NoSQL model

and it is characterized by it performance (Lemberger

et al., 2015).

To transform a galaxy model into a column-

oriented NoSQL model, we define two types of

transformations: simple and hierarchical. The first

type of transformation transforms the galaxy model

into a column-oriented NoSQL model without

detailing the hierarchy concept. Whereas, the

hierarchical transformation describes hierarchies

when converting the galaxy model into a column-

oriented NoSQL model.

In the remaining of this paper, we define the

fundamental concepts of the NoSQL column-oriented

model and a set of appropriate transformation rules

for both simple and hierarchical transformations.

ICSOFT 2017 - 12th International Conference on Software Technologies

Figure 2: Basic concepts of the column-oriented NoSQL model.

4.1 Column-Oriented NoSQL Model

The column-oriented NoSQL model provides a

flexible schema characterized with a large number of

columns that can differ between each row without

generating null values. For this reason, it can be seen

as a set of tables defined row by row. In this section,

we define the five basic concepts of this model

namely Table, Row, Column-Family, Super-column

and Column.

Table. A table is a set of lines and a line is

composed of a key and a set of column-families. It is

defined by (T

, T

) where:

 T

: is the name of the table.

 T

= {R

, ..., R

}: is a set of rows.

Row. A row represents a record in the table of the

database. Formally, it is a couple (R

, R

) where:

 R

: is the key of the row.

 R

= {CF

, ..., CF

}: is set of column-families.

Column-Family. A column-family consists

of a set of columns or super-columns. It is

defined by (CF

, CF

) where:

 CF

: is the name of the column-family.

 CF

= {C

,…,C

}: is a set of columns or super-

columns.

Super-Column. A super-column allows the

grouping of columns with data semantically linked

(Lemberger et al., 2015). It is defined by the couple

(SC

, SC

) where:

 SC

: is the name of the super-column.

 SC

= {C

,…,C

}: is a set of columns.

Column. A column is characterised by a name

and an atomic value. We note that each atomic value

can be historised with a time label: timestamp. In fact,

this principle is useful for historical management

(Wrembel, 2009). Formally, a column is a couple (C

) where:

 C

: is the name of the column.

 C

: is the column value.

Figure 2 shows the basic concepts of the column-

oriented NoSQL model.

4.2 Simple Transformation

The simple transformation aims to convert the

concepts of the galaxy model into a NoSQL model

but without detailing the hierarchy concept;

transforming parameters of a dimension uses a unique

structure of the NoSQL model. To achieve this goal,

we define a set of four rules.

Rule 1: Each galaxy model G transforms into a

table T in the NoSQL column-oriented database.

Obviously, the galaxy is the most generic concept

of a multidimensional galaxy model; consequently, it

can represent only a table that presents the global

concept in the column-oriented NoSQL model.

Rule 2: Each node N belonging to a galaxy G and

connected to a set of compatible dimensions {D

, ...,

} transforms into a column-family CF.

Remember that in a NoSQL column-oriented

model, the column-family is a concept that contains a

set of columns. In addition to this, in the galaxy

model, a node is connected to a set of dimensions.

Consequently, the node merits to be transformed into

a column-family and each dimension connected to the

node is presented by a column.

Rule 3: Each dimension D

of the galaxy is

transformed into a column-family CF

where each

attribute (parameters and weak attributes) of D

transformed into a column C

of CF

∈ CF

From Document Warehouse to Column-Oriented NoSQL Document Warehouse

Figure 3: Example of a row in a table (Simple transformation).

In a galaxy, a dimension is characterized by a set

of attributes. In the NoSQL column-oriented model,

the column-family contains columns. Thereafter,

each dimension of the galaxy is transformed into a

column-family and each dimension attribute is

transformed into a column.

Rule 4: The nodes and dimensions instances are

transformed into a row of the table of the column-

oriented NoSQL model.

Recall that a row in a table of the column-oriented

NoSQL model is composed of a key, a set of column

families denoted CF. Consequently, the instances of

node, and the associated instances of dimensions

transform into a row of the table of the column-

oriented NoSQL model.

Figure 3 depicts an example of a row in a table of

the NoSQL column-oriented database. This row is

obtained by applying the simple transformation rules

on a galaxy composed of four dimensions: D-Date,

D-Author, D-Article and D-References, and a single

node. We note that this transformation does not take

into account hierarchies dimensions. To do so, we

propose a hierarchical transformation in order to

describe parameters hierarchies in the column-

oriented NoSQL model.

4.3 Hierarchical Transformation

Unlike the simple transformation, the hierarchical

transformation explain the hierarchy concept when

transforming the multidimensional model into a

column-oriented NoSQL model. This transformation

exhibits parameters of each hierarchy into a separate

structure of the NoSQL model. For the hierarchical

transformation, we keep three rules from the simple

transformation (Rules 1, 2 and 4) and one specific

rule we have defined for this transformation.

Rule 5: Each dimension D

of the galaxy

transforms into a column-family CF

where each

hierarchy is transformed into a super-column. The

name of the super-column is the name of the

hierarchy, whereas parameters and weak attributes

become columns.

In a column-oriented NoSQL model, a column-

family is composed of a set of super-columns and a

super-column is composed of columns. In the other

hand, in a galaxy model, a dimension is composed of

a set of hierarchies, and a hierarchy is composed of a

set parameters. Thereafter, we can use the concepts

column-family, super-column and column to present

dimensions and hierarchies into the column-oriented

NoSQL model.

Figure 4 shows an example of a row in a table of

the NoSQL column-oriented database obtained by

applying the hierarchical transformation rules on the

galaxy at the top of the figure.

ICSOFT 2017 - 12th International Conference on Software Technologies

Figure 4: Example of a row in a table (Hierarchical transformation).

Figure 5: Galaxy model for the medical collection (Ben Messaoud et al., 2015).

5 EXPERIMENT AND

EVALUATION

Due to the absence of a benchmark for the galaxy

model, we conduct our experiment on the galaxy

model generated from a set of 1691 XML documents

issued from the medical collection Clef-2007 and

described by three DTDs (Ben Messaoud et al.,

2015). This galaxy model is composed of five

dimensions: D-Casimage-Case, D-Author, D-

Keywords, D-Reviewer and D-References. Figure 5

presents galaxy model for the medical collection. We

note that weak attributes are voluntary withdrawn to

simplify the galaxy model.

From Document Warehouse to Column-Oriented NoSQL Document Warehouse

Figure 6: Evaluation of the implemented NoSQL column-oriented XML DocWs.

In order to implement the galaxy XML Document

Warehouse as a column-oriented NoSQL model, we

have elected Cassandra as a NoSQL database

management system. In fact, it can manipulate a very

large amount of data like any NoSQL database

management system. It includes the concepts of

columns, super-column and column-families that we

have used in our proposed rules.

To define the schema of the NoSQL DocW with

Cassandra, we used the Cassandra Query Language

(CQL). By using this language, it is useful to create

table, column-family, etc.

In the context of our work, we distinguish two

NoSQL DocW: DocW-S and DocW-H implemented

by applying respectively the simple and hierarchical

transformation rules.

To load the data of the galaxy model into the

NoSQL column-oriented DocW, we opted for the use

of the Talend Data Integration tool. This tool allows

extracting data from large, heterogeneous data

sources and integrating them into the NoSQL

database. In the context of our work, data integration

is carried out in accordance with our transformation

rules.

In order to evaluate the two-implemented NoSQL

document warehouses, we use the two metrics: Write

Request Latency (WRL) and Read Request Latency

(RRL). The WRL metric assesses the speed of the

system during the data loading stage. The RRL metric

measures the response time of the system to answer

user requests.

Figure 6 illustrates the evaluation of the two

implemented NoSQL XML DocW.

By comparing the values obtained with each

metric on the two implementations (Simple and

Hierarchical) of the DocW, we conclude that the

DocW implemented using the hierarchical

transformation is better than the simple one. Indeed,

the loading time and the user request response time

for the DocW-H are better than the DocW-S; To load

data in DocW-S, we need approximately the double

of loading time of DocW-H and to answer to decision

maker queries with the DocW-S we need more than

the twice of the response time of DocW-H.

6 CONCLUSION AND FUTURE

WORKS

Documents represent an important source of

information for decisional analyses. So, they merit to

be integrated in the decision support system.

Likewise, documents should be warehoused. As a

result, the document warehouse concept has emerged.

Nevertheless, the increasing production of

documents, the sharing of information between users

and the diffusion of data via networks generate a very

large volumes of available and interesting data to be

analysed. This huge amount of data require

appropriate storage means (Favre et al., 2013).

The NoSQL environment provides a solution to

answer the limitations of the relational systems in

terms of scalability and handling a large volume of

ICSOFT 2017 - 12th International Conference on Software Technologies

data. To benefit from this new technology, we

proposed, in this paper, an approach to build a

NoSQL Document Warehouse. More accurately, we

transform the multidimensional galaxy model of the

DocW into the column-oriented NoSQL model. From

NoSQL models, we elected the column-oriented

model because its performance has been proven in the

literature works.

To build a NoSQL DocW, we distinguish two

transformation types: simple and hierarchical. The

first transformation converts the concepts of the

galaxy model into a NoSQL model without detailing

the hierarchy concept. For this transformation, we

define a set of four rules. While, the hierarchical

transformation explains the hierarchy concept when

transforming the multidimensional model into a

column-oriented NoSQL model. It retains three rules

from the simple transformation and defines one

specific rule for this transformation.

To substantiate these rules, we use the NoSQL

database management system Cassandra and

Cassandra Query Language (CQL) to apply the

simple and hierarchical transformation rules. We

obtain respectively DocW-S

and DocW-H. Moreover,

the evaluation of the obtained NoSQL DocW in terms

of the two metrics Write Request Latency and Read

Request Latency on a medical collection shows that

the DocW-H is better than the DocW-S.

As a future work, we will propose rules to

transform the multidimensional model of the

document warehouse into the document-oriented

NoSQL model and compare the performance of the

two NoSQL DocWs: column-oriented DocW and

document-oriented DocW. In addition, we expect

define a set of analytical operations dedicated to the

galaxy model of the NoSQL DocW.

REFERENCES

Agrawal, D., Das, S., El Abbadi, A., 2011. Big data and

cloud computing: current state and future opportunities.

In EDBT/ICDT’11, 14

International Conference on

Extending Database Technology, pp. 530–533.

Ben Mefteh S., Khrouf K., Feki J., Ben Kraiem M., Soule-

Dupuy C., 2016. A Semantic Approach for XML

Document Warehousing and OLAP Analysis. In

IJIDS’16, International Journal of Information &

Decision Sciences, vol.8, n°.3, pp. 254-283,

DOI: 10.1504/IJIDS.2016.078587.

Ben Messaoud, I., Feki, J., Zurfluh, G., 2015. A Semi-

automatic Approach to Build XML Document

Warehouse, in CCIS’15, Communications in Computer

and Information Science, Springer International

Publishing Switzerland 2015, A. Fred et al. (Eds.), pp.

347–363.

Chandawni, G., 2016. NOSQL DATA-WAREHOUSE. In

IJIRCCE’16, International Journal of Innovative

Research in Computer and Communication

Engineering, Vol. 4, Special Issue 4, pp. 96-104.

Chevalier, M., Elmalki, M., Kupliku, A., Teste, O.,

Tournier, R., 2015a. Entrepôts de données

multidimensionnelles NoSQL. In EDA’15, 11

ème

Journées francophones sur les Entrepôts de Données et

l’Analyse en ligne, p.161-176

Chevalier, M., El Malki, M., Kopliku, A., Teste, O.,

Tournier, R., 2015b. Implementing multidimensional

data warehouses into NoSQL. In ICEIS’15, 17

International Conference on Enterprise Information

Systems, pp.108-130.

Chevalier, M., El Malki, M., Kopliku, A., Teste, O.,

Tournier, R., 2015c. Implementation of

Multidimensional Databases in Column-Oriented

NoSQL Systems. In ADMIS’15, 19th East-European

Conference on Advances in Databases and Information

Systems, pp. 79-91.

Chevalier, M., El Malki, Teste, O., Tournier, R., 2015d.

Implementation of Multidimensional Databases with

Document-Oriented NoSQL. In DAWAK’15, 17th

International Conference on Big Data Analytics and

Knowledge Discovery, pp. 379-390.

Dehdouh, K., Bentayeb, F., Boussaid, O., 2014. Columnar

NoSQL Star Schema Benchmark. In MEDI’14, 4

Model and Data Engineering, pp. 281-288.

Dehdouh, K., Bentayeb, F., Boussaid, O., Kabachi, N.,

2015. Using the column oriented NoSQL model for

implementing big data warehouses. In PDPTA'15, 21

International Conference on Parallel and Distributed

Processing Techniques and Applications, pp.469-475.

Favre, C., Bentayeb, F., Boussaid, O., Darmont, J., Gavin,

G., Harbi, N., Kabachi, N., Loudcher, S., 2013. Les

entrepôts de données pour les nuls ou pas !. In EGC’13,

conference francophone sur l’extraction et la

gestion de connaissance, pp. 1-18.

Feki, J., Ben Messaoud, I., Zurfluh, G., 2013. Building an

XML Document Warehouse. In JDS, Journal of

Decision Systems, Vol. 22 No. 2. pp. 122-148, DOI:

10.1080/12460125.2013.780322.

Jacobs, A., 2009. The pathologies of big data. In

Communications of the ACM 52(8), pp. 36–44.

Lemberger, P., Batty, M., Morel, M., Rafaelli, JL, 2015.

Big Data et Machine Learning, Dunod. 1

edition,

ISBN: 978-2-10-072074-3.

Li, C., 2010. Transforming relational database into HBase:

A case study. In ICSESS’10, International Conference

on Software Engineering and Service Sciences, pp.

683–687.

McCabe, C., Lee, J., Chowdhury, A., Grossman, D.,

Frieder, O., 2000. On the design and evaluation of a

multi-dimensional approach to information retrieval. In

SIGIR’00, 23rd International Conference on Research

and Development in Information Retrieval, pp. 363-

365.

From Document Warehouse to Column-Oriented NoSQL Document Warehouse

Pujolle, G., Ravat, F., Teste, O., Tournier, R., Zurfluh, G.,

2011. Multidimensional database design from

document-centric XML documents. In DAWAK’11,

International Conference on Data Warehousing

and Knowledge Discovery, pp. 51–65.

Stonebraker, M., 2012. New opportunities for new sql.

Commun. In ACM 55(11), pp. 10–11.

Tournier, R., 2007. Analyse en ligne (OLAP) des

documents. Thèse de doctorat en Informatique,

Université Toulouse III, Paul Sabatier, Toulouse,

France.

Tseng, F. S. C., Chou A. Y. H., 2006. The concept of

document warehousing for multi-dimensional

modeling of textual-based business intelligence. In

Decision Support Systems, pp. 727-744.

Wrembel, R., 2009. A survey of managing the evolution of

data warehouses. In IJDWM’09, International Journal

of Data Warehousing and Mining, pp. 24–56.

Yangui, R., Nabli, A., Gargouri, F., 2016. Automatic

Transformation of Data Warehouse Schema to NoSQL

Data Base: Comparative Study. In KES’16, 20

International Conference on Knowledge-Based and

Intelligent Information & Engineering Systems, pp.

255-264.

ICSOFT 2017 - 12th International Conference on Software Technologies