documents de recherche working papers – n° 01021
Transcription
documents de recherche working papers – n° 01021
CENTRE DE RECHERCHE RESEARCH CENTER DOCUMENTS DE RECHERCHE WORKING PAPERS – N° 01021 – From UML to ROLAP Multidimensional Databases Using a Pivot Model Jacky AKOKA CEDRIC –CNAM & INT Isabelle COMYN-WATTIAU Université de Cergy & ESSEC Nicolas PRAT ESSEC July 2001 GROUPE ESSEC CERNTRE DE RECHERCHE / RESEARCH CENTER AVENUE BERNARD HIRSCH - BP 105 95021 CERGY-PONTOISE CEDEX FRANCE TÉL. : 33 (0) 1 34 43 30 91 FAX : 33 (0) 1 34 43 30 01 Mail : [email protected] GROUPE ESSEC, ÉTABLISSEMENTS PRIVÉS D'ENSEIGNEMENT SUPÉRIEUR, ASSOCIATION LOI 1901, ACCRÉDITÉ AACSB - THE INTERNATIONAL ASSOCIATION FOR MANAGEMENT EDUCATION, AFFILIÉ A LA CHAMBRE DE COMMERCE ET D'INDUSTRIE DE VERSAILLES VAL D'OISE - YVELINES. WEB : WWW.ESSEC.FR FROM UML TO ROLAP MULTIDIMENSIONAL DATABASES USING A PIVOT MODEL Jacky AKOKA1, Isabelle COMYN-WATTIAU2 and Nicolas PRAT3 Abstract Effective data warehouse design requires a conceptual modeling phase. This paper describes a method for data warehouse design. The method relies on the three modeling levels that have been defined for transactional systems design : conceptual, logical and physical. The conceptual phase creates an object-oriented schema which is represented using the UML notation. To this end, UML is enriched with concepts relevant to multidimensional systems. The logical phase consists in mapping the enriched UML schema into a multidimensional one. We propose a set of rules facilitating the design process. The rules are based on a multidimensional model unifying the main decisional concepts as provided by ROLAP tools. The physical phase allows the designer to map the multidimensional schema into a physical schema, depending on the target decisional tool. We illustrate our approach on a case study and compare it to the state-of-the-art. Keywords Data warehouse, multidimensional system, design method, UML, ROLAP. Résumé Pour être efficace, la conception d’entrepôts de données doit s’appuyer sur une phase de modélisation conceptuelle. Ce document de recherche décrit une méthode des conception d’entrepôts de données. La méthode s’appuie sur les trois niveaux de modélisation qui ont été définis pour la conception de systèmes transactionnels: les niveaux conceptuel, logique et physique. La phase conceptuelle crée un schéma orienté-objet représenté avec le formalisme UML. A cet effet, UML est enrichi de concepts propres aux systèmes multidimensionnels. La phase logique consiste à transformer le schéma UML enrichi en schéma multidimensionnel. Nous proposons un ensemble de règles qui facilitent le processus de conception. Les règles se fondent sur un modèle multidimensionnel qui unifie les principaux concepts du décisionnel, tels que les concepts mis en œuvre dans les outils ROLAP. La phase physique permet au concepteur de transformer le schéma multidimensionnel en un schéma physique, en fonction de l’outil décisionnel cible. Nous illustrons notre approche par une étude de cas et la situons par rapport à l’état de l’art. Mots clés Entrepôt de données, système multidimensionnel, méthode de conception, UML, ROLAP. 1 CEDRIC-CNAM & INT, 292 rue Saint-Martin, 75141 Paris cedex 03, France. Email: [email protected] 2 Université de Cergy & ESSEC, 2 rue Chauvin, 95302 Pontoise cedex, France. Email: [email protected] 3 ESSEC, Department of Information and Decision Systems, avenue Bernard Hirsch, BP 105, 95021 Cergy cedex, France. Email: [email protected] 1. Introduction The data warehousing market is growing rapidly with an estimated average growth rate of 30% over the last four years. It is very likely to approach 5 billion dollars by 2003, compared to 1 billion in 1996 [OLAP report, 2001]. Like the relational database market at its beginning, the OLAP market has no dominant players. However, in the near future, analysts expect this market to be consolidated. All these factors contribute to suggest a strong analogy between the data warehousing field and the relational database one in the early eighties. OLAP design process is described by tool vendors as much easier than classical database design. We claim that this process is at least as difficult as the design of transactional databases. Since data warehouses are developed to provide managers with data on which they will build queries depending on their constantly evolving needs, the design process is crucial. Each OLAP tool is based on a specific underlying logical model. Although the various logical models rely on very similar concepts, no real normalization was provided nor agreed on. We claim that a standardization of the multidimensional model will allow both the users to better understand the underlying concepts and the tool editors to adopt a unified view. Moreover, companies using several OLAP tools will profit from having a unique logical description of their decisional data. Unlike the database world which has taken some time to adopt the three levels of abstraction recommended by ANSI/X3/SPARC, the data warehouse actors should rapidly divide the modeling task according to the three conceptual, logical and physical levels. The conceptual level allows the data warehouse designer to build a high level abstraction of decisional data independently from implementation issues. The logical level maps this abstraction into a standard multidimensional representation. Finally, taking into account the target tool, the physical step aims at building a schema to be implemented on a specific platform. The main objective of this paper is to propose a four-step design method for data warehouse development, including rules for systematic mapping between each step. In order to address several target tools, we define a unified multidimensional model at the logical level and build our process on it. The paper is organized as follows. Section 2 describes the unified multidimensional model. Section 3 presents the four-step design method. A case study illustrating the process is presented in Section 4. Section 5 compares our approach with the state-ofthe-art. Finally, Section 6 concludes and describes further research. 2 2. A Unified Multidimensional Model In contrast with the relational model, there is no standard multidimensional database model. More precisely, there is no commonly accepted formal multidimensional data model. As a consequence, many multidimensional models have been proposed in the literature [Agrawal et al, 1997 ; Blaschka et al, 1998 ; Cabibbo & Torlone, 1998 ; Chaudhuri & Dayal, 1997 ; Golfarelli et al, 1998 ; Gyssens & Lakshmanan, 1997 ; Kimball, 1996 ; Li & Wang, 1996; Pedersen & Jensen, 1999 ; Vassiliadis & Sellis, 1999]. The concepts used vary depending on the authors and some concepts, e.g. the notion of “fact”, are employed with various meanings. Furthermore, there is no consensus concerning the level of the multidimensional model (physical, logical or conceptual). The star and snowflake models presented in [Kimball, 1996] have often been considered to be at the physical level, since the choice between stars and snowflakes is based on performance considerations (trade-off between query performance and optimization of disk space). More recent publications have placed the multidimensional model at the logical level [Vassiliadis, 1999] or even at the conceptual level [Golfarelli et al, 1998 ; Hüsemann et al, 2000]. Our strong belief is that the multidimensional model belongs to the logical level. Even though there is no consensus on this model, it clearly exists independently of physical implementations. However, the multidimensional model should not be situated at the conceptual level since the concepts of this model (e.g. the concept of dimension) are not as close to reality as concepts like the object (used in conceptual object models like UML [OMG, 2001] for example). There is indeed a strong parallel between the relational model and the multidimensional model - e.g. the definitions or attempts to define an associated query language and a normalization theory. This is the reason why we argue that both models should be considered as belonging to the same level, i.e., the logical level. In the multidimensional model, data are organized in (hyper)cubes. Although the detailed concepts of this model vary depending on the authors, we can describe multidimensional semantics using four concepts which appear recurrently, namely the notions of measure, dimension, hierarchy and attribute. Our model is composed of these four concepts, which are illustrated in Figure 1. 3 3 March 99 4 4 March 99 6 5 March 99 9 6 March 99 9 7 March 99 3 8 March 99 1 9 March 99 12 P1 6 8 11 5 9 9 Bordeaux Brest Lyon Nantes Paris P2 P3 P4 P5 P6 P7 PRODUCT CI TY RE GI ON DAY MONTH YEAR QUARTER Quantity sold product name unit price LEGEND: Measure DIMENSION Attribute Hierarchy CATEGORY Figure 1 : Multidimensional representation of data The key concept of the multidimensional model is the notion of measure. A measure is typically a quantitative data, a numeric value of interest for the analysis. Examples of measures are the quantity sold and the total dollar amount by product, date and city. A measure needs not to be of numeric type, as long as its values are totally ordered. For example, it can be an enumeration type. Thus, the satisfaction of customers with a product may be measured on a four-value scale (unsatisfied – mitigated – satisfied – enthusiastic). In a cube, the measures correspond to the content of the cells. The dimensions form the edges of the cube. Each measure is associated with one or several dimensions, which specify the context of the measure. In the example of Figure 1, the quantity sold is dimensioned by the dimensions product, day and city, in other words, by “quantity sold”, we mean the quantity sold for a particular product at a particular date in a particular city. Note that a dimension is represented by its identifier (e.g. the values of the dimension product are product codes). Sometimes, we need to represent an event linking several dimensions without having any measure associated with this event [Kimball, 1996]. For this purpose, we use a specific type of measure, called dummy measure. Consider for example the relationship “reservation” linking a borrower, a book and a reservation date, without any specific attributes characterizing the reservation. The dummy measure “reservation” will serve to indicate which books have been reserved by which borrowers and at which date. The dimensions are organized in hierarchies. A hierarchy is an aggregation path between dimensions. In Figure 1, “city->region” and “day->month->quarter->year” are examples of hierarchies. A hierarchy is oriented from the 4 lower to the upper abstraction levels (here, from city to region and from day to year). The arrow between two successive dimensions may be interpreted as a functional dependency. Hierarchies are of paramount importance in the multidimensional model since they are used to drill up and down measures. For example, the total quantity sold by product, quarter and region may be computed by aggregating the quantities sold along the day and city dimensions. Due to the ubiquity of time in data warehouses, time hierarchies are very frequent and even necessary in multidimensional models. Dimensions may be described by attributes. For example, the dimension product (i.e. product code) is described by the product name and its unit price. Attributes are not the object of multidimensional analysis, as opposed to measures. In other words, if a dimension is described by a feature that is a measure of interest, this feature should be defined as a one-dimensional measure associated with this dimension. To specify a multidimensional model, we use the following notation : • A dimension is specified with the key word “dimension”. • A measure is specified with the key word “measure” and followed by the dimensions which provide the context for the measure (i.e. the edges of the cube). Among these dimensions, the ones that do not functionally determine the measure are represented between parentheses. For example, consider the dollar amount paid for a car rental. This measure is dimensioned by the hour of rental, the registration number of the car rented and the code of the customer renting the car. However, since the same car cannot be rented at the same time by different customers, the hour and the car registration number are sufficient to determine the rental. Therefore, in the specification of the measure “dollar amount”, the dimension “customer code” shall be indicated between parentheses. Dummy measures are preceded by the key word “dummy”. We need to distinguish explicitly dummy measures from other measures since they will be implemented differently depending on the target decisional tool. • A hierarchy is specified with the key word “hierarchy”, followed by the aggregation path. • An attribute is specified with the key word “attribute”, followed by the dimension it characterizes. Following this notation, the multidimensional schema illustrated in Figure 1 is specified as follows : measure quantity sold [product code, day, city] dimension day dimension month dimension quarter hierarchy day->month->quarter->year dimension year hierarchy product code->category dimension product code hierarchy city->region dimension category dimension city attribute product name [product code] dimension region attribute unit price [product code] 5 A graphical representation is also proposed (Figure 2). Measures are stored in trapezoidal boxes and linked to dimensions in rectangle boxes. Hierarchical links between dimensions are represented by arrows. Finally, attributes are stored in dimension boxes. Category Product Code quantity sold City Region Product name Unit price Day Month Quarter Year Figure 2 : A graphical representation of a multidimensional schema Our unified multidimensional model is generic and can be easily mapped into the usual multidimensional models that can be found in the literature and/or in decisional tools. This model is used as a pivot model in our design method, as described in the next section. 3. The Design Method Starting from user requirements, our method is based on the three usual abstraction levels : conceptual, logical and physical (Figure 3). It is therefore decomposed into four phases : - In the conceptual phase, the designer represents the universe of discourse using the UML notation [OMG, 2001] along with the associated approach of development[Jacobson et al, 1999] (step 1) ; the UML schema is then enriched and transformed to take into account the specific features of multidimensional modeling (step 2). - In the logical phase, the enriched and transformed UML schema is mapped into our unified multidimensional model, using mapping rules. 6 - The physical phase allows the designer to convert the multidimensional schema into a physical schema, depending on the target decisional tool. In this paper, we focus on ROLAP tools (tools using the star model, the snowflake model, a combination of the star and snowflake models…). A specific set of mapping rules from the logical to the physical model is defined for each type of tool. - The data confrontation phase consists in mapping the physical schema data elements with the data sources. It leads to the definition of queries for extracting the data corresponding to each component in the physical schema. This is a very complex problem, going beyond the scope of this paper. However, it is important to mention this crucial phase in the data warehouse design process. U n iverse of discourse con c e p t u a l m odelin g U M Lsch e m a CONCEPTUAL DESIGN enrichm e n t/transformation Enriched/transform ed U M L sch e m a Logical m apping LOGICAL DESIGN U n ified m ultidim ensional schem a PHYSICAL DESIGN P h ysical m a p p ing MOLAP sch e m a ROLAP star sch e m a R O L A P snowflake sch e m a … Source confrontation DATA CONFRONTATION D a ta W a r ehouse M etadata Figure 3 : The four phases of the design method 3.1. Conceptual design OLAP systems are emerging as the dominant approach in data warehousing. OLAP allows designers to model data in a multidimensional way as hyper-cubes. ROLAP snowflakes and stars as well as MOLAP cubes do not 7 offer a visualization of data structures independently from implementation issues. Therefore, they do not ensure a sound data warehouse conceptual design. Our design method uses the Unified Modeling Language (UML) at the conceptual level. This choice can be justified along at least three considerations : - The UML is now a well-known language for software engineers. - It provides simple basic constructs to describe at a high level of abstraction the important concepts of the application domain. - It can be easily mapped to relational as well as to multidimensional logical models. Due to these considerations, many authors use the UML notation at a first step of the transactional database design. To the best of our knowledge, it has not been used in ROLAP systems. Our design method consists of a two-step conceptual design : - The first step leads to a UML schema, more precisely to a class diagram without operations. - The second step enriches and transforms this schema in order to facilitate its automatic mapping to the multidimensional model. Four types of operations are conducted : the determination of identifying attributes, the determination of attributes representing measures, the migration of association attributes and the suppression of generalizations. a) Determination of identifying attributes In contrast to the ER model or the relational model, the notion of identifying attribute is not defined in the standard UML notation; this notion is replaced by the concept of object identity. However, we need to determine the identifying attributes of classes in order to define the dimensions of the multidimensional model at the logical level. Note that since an association class is identified by the n-uple of identifiers of the participating classes, the determination of identifiers is necessary only for the other classes (called ordinary classes). For each ordinary class of the UML schema, the user and the data warehouse designer have to decide which attribute or combination of attributes identify the class. If necessary, a specific attribute is created in order to identify the class. Identifying attributes are specified using the UML construct of tagged value: the suffix {id} is added to each identifying attribute, as in [Morley et al, 2000]. This process can be synthesized by the following rule : Rule Rcc1: Each attribute of an ordinary class is either an identifying attribute or not. 8 b) Determination of attributes representing measures We differentiate between attributes representing measures, and attributes which can be defined as qualitative values. As described in the previous section, this distinction is not based on data types even if, generally, measures are numerical and qualitative attributes are not. Therefore this differentiation cannot be performed automatically. The user and the data warehouse designer have to decide which attributes must be considered as measures. Note that this does not deal with identifying attributes determined previously, since an identifying attribute cannot be a measure. In the UML schema, attributes representing measures are specified by the tagged value {meas}. This process can be synthesized by the following rule : Rule Rcc2: Each attribute is either a measure or not. c) Migration of association attributes This step is concerned with 1-1 and 1-N associations having specific attributes (these associations are actually association classes, since an ordinary association cannot bear attributes in UML). Let us mention that this case is rarely encountered. If specific attributes are present in these associations, the designer has first to check the validity of this representation. Even if their presence cannot be questioned, they cannot be mapped into multidimensional models by using hierarchies. The reason is that, in multidimensional models, these hierarchies do not contain information. Therefore, they must migrate from the association to the participating class on the N side. In case of 1-1 association, they can indifferently migrate into one of the two classes. After migrating the attributes of a 1-N or 1-1 association, the latter is transformed into an ordinary association unless it is connected to other associations or classes. The rules for migrating association attributes are expressed as folows : Rule Rcc3 : Each attribute belonging to a 1-1 association is transferred to one of the classes involved in the association. Rule Rcc4 : Each attribute belonging to a 1-N association is transferred to the N-class, i.e. the class involved several times in the association. d) Suppression of generalizations The inheritance links of the UML notation cannot be mapped directly to hierarchies in the multidimensional model, since the semantics of hierarchies in object-oriented models and multidimensional model differ. However, we want to preserve the information contained in UML generalizations and transform these hierarchies to enable their correct mapping to multidimensional hierarchies in the logical phase. To this end, we transform the generalizations into aggregations and classes following the proposal of [Moody & Kortink, 2000] for ER models. 9 We have adapted this rule to UML and extended it to consider the different cases of partial specialization and/or overlapping specialization. The corresponding rule is informally described below : Rule Rcc5 : For each level i of specialization of a class C, a class named Type-C-i is created. The occurrences of these classes define all the specializations of C. In case of overlapping between specializations, a special value is created for each overlapping between two or more sub-classes of C. In case of partial specialization, the special value “others” is created. A N-1 aggregation is created between the classes C and Type-C-i. CONCEPTUAL RULES Attribute Rcc2 X X 1-1 association attribute Rcc3 Rcc4 Rcc5 X 1-N association attribute X Generalization X Identifying attribute CONCEPTUAL Rcc1 X Attribute - measure X Attribute - not a measure X Class attribute X X Class X N-1 aggregation X Figure 4: Transformation of the UML schema These rules are sketched in Figure 4. Thanks to these transformations, the resulting UML schema can then be automatically mapped into a logical multidimensional schema, as described in the following section. 3.2. Logical design The aim of the logical design phase is to convert the enriched UML conceptual schema to a logical one expressed with the multidimensional concepts of our unified model. This schema is generated using specific rules represented as production rules (Figure 5). 10 RULES Rcl1 Rcl2 Rcl3 Rcl4 Rcl5 Rcl6 CONCEPTUAL Identifier of an ordinary class X Non-identifying attribute of an ordinary class, which is not a measure of interest Non-identifying attribute of an ordinary class, which is a measure of interest X X Attribute of an association class X Path formed by N-1 associations X N-M or N-ary association without at least one attribute that is always defined LOGICAL Dimension X X Measure X X Dummy measure X Dimension attribute X Hierarchy X Figure 5 : Conceptual UML towards logical multidimensional concepts mapping rules Figure 5 presents the main rules to be used. Needless to say that mapping an enriched UML schema to a multidimensional schema requires more rules. The rules map first the ordinary classes – classes that are not association classes – and their attributes (rules Rcl1 to Rcl3) and then the associations – association classes or ordinary associations – and their attributes (rules Rcl4 to Rcl6). The ordinary classes of the conceptual UML model are mapped by rule Rcl1: Rule Rcl1: The identifier of each ordinary class is mapped into a dimension in the multidimensional model. If an identifier is composed of several attributes in the UML model, it is mapped into a single dimension, formed by concatenating the attributes composing the identifier. The attributes of ordinary classes are mapped by rules Rcl2 and Rcl3: Rule Rcl2: The non-identifying attributes of each ordinary class are mapped into dimension attributes in the multidimensional model if these non-identifying attributes are not measures of interest. The multidimensional attributes are associated with the dimension obtained by mapping the identifier of the ordinary class (rule Rcl1). Rule Rcl3: The non-identifying attributes of each ordinary class are mapped into measures in the multidimensional model if these non-identifying attributes are measures of interest. The measures are associated with the dimension obtained by mapping the identifier of the ordinary class (rule Rcl1). The attributes of association classes are mapped using rule Rcl4 : 11 Rule Rcl4: The attributes of each association class are mapped into measures, associated with dimensions obtained by mapping the identifiers of the ordinary classes directly or indirectly participating in the association class (rule Rcl1). If the association class bearing the attributes has one (or several) participating class(es) with a maximal cardinality of 1, the dimension(s) obtained by mapping the identifier(s) of this (these) class(es) should be indicated between parentheses in the specification of the measures, to express the fact that the dimension(s) are not necessary to functionally determine the measures. Finally, associations are mapped using rules Rcl5 (for binary N-1 associations) and Rcl6 (for other associations) : Rule Rcl5: A path formed by N-1 associations is mapped into a hierarchy in the multidimensional model. Very often, N-1 associations are aggregations or compositions. Therefore, the hierarchies of the multidimensional model generally correspond to aggregation/composition paths in the UML model. Rule Rcl6: Every N-M or N-ary association without at least one attribute that is always defined is mapped into a dummy measure, associated with dimensions obtained by mapping the identifiers of the ordinary classes directly or indirectly participating in the association (rule Rcl1). Note that if an N-M association or an N-ary association has attributes, these attributes have already been mapped into measures (rule Rcl4). If one of these attributes is always defined, the corresponding measure is also always defined, making the definition of a “dummy measure” unnecessary. At the end of the logical design phase, the universe of discourse is described through a unified multidimensional schema. Depending on the decisional tool to be used, this schema must be implemented, i.e. mapped into physical concepts. 3.3. Physical design The physical design phase depends heavily on the target system. ROLAP systems implement the multidimensional schema in a relational database system (RDBMS). This category of systems may be subdivided depending on the model used to implement the multidimensional schema in the RDBMS. The models used are typically the star model, the snowflake model, or any combination or extension of these two models. For each type of target system i.e. for each physical model, a specific set of mapping rules from the logical multidimensional model has to be defined. In this paper, we consider two types of target systems : ROLAP star and ROLAP snowflake. Figure 6 describes the mapping rules, assuming the star model is used for implementation. 12 LOGICAL RULES Rls1 Dimension which dimensions at least one measure X Non-dummy measure Rls2 Rls3 Rls4 Rls5 X Dummy measure X Hierarchy X PHYSICAL Dimension attribute X Dimension table X Fact table X Fact table attribute X Dimension table attribute X X X Figure 6 : Logical multidimensional towards physical ROLAP star concepts mapping rules The dimensions of the logical model are mapped by rule Rls1: Rule Rls1: Every dimension dimensioning at least one measure is mapped into a dimension table and an associated primary key. Note that the logical dimensions that do not dimension any measure, i.e. the ones that only participate in hierarchies, will be taken care of by rule Rls4. Measures are mapped using rules Rls2 and Rls3: Rule Rls2: Every non-dummy measure is mapped into a fact table attribute (i.e. a fact) in table T, whose foreign keys correspond to the logical dimensions of the measure and whose primary key corresponds to the subset of these dimensions which are not indicated in parentheses. If table T does not exist, it is defined when mapping the measure. Note that the logical dimensions indicated in parentheses in the specification of a measure are the ones that are functionally determined by the others. Therefore, they are not used in the definition of the primary key of table T. Rule Rls3: Every dummy measure is mapped into a fact table whose foreign keys correspond to the logical dimensions of the measure and whose primary key corresponds to the subset of these dimensions which are not indicated in parentheses. The fact tables generated by rule Rls3 are thus fact tables with no fact [Kimball, 1996]. The hierarchies of the logical model are mapped using rule Rls4: Rule Rls4: Every hierarchy D1->D2->…->Dn of the logical model is mapped by considering all the subhierarchies Dj->Dj+1…->Dn where 1<=j<n and Dj dimensions at least one measure. A sub-hierarchy Dj>Dj+1…->Dn is mapped in the physical model by defining in the dimension table identified by Dj an attribute corresponding to each of the Di (where j<i<=n). Dimension attributes are mapped with rule Rls5: 13 Rule Rls5: Every attribute of every dimension Di of the logical model is mapped into a dimension table attribute, in all the dimension tables which possess an (identifying or non-identifying) attribute corresponding to Di. Figure 7 describes the mapping rules, assuming the snowflake model is used for implementation. The difference between the star and snowflake models lies in the explicit representation of hierarchies in the snowflake model. As a result, all the dimensions of the logical model are mapped into dimension tables in the physical model (rule Rlf1); the link between any consecutive dimensions (Di,Di+1) in any hierarchy is mapped into a foreign key (dimension table attribute + referential integrity constraint) in the dimension table corresponding to Di (rule Rlf4); each attribute of each dimension Di of the logical model is mapped into a dimension table attribute in the dimension table corresponding to Di (rule Rlf5). RULES Rlf1 LOGICAL Dimension Rlf2 Rlf3 Rlf4 X Non-dummy measure X Dummy measure X Hierarchy X Dimension attribute PHYSICA Rlf5 X Dimension table X Fact table X Fact table attribute X Dimension table attribute Referential integrity constraint X X X X Figure 7 : Logical multidimensional towards physical ROLAP snowflake concepts mapping rules The rules proposed above allow the designer to map a multidimensional logical model to the most standard ROLAP physical models. The next section illustrates this design method through an example. 4. Case study A firm is faced with the definition of an optimal media-planning system. The company wishes to launch advertising campaigns for its products using several types of media (radio, TV, newspapers, magazines, etc.). Its objective is to maximize the number of consumers being exposed to the advertising campaign. To support the decision-making process, we need to define a multidimensional model with all the data relevant to the mediaplanning problem. 14 4.1. Conceptual design The conceptual schema is represented in Figure 8. It contains data related to the products concerned by the advertising campaigns. The consumers are represented as targets located at different regions. The consumers are defined according to their purchase behavior over time which is strongly influenced by the advertising campaigns. The consumers are exposed to several types of media. This exposure is measured over time. The schema includes all the key information about the media shareholding. The real schema has been simplified in order to be more readable. Product_type Media_type 1..* product_type product_unit * media_type insertion may_be_advertised_in 1 1 * * Product Media Region product_code product_name region number_of_inhabitants percentage_of_region 1..* 1..* media_name advertising_price gets * * 1 for * * 1 exposure consumption 1..* * product_consumption Target * * * target_code status minimum_age maximum_age sex is_strongly_influenced_by main_shareholder media_exposure 1 Shareholder shareholder_name * Advertising_campaign Year campaign_code year * * Private_shareholder Public_shareholder 1 during * 1 public_shareholder_level 1..* Quarter * quarter 1 in Person 1..* Date Company manager_name * dd_mm_yy Figure 8 : The UML conceptual schema for the media-planning example The UML schema is then enriched by determination of identifying attributes {id}, attributes representing measures {meas}, migration of association attributes and suppression of generalizations. Two classes Typeshareholder-1 and Type-shareholder-2 are created and renamed as Shareholder_type and Private_shareholder_type. The set of occurrences of Shareholder_type is {private, public, both}. The set of occurrences of Private_shareholder_type is {person, company, others}. The attribute percentage_of_region is transferred to the class Target (Rule Rcc4). The result is represented in Figure 9. 15 Product_type Media_type 1..* product_type {id} product_unit * media_type {id} insertion may_be_advertised_in 1 1 * * Product Media Region product_code {id} product_name region {id} number_of_inhabitants {meas} * 1..* * 1 * * * 1 for main_shareholder 1..* consumption Target product_consumption {meas} target_code {id} status minimum_age maximum_age sex percentage_of_region {meas} * is_strongly_influenced_by * * exposure media_exposure {meas} 1 Shareholder shareholder_name {id} public_shareholder_level manager_name * Advertising_campaign Year campaign_code {id} year {id} * * 1..* media_name {id} advertising_price gets * * 1 during * 1 1..* Quarter * quarter {id} 1 Shareholder_type 1 in shareholder_type {id} 1 Private_shareholder_type private_shareholder_type {id} 1..* Date * dd_mm_yy {id} Figure 9 : Enriched/transformed UML schema 4.2. Logical design We give below the unified multidimensional representation resulting from the application of the mapping rules (Figure 10). Measures are either in trapezoidal boxes or in italics inside dimension boxes. Dummy measures are named using relationship names. A dotted line between the trapezoidal box of a measure and the rectangle box of a dimension indicates that the dimension is not necessary to functionally determine the measure (i.e., the dimension is represented between parentheses in the textual specification of the measure). 16 PRODUCT_ TYPE product_type product_unit Dummy_may_be_advertised_in MEDIA_TYPE media_type insertion REGION region number_of_inhabitants PRODUCT Dummy_gets MEDIA product_code product_name media_name advertising_price TARGET Dummy_is_ strongly_ influenced_ by ADVERTISING_ CAMPAIGN product_consumption target_code status minimum_age maximum_age sex percentage_of_region YEAR year media_exposure Dummy_main_shareholder SHAREHOLDER campaign_code shareholder_name public_shareholder_level manager_name QUARTER quarter Dummy_in DATE dd_mm_yy SHAREHOLDER_TYPE PRIVATE_SHAREHOLDER_TYPE shareholder_type private_shareholder_type Figure 10 : Media-planning multidimensional schema This graphical representation allows the designer and the users to immediately recognize and differentiate the dimensions, measures and hierarchies. 4.3. Physical design The mapping rules described in Section 3.3 can also be applied to derive ROLAP physical schemas. Figure 11 presents the result of the application of the mapping rules in case the star model is used. Note that since many dimension tables are shared by different fact tables, Figure 11 actually represents a constellation of stars. 17 PRODUCT_ TYPE product_type product_unit REGION PRODUCT product_code product_name product_type product_unit CONSUMPTION FK product_code FK target_code FK quarter product_consumption ADVERTISING_ CAMPAIGN campaign_code quarter year product_code product_name product_type product_unit region MAY_BE_ADVERTISED_IN MEDIA_TYPE FK product_type FK media_type media_type insertion MEDIA REGION_FIGURES GETS FK region number_of_inhabitants FK region FK media_name TARGET target_code status minimum_age maximum_age sex region media_name advertising_price media_type insertion TARGET_FIGURES FK target_code percentage_of_region EXPOSURE FK media_name FK target_code FK quarter media_exposure QUARTER quarter year SHAREHOLDER shareholder_name public_shareholder_level manager_name shareholder_type private_shareholder_type DATE dd_mm_yy quarter year MAIN_SHAREHOLDER FK media_name FK dd_mm_yy FK shareholder_name IS_STRONGLY_INFLUENCED_BY FK product_code FK target_code FK quarter FK campaign_code IN FK campaign_code FK media_name Figure 11 : Media-planning ROLAP star physical schema After the definition of the physical schema, the data confrontation phase is performed in order to map the physical schema data elements with the data sources. Since it is beyond the scope of this paper, this confrontation phase is not performed. 5. State-of-the-Art The deficit in data warehouse design is real. Very few methods have been proposed until now. Let us mention [Akoka & Prat, 1997 ; Golfarelli et al., 1998 ; Cabibbo & Torlone, 1998 ; Moody & Kortink, 2000]. Unlike papers describing design methods, as stated by [Sapia et al., 1998], a fair number of publications is available concerning multidimensional data modeling but with very few recognizing the importance of the separation of conceptual, logical and physical issues. Moreover, even if the three levels are considered, some 18 confusion exists between, on the one hand, conceptual and logical models and, on the other hand, logical and physical models. A real confusion also seems to exist between the conceptual and physical aspects. As an example, the multidimensional modeling manifesto of Kimball is inadequate for conceptual modeling [Kimball, 1997]. His approach tends to include physical design issues, especially with his propositions of star and snowflake schemas which appear to be not independent from implementation issues. a) Conceptual-logical models [Cabibbo & Torlone, 1998] focus on logical design issues and propose a logical model for OLAP systems. They assume that there exists an integrated ER schema of the operational data sources. They provided a methodology to transform the ER schema into a dimensional graph. [Golfarelli & Rizzi, 1998] proposed a conceptual model called Dimensional-Fact schema. They provided a methodology to transform the ER model of the data sources into a Dimensional-Fact model. Note that their approach is not based on a formal data model. [Lehner et al., 1998] proposed a conceptual multidimensional model which includes some mechanism to structure qualifying information. Note that no formal graphical notation is provided. [Sapia et al., 1998] presented a specialization of the ER model called Multidimensional Entity-Relationship Model (MER) expressing the multi-dimensional structure of the data by means of two specialized relationship sets and a specialized entity set. Their approach models user requirements independently from the structure of the data sources. [Gyssens & Laskshmanan, 1997] proposed a multi-dimensional database model providing the functionalities necessary for OLAP-based applications. They made a clear separation between the structural aspects and the contents, allowing them to define data manipulation languages in a transparent way. They defined an algebra and a calculus. b) Logical-physical models An interesting survey of logical models for OLAP databases can be found in [Vassiliadis & Sellis, 1999]. The main features of these models are that they offer systematically a logical view of data to be queried by a set of operators and, usually, a set of implementation mechanisms. Among these models, let us mention [Agrawal et al., 1997] who provide a logical model in which dimensions and measures are treated in a symmetric way and where multiple hierarchies of dimensions allow ad hoc aggregates. [Pedersen & Jensen, 1999] propose a multidimensional logical data model for complex data, justifying nine requirements to be satisfied in order to support complex data. They showed that their model covers the nine requirements in a better way than previous models 19 [Rafanelli & Shoshani, 1990 ; Agrawal et al, 1997 ; Gray et al, 1996 ; Kimball, 1996 ; Li & Wang, 1996 ; Gyssens & Laskshmanan, 1997 ; Datta & Thomas, 1997 ; Lehner, 1998]. [Harinarayan et al, 1996] investigated mainly physical design and implementation issues. Finally, our method is an attempt to define a generic framework based on the following principles : - it makes a clear distinction between the classical steps of database design (conceptual, logical, physical), - it unifies the different multidimensional concepts into a single and generic model, - it can be adapted to all OLAP vendor engines, - it capitalizes on the existing UML schemas, - it is based on well-established UML concepts. 6. Conclusion and Further Research We have described a method for designing and developing data warehouses. Capitalizing on database design techniques, we proposed a conceptual design phase based on UML notation, followed by an enrichment/transformation of this schema. This enrichment/transformation allows the designer to automatically convert this conceptual representation into a logical multidimensional model. At this step, we proposed a generic multi-dimensional model independent from implementation issues and unifying ROLAP and MOLAP concepts. Using mapping rules, this generic logical schema can be mapped to any physical multi-dimensional platform. A case study was described to illustrate the main features of the method. Several questions still remain open. A formal foundation for the unifying multidimensional model needs to be developed. The different mapping rules sets have to be enriched and efficiently implemented. A reverse engineering approach taking into account existing data warehouses and/or data marts must be developed [Akoka & Comyn-Wattiau, 1999]. An intensive real life case study should be undertaken and used to validate the approach. We are currently working on these issues. 7. References [Agrawal et al, 1997] R. Agrawal, A. Gupta, S. Sarawagi, “Modeling multidimensional databases”, 13th International Conference on Data Engineering (ICDE ’97), Birmingham, UK, April. [Akoka & Comyn-Wattiau, 1999] J. Akoka, I. Comyn-Wattiau, “Rétro-conception des « datawarehouses » et des systèmes multidimensionnels”, 17ème Congrès INFORSID, La Garde, June. [Akoka & Prat, 1997] J. Akoka, N. Prat, “Modélisation logique des données dans les Systèmes Multidimensionnels d'Aide à la Décision : la méthode MAP”, Revue des Systèmes de Décision, Vol 6(2), June. 20 [Blaschka et al, 1998] M. Blaschka, C.Sapia, G.Höfling, B.Dinter, “Finding your way through multidimensional data models”, DEXA Workshop on Data Warehouse Design and OLAP Technology (DWDOT ’98), Vienna, Austria. [Cabibbo & Torlone, 1998] L. Cabibbo, R. Torlone, “A Logical Approach to Multidimensional Databases”, Proceedings of 6th International Workshop on Extending Database Technology (EDBT’1998), Valencia (Spain), March. [Chaudhuri & Dayal, 1997] S.Chaudhuri, U.Dayal, “An overview of data warehousing and OLAP Technology”, SIGMOD Record, volume 26, number 1, March. [Datta & Thomas, 1997] A. Datta, H. Thomas, “A Conceptual Model and Algebra for On-Line Analytical Processing in Decision Support Databases”, Proceedings of WITS. [Golfarelli et al, 1998] M.Golfarelli, D.Maio, S.Rizzi, “Conceptual design of data warehouses from E/R schemes”, 31st Hawaii International Conference on System Sciences, Hawaii, USA, January. [Golfarelli & Rizzi, 1998] M.Golfarelli, S.Rizzi, “A methodological framework for data warehousing design”, ACM workshop on data warehousing and OLAP. [Gray et al, 1996] J.Gray et al, “Data Cube : A Relational Aggregation Operator Generalizing Group-By, CrossTab and Sub-Totals”, Proceedings of ICDE. [Gyssens & Lakshmanan, 1997] M.Gyssens, L.V.S.Lakshmanan, “A foundation for multi-dimensional databases” 23rd VLDB Conference, Athens, Greece. [Harinarayan et al, 1996] V. Harinarayan, A. Rajaraman, J.D. Ullman, “Implementing Data Cubes Efficiently”, Proceedings of SIGMOD conference. [Hüsemann et al, 2000] B.Hüsemann, J.Lechtenbörger, G.Vossen, “Conceptual data warehouse design”, 2nd International Workshop on Design and Management of Data Warehouses (DMDW 2000), Stockholm, Sweden, June. [Jacobson et al, 1999] I. Jacobson, G. Booch, J. Rumbaugh, “The Unified Software Development Process”, Addison Wesley Publishing Company. [Kimball, 1996] R.Kimball, “The data warehouse toolkit”, John Wiley & Sons. [Kimball, 1997] R. Kimball, “A Dimensional Modeling Manifesto”, DBMS on-line, http://www.dbmsmag.com/. [Lehner, 1998] W. Lehner, “Modeling Large Scale OLAP Scenarios”, Proceedings of EDBT. [Lehner et al, 1998] W. Lehner, J. Albrecht, H. Wedekind, “Normal Forms for Multidimensional Databases”, Proceedings 10th SSDBM conference, Italy, July. [Li & Wang, 1996] C.Li, X.S.Wang, “A data model for supporting on-line analytical processing”, Proceedings Conference on Information and Knowledge Management (CIKM ’96), Baltimore, USA, November. [Moody & Kortink, 2000] D.L. Moody, M.A.R. Kortink, “From Enterprise Models to Dimensional Models : A Methodology for Data Warehouse and Data Mart Design”, 2nd International Workshop on Design and Management of Data Warehouses (DMDW 2000), Stockholm, Sweden, June. [Morley et al, 2000] C. Morley, J. Hugues, B. Leblanc, "UML, pour l'analyse d'un système d'information", Informatique Dunod, Paris. [OLAP report, 2001] The OLAP Report – Market share analysis, http://www.olapreport.com/Market.htm [OMG, 2001] Unified Modeling Language, http://www.omg.org/technology/documents/formal [Pedersen & Jensen, 1999] T.B.Pedersen, C.S.Jensen, “Multidimensional data modeling for complex data”, 15th International Conference on Data Engineering (ICDE ’99), Sydney, Australia, March. [Rafanelli & Shoshani, 1990] M. Rafanelli, A. Shoshani, “STORM : A Statistical Object Representation Model”, Proc. of SSDBM. [Sapia et al, 1998] C. Sapia, M. Blaschka, G. Höfling, B. Dinter, “Extending the E/R Model for the Multidimensional Paradigm”, International Workshop on Data Warehousing and Data Mining in conjunction with ER98, Singapore, 1998. [Vassiliadis, 1999] P. Vassiliadis, “Gulliver in the land of data warehousing : practical experiences and observations of a researcher”, Proceedings of the International Workshop on Design and Management of Data Warehouses (DMDW’2000), Stockholm, June. [Vassiliadis & Sellis, 1999] P.Vassiliadis, T.Sellis, “A survey of logical models for OLAP databases”, SIGMOD Record, volume 28, number 4, December. 21 ESSEC CE NTRE DE RECHERCHE LISTE DES DOCUMENTS DE RECHERCHE DU CENTRE DE RECHERCHE DE L’ESSEC (Pour se procurer ces documents, s’adresser au CENTRE DE RECHERCHE DE L’ESSEC) LISTE OF ESSEC RESEARCH CENTER WORKING PAPERS (Contact the ESSEC RESEARCH CENTER for information on how to obtain copies of these papers) [email protected] 1997 97001 BESANCENOT D., VRANCEANU Radu Reputation in a Model of Economy-wide Privatization. 97002 GURVIEZ P. The Trust Concept in the Brand-consumers Relationship. 97003 POTULNY S. L’utilitarisme cognitif de John Stuart Mill. 97004 LONGIN François From Value at Risk to Stress Testing: The Extreme Value Approach. 97005 BIBARD Laurent, PRORIOL G. Machiavel : entre pensée du pouvoir et philosophie de la modernité. 97006 LONGIN François Value at Risk: une nouvelle méthode fondée sur la théorie des valeurs extrêmes. 97007 CONTENSOU François, VRANCEANU Radu Effects of Working Time Constraints on Employment: A Two-sector Model. 97008 BESANCENOT D., VRANCEANU Radu Reputation in a Model of Exchange Rate Policy with Incomplete Information. 97009 AKOKA Jacky, BRIOLAT Dominique, WATTIAU Isabelle La reconfiguration des processus inter-organisationnels. 97010 NGUYEN. P Bank Regulation by Capital Adequacy and Cash Reserves Requirements. 97011 LONGIN François Beyond the VaR. 97012 LONGIN François Optimal Margin Level in Futures Markets: A Method Based on Extreme Price Movements. 97013 GROUT DE BEAUFORT Viviane Maastricth II ou la copie à réviser. 97014 ALBIGOT J.G., GROUT DE BEAUFORT V., BONFILLON P.O., RIEGER B . Perspectives communautaires et européennes sur la réduction du temps de travail. Page 1 97015 DEMEESTERE René, LORINO Philippe, MOTTIS Nicolas Business Process Management: Case Studies of Different Companies and Hypotheses for Further Research. 97016 PERETTI Jean-Marie, HOURQUET P.G., ALIS D. Hétérogénéité de la perception des déterminants de l’équité dans un contexte international. 97017 NYECK Simon, ROUX Elyette WWW as a Communication Tool for Luxury Brands: Compared Perceptions of Consumers and Managers. 97018 NAPPI-CHOULET Ingrid L’analyse économique du fonctionnement des marchés immobiliers. 97019 BESANCENOT D., ROCHETEAU G., VRANCEANU Radu Effects of Currency Unit Substitution in a Search Equilibrium Model. 97020 BOUCHIKHI Hamid Living with and Building on Complexity: A Constructivist Perspective on Organizations. 97021 GROUT DE BEAUFORT V., GRENOT S., TIXIER A . TSE K.L Essai sur le Parlement Européen. 97022 BOULIER J.F., DALAUD R., LONGIN François Application de la théorie des valeurs extrêmes aux marchés financiers. 97023 LORINO Philippe Théorie stratégique : des approches fondées sur les ressources aux approches fondées sur les processus. 97024 VRANCEANU Radu Investment through Retained Earnings and Employment in Transitional Economies. 97025 INGHAM M., XUEREB Jean-Marc The Evolution of Market Knowledge in New High Technology Firms: An Organizational Learning Perspective. 97026 KOENING Christian Les alliances inter-entreprises et la coopération émergente. 97027 LEMPEREUR Alain Retour sur la négociation de positions : pourquoi intégrer l’autre dans mon équation personnelle ? 97028 GATTO Riccardo Hypothesis Testing by Symbolic Computation. 97029 GATTO Riccardo , JAMMALAMADAKA S. Rao A conditional Saddlepoint Approximation for Testing Problems. 97030 ROSSI (de) F.X., GATTO Riccardo High-order Asymptotic Expansions for Robust Tests. 97031 LEMPEREUR Alain Negotiation and Mediation in France: The Challenge of Skill-based Learnings and Interdisciplinary Research in Legal Education. 97032 LEMPEREUR Alain Pédagogie de la négociation : allier théorie et pratique. 97033 WARIN T. Crédibilité des politiques monétaires en économie ouverte. 97034 FRANCOIS P. Bond Evaluation with Default Risk: A Review of the Continuous Time Approach. 97035 FOURCANS André, VRANCEANU Radu Fiscal Coordination in the EMU: A Theoretical and Policy Perspective. Page 2 97036 AKOKA Jacky, WATTIAU Isabelle MeRCI: An Expert System for Software Reverse Engineering. 97037 MNOOKIN R. (traduit par LEMPEREUR Alain) Surmonter les obstacles dans la résolution des conflits. 97038 LARDINOIT Thierry, DERBAIX D. An Experimental Study of the Effectiveness of Sport Sponsorship Stimuli. 97039 LONGIN François, SOLNIK B. Dependences Structure of International Equity Markets during Extremely Volatile Periods. 97040 LONGIN François Stress Testing : application de la théorie des valeurs extrêmes aux marchés des changes. 1998 98001 TISSOT (de) Olivier Quelques observations sur les problèmes juridiques posés par la rémunération des artistes interprètes. 98002 MOTTIS Nicolas, PONSSARD J.P. Incitations et création de valeur dans l’entreprise. Faut-il réinventer Taylor ? 98003 LIOUI A., PONCET Patrice Trading on Interest Rate Derivatives and the Costs of Marking-to-market. 98004 DEMEESTERE René La comptabilité de gestion : une modélisation de l’entreprise ? 98005 TISSOT (de) Olivier La mise en œuvre du droit à rémunération d’un comédien ayant « doublé » une œuvre audiovisuelle er (film cinématographique ou fiction télévisée ) avant le 1 janvier 1986. 98006 KUESTER Sabine, HOMBURG C., ROBERTSON T.S. Retaliatory Behavior to New Product Entry. 98007 MONTAGUTI E., KUESTER Sabine, ROBERTSON T.S. Déterminants of « Take-off » Time for Emerging Technologies: A Conceptual Model and Propositional Inventory. 98008 KUESTER Sabine, HOMBURG C . An Economic Model of Organizational Buying Behavior. 98009 BOURGUIGNON Annick Images of Performance: Accounting is not Enough. 98010 BESANCENOT D., VRANCEANU Radu A model of Manager Corruption in Developing Countries with Macroeconomic Implications. 98011 VRANCEANU Radu, WARIN T. Une étude théorique de la coordination budgétaire en union monétaire. 98012 BANDYOPADHYAU D. K. A Multiple Criteria Decision Making Approach for Information System Project Section. 98013 NGUYEN P., PORTAIT Roland Dynamic Mean-variance Efficiency and Strategic Asset Allocation with a Solvency Constraint. 98014 CONTENSOU François Heures supplémentaires et captation du surplus des travailleurs. 98015 GOMEZ M.L. De l’apprentissage organisationnel à la construction de connaissances organisationnelles. Page 3 98016 BOUYSSOU Denis Using DEA as a Tool for MCDM: some Remarks. 98017 INDJEHAGOPIAN Jean-Pierre, LANTZ F., SIMON V. Dynamique des prix sur le marché des fiouls domestiques en Europe. 98019 PELISSIER-TANON Arnaud La division du travail, une affaire de prudence. 98020 PELISSIER-TANON Arnaud Prudence et qualité totale. L’apport de la philosophie morale classique à l’étude du ressort psychologique par lequel les produits satisfont les besoins de leurs utilisateurs. 98021 BRIOLAT Dominique, AKOKA Jacky, WATTIAU Isabelle Le commerce électronique sur Internet. Mythe ou réalité ? 98022 DARMON René Equitable Pay for the Sales Force. 98023 CONTENSOU François, VRANCEANU Radu Working Time in a Model of Wage-hours Negociation. 98024 BIBARD Laurent La notion de démocratie. 98025 BIBARD Laurent Recherche et expertise. 98026 LEMPEREUR Alain Les étapes du processus de conciliation. 98027 INDJEHAGOPIAN Jean-Pierre, LANTZ F., SIMON V. Exchange Rate and Medium Distillates Distribution Margins. 98028 LEMPEREUR Alain Dialogue national pour l’Europe. Essai sur l’identité européenne des français. 98029 TIXIER Maud What are the Implications of Differing Perceptions in Western, Central and Eastern Europe for Emerging Management. 98030 TIXIER Maud Internal Communication and Structural Change. The Case of the European Public Service: Privatisation And Deregulation. 98031 NAPPI-CHOULET Ingrid La crise des bureaux : retournement de cycle ou bulle ? Une revue internationale des recherches. 98032 DEMEESTERE René La comptabilité de gestion dans le secteur public en France. 98033 LIOUI A., PONCET Patrice The Minimum Variance Hedge Ratio Revisited with Stochastic Interest Rates. 98034 LIOUI A., PONCET Patrice Is the Bernoulli Speculator always Myobic in a Complete Information Economy? 98035 LIOUI A., PONCET Patrice More on the Optimal Portfolio Choice under Stochastic Interest Rates. 98036 FAUCHER Hubert The Value of Dependency is Plant Breeding: A Game Theoretic Analysis. 98037 BOUCHIKHI Hamid, ROND (de) Mark., LEROUX V. Alliances as Social Facts: A Constructivist of Inter-Organizational Collaboration. 98038 BOUCHIKHI Hamid, KIMBERLY John R. Page 4 In Search of Substance: Content and Dynamics of Organizational Identity. 98039 BRIOLAT Dominique, AKOKA Jacky, COMYN-WATTIAU Isabelle Electronic Commerce on the Internet in France. An Explanatory Survey. 98040 CONTENSOU François, VRANCEANU Radu Réduction de la durée du travail et complémentarité des niveaux de qualification. 98041 TIXIER Daniel La globalisation de la relation Producteurs-Distributeurs. 98042 BOURGUIGNON Annick L’évaluation de la performance : un instrument de gestion éclaté. 98043 BOURGUIGNON Annick Benchmarking: from Intentions to Perceptions. 98044 BOURGUIGNON Annick Management Accounting and Value Creation: Value, Yes, but What Value? 98045 VRANCEANU Radu A Simple Matching Model of Unemployment and Working Time Determination with Policy Implications. 98046 PORTAIT Roland, BAJEUX-BESNAINOU Isabelle Pricing Contingent Claims in Incomplete Markets Using the Numeraire Portfolio. 98047 TAKAGI Junko Changes in Institutional Logics in the US. Health Care Sector: A Discourse Analysis. 98048 TAKAGI Junko Changing Policies and Professionals: A Symbolic Framework Approach to Organizational Effects on Physician Autonomy. 98049 LORINO Philippe L’apprentissage organisationnel bloquée (Groupe Bull 1986-1992) : du signe porteur d’apprentissage au Piège de l’habitude et de la représentation-miroir. 98050 TAKAGI Junko, ALLES G. Uncertainty, Symbolic Frameworks and Worker Discomfort with Change. 1999 99001 CHOFFRAY Jean-Marie Innovation et entreprenariat : De l’idée… au Spin-Off. 99002 TAKAGI Junko Physician Mobility and Attidudes across Organizational Work Settings between 1987 and 1991. 99003 GUYOT Marc, VRANCEANU Radu La réduction des budgets de la défense en Europe : économie budgétaire ou concurrence budgétaire ? 99004 CONTENSOU François, LEE Janghyuk Interactions on the Quality of Services in Franchise Chains: Externalities and Free-riding Incentives. 99005 LIOUI Abraham, PONCET Patrice International Bond Portfolio Diversification. 99006 GUIOTTO Paolo, RONCORONI Andrea Infinite Dimensional HJM Dynamics for the Term Structure of Interest Rates. 99007 GROUT de BEAUFORT Viviane, BERNET Anne-Cécile Les OPA en Allemagne. Page 5 99008 GROUT de BEAUFORT Viviane, GENEST Elodie Les OPA aux Pays-Bas. 99009 GROUT de BEAUFORT Viviane Les OPA en Italie. 99010 GROUT de BEAUFORT Viviane, LEVY M. Les OPA au Royaume-Uni. 99011 GROUT de BEAUFORT Viviane, GENEST Elodie Les OPA en Suède. 99012 BOUCHIKHI Hamid, KIMBERLY John R. st The Customized Workplace: A New Management Paradigm for the 21 Century. 99013 BOURGUIGNON Annick The Perception of Performance Evaluation Criteria (1): Perception Styles 99014 BOURGUIGNON Annick Performance et contrôle de gestion. 99015 BAJEUX-BESNAINOU Isabelle, JORDAN J., PORTAIT Roland Dynamic Asset Allocation for Stocks, Bonds and Cash over Long Horizons. 99016 BAJEUX-BESNAINOU Isabelle, JORDAN J., PORTAIT Roland On the Bonds-stock Asset Allocation Puzzle. 99017 TIXIER Daniel La logistique est-elle l’avenir du Marketing ? 99018 FOURCANS André, WARIN Thierry Euroland versus USA: A Theoretical Framework for Monetary Strategies. 99019 GATTO Riccardo, JAMMALAMADAKA S.R. Saddlepoint Approximations and Inference for Wrapped α-stable Circular Models. 99020 MOTTIS Nicolas, PONSSARD Jean-Pierre Création de valeur et politique de rémunération. Enjeux et pratiques. 99021 STOLOWY Nicole Les aspects contemporains du droit processuel : règles communes à toutes les juridictions et procédures devant le Tribunal de Grande Instance. 99022 STOLOWY Nicole Les juridictions civiles d’exception et l’étude des processus dans le droit judiciaire privé. 99023 GATTO Riccardo Multivariate Saddlepoint Test for Wrapped Normal Models. 99024 LORINO Philippe, PEYROLLE Jean-Claude Enquête sur le facteur X. L’autonomie de l’activité pour le management des ressources humaines et pour le contrôle de gestion. 99025 SALLEZ Alain Les critères de métropolisation et les éléments de comparaison entre Lyon et d’autres métropoles françaises. 99026 STOLOWY Nicole Réflexions sur l’actualité des procédures pénales et administratives. 99027 MOTTIS Nicolas, THEVENET Maurice Accréditation et Enseignement supérieur : certifier un service comme les autres… 99028 CERDIN Jean-Luc International Adjustment of French Expatriate Managers. Page 6 99029 BEAUFORT Viviane, CARREY Eric L’union européenne et la politique étrangère et de sécurité commune : la difficile voie de la construction d’une identité de défense européenne. 99030 STOLOWY Nicole How French Law Treats Fraudulent Bankruptcy. 99031 CHEVALIER Anne, LONGIN François Coût d’investissement à la bourse de Paris. 99032 LORINO Philippe Les indicateurs de performance dans le pilotage organisationnel. 99033 LARDINOIT Thierry, QUESTER Pascale Prominent vs Non Prominent Bands: Their Respective Effect on Sponsorship Effectiveness. 99034 CONTENSOU François, VRANCEANU Radu Working Time and Unemployment in an Efficiency Wage Model. 99035 EL OUARDIGHI Fouad La théorie statistique de la décision (I). 2000 00001 CHAU Minh, LIM Terence The Dynamic Response of Stock Prices Under Asymetric Information and Inventory Costs: Theory and Evidence 00002 BIBARD Laurent Matérialisme et spiritualité 00003 BIBARD Laurent La crise du monde moderne ou le divorce de l’occident. 00004 MATHE Hervé Exploring the Role of Space and Architecture in Business Education. 00005 MATHE Hervé Customer Service: Building Highly Innovative Organizations that Deliver Value. 00006 BEAUFORT (de) Viviane L’Union Européenne et la question autrichienne, ses conséquences éventuelles sur le champ de révision de la CIG. 00007 MOTTIS Nicolas, PONSSARD Jean-Pierre Value Creation and Compensation Policy Implications and Practices. 00009 BOURGUIGNON Annick The Perception of Performance Evaluation Criteria (2): Determinants of Perception Styles. 00010 EL OUARDIGHI Fouad The Dynamics of Cooperation. 00011 CHOFFRAY Jean-Marie Innovation et entrepreneuriat : De l’Idée…au Spin-Off. (Version révisée du DR 99001). 00012 LE BON Joël De l’intelligence économique à la veille marketing et commerciale : vers une nécessaire mise au point conceptuelle et théorique. 00013 ROND (de) Mark Reviewer 198 and Next Generation Theories in Strategy. 00014 BIBARD Laurent Amérique latine : identité, culture et management. Page 7 00016 BIBARD Laurent Les sciences de gestion et l’action. 00017 BEAUFORT (de) V. Les OPA au Danemark. 00018 BEAUFORT (de) V. Les OPA en Belgique. 00019 BEAUFORT (de) V. Les OPA en Finlande. 00020 BEAUFORT (de) V. Les OPA en Irlande. 00021 BEAUFORT (de) V. Les OPA au Luxembourg. 00022 BEAUFORT (de) V. Les OPA au Portugal. 00023 BEAUFORT (de) V. Les OPA en Autriche. 00024 KORCHIA Mickael Brand Image and Brand Associations. 00025 MOTTIS Nicolas, PONSSARD Jean-Pierre L’impact des FIE sur les firmes françaises et allemandes : épiphénomène ou influence réelle ? 00026 BIBARD Laurent Penser la paix entre hommes et femmes. 00027 BIBARD Laurent Sciences et éthique (Notule pour une conférence). 00028 MARTEL Jocelyn, C.G. FISHER Timothy Empirical Estimates of Filtering Failure in Court-supervised Reorganization. 00029 MARTEL Jocelyn Faillite et réorganisation financière : comparaison internationale et évidence empirique. 00030 MARTEL Jocelyn, C.G. FISHER Timothy The Effect of Bankruptcy Reform on the Number of Reorganization Proposals. 00031 MARTEL Jocelyn, C.G. FISHER Timothy The Bankruptcy Decision: Empirical Evidence from Canada. 00032 CONTENSOU François Profit-sharing Constraints, Efforts Output and Welfare. 00033 CHARLETY-LEPERS Patricia, SOUAM Saïd Analyse économique des fusions horizontales. 00034 BOUYSSOU Denis, PIRLOT Marc A Characterization of Asymmetric Concordance Relations. 00035 BOUYSSOU Denis, PIRLOT Marc Nontransitive Decomposable Conjoint Measurement. 00036 MARTEL Jocelyn, C.G. FISHER Timothy A Comparison of Business Bankruptcies across Industries in Canada, 1981-2000. Page 8 2001 01001 DEMEESTERE René Pour une vue pragmatique de la comptabilité. 01003 EL OUARDIGHI Fouad, GANNON Frédéric The Dynamics of Optimal Cooperation. 01004 DARMON René Optimal Salesforce Quota Plans Under Salesperson Job Equity Constraints. 01005 BOURGUIGNON Annick, MALLERET Véronique, NORREKLIT Hanne Balanced Scorecard versus French tableau de bord : Beyond Dispute, a Cultural and Ideological Perspective. 01006 CERDIN Jean-Luc Vers la collecte de données via Internet : Cas d’une recherche sur l’expatriation. 01007 LADHARI Riahd, MORALES Miguel, NYECK Simon, Pons Franck Profils d’attitudes par rapport aux loisirs culturels : L’exemple des spectacles cinématographiques. 01008 NYECK Simon, PARADIS Sylvie, de COSTER Louis, BOURDEAU Laurent Profils de satisfaction des consommateurs par rapport aux événements culturels : L’exemple du festival de jazz. 01009 NYECK Simon, Pons Franck Orientation des consommateurs par rapport aux événements sportifs (oes) : Proposition et validation d’un outil de mesure . 01010 NYECK Simon Représentations masculines des produits cosmétiques : Etude exploratoire auprès de la population "gay". 01011 LADHARI Riahd, MORALES Miguel, NYECK Simon Assessment of SERVQUAL Validity : An Evaluation of 10 Years of Use of the Measurement of the Quality. 01012 VRANCEANU Radu, CERNAT Lucian Globalisation and Growth: New Evidence from Central and Eastern Europe. 01013 BIBARD Laurent De quoi s’occupe la sociologie ? 01014 BIBARD Laurent Introduction aux questions que posent les rapports entre éthique et entreprise. 01015 BIBARD Laurent Quel XXIème siècle pour l’humanité ? 01016 MOTTIS Nicolas, PONSSARD Jean-Pierre Value-based Management at the Profit Center Level. 01017 BESANCENOT Damien, HUYNH Kim, VRANCEANU Radu Public Debt : From Insolvency to Illiquidity Default. 01018 BIBARD Laurent Ethique de la vie bonne et théorie du sujet : nature et liberté, ou la question du corps. 01019 INDJEHAGOPIAN Jean-Pierre, JUAN S . LANTZ F., PHILIPPE F. La pénétration du Diesel en France : tendances et ruptures. 01020 BARONI Michel, BARTHELEMY Fabrice, MOKRANE Mahdi Physical Real Estates: Risk Factors and Investor Behaviour. 01022 BESANCENOT Damien, VRANCEANU Radu Quality Leaps and Price Distribution in an Equilibrium Search model Page 9 01023 BIBARD Laurent Gestion et Politique 01024 BESANCENOT Damien, VRANCEANU Radu Technological Change, Acquisition of Skills and Wages in a search Economy 01025 BESANCENOT Damien, VRANCEANU Radu Quality Uncertainty and Welfare in a search Economy 01026 MOTTIS N. , PONSARD J.P., L’impact des FIE sur le pilotage de l’entreprise 01027 TAPIERO Charles, VALOIS Pierre The inverse Range Process in a Random Volatibility Random Walk 01028 ZARLOWSKI Ph., MOTTIS N. Making Managers into Owners An Experimental Research on the impact of Incentive Schemes on Shareolder Value Creation 01029 BESANCENOT Damien, VRANCEANU Radu Incertitude, bien-être et distribution des salaires dans un modèle de recherche d’emploi 01030 BOUCHIKHI H. De l’entrepreneur au gestionnaire et du gestionnaire à l’entrepreneur 01031 TAPIERO A., SULEM A. Inventory Control wth Supply Delays, On Going Order and Emergency Supplies 01032 ROND (DE) M., MILLER A.N. The playground of Academe : The Rhetoric and Reality of Tenure and Terror 01033 BIBARD L. Décision et écoute 01034 GEHRKE I., HORVATH P. Implementation of Performance Measurement. A comparative Study of French and German Organizations 01035 NAPPI-CHOULET I. The Recent Emergence Of Real Estate Education in French Business Schools : The Paradox of the French Experience Page 10 2002 Page 11