The Integrated Approach to ERP: Embracing the Web

1 he ntegrated pproach to P: mbracing the Web ergey V. Zykov nternational Group of Companies oscow, ussia bstract ntegrated approach to enterprise resource planning (P) software design and implementation can significantly improve the entire corporate information infrastructure and it helps to benefit from power of nternet services. he approach proposed provides for corporate Web portal integrity, consistency, urgency and front-end data processing. Human resources (H) P component implementation is discussed as an essential instance. 1. ntroduction Frequent priority changes in conglomerate development require fast and flexible adaptability of management to match rapidly changing market conditions. uch adaptability should be based on strategic software integration and its connection to nternet, especially for comprehensive P. uring the two recent decades, the data models () and architectures underlying software development process have been changed significantly to support object met hodologies and interoperability. ince early 90 s, rather primitive file-server systems supporting relational [3] have evolved into comprehensive P systems based on extendable relational supporting objectoriented B. ttempts of enterprise application integration have also been undertaken [2,4-6]. Permission to copy without fee all or part of this material is granted provided that the copies are not made or distributed for direct commercial advantage, the C copyright notice and the title of the publication and its date appear, and notice is given that copying is by permission of the nstitute for Contemporary ducation JUC. o copy otherwise, or to republish, requires a fee and/or special permission from the JUC. Proceedings of the 4 th nternational Workshop on Computer cience and nformation echnologies C 2002 Patras, Greece, 2002 he main objectives of the paper are development of integrated data and metadata model, application of the model for extending corporate nternet presence, formal approach to Web-enhanced P construction and overview of improved P implementation process. esearch methods meeting the problem domain specific features are based on a creative synthesis of fundamental statements of lambda calculus [11], categories [1] and semantic networks [9]. he data model introduced provides an open, integrated, problem-oriented, event-driven data and metadata management of dynamic, heterogeneous, weak-structured problem domains in a more adequate way than previously known ones. he model suggested allows to generate system architecture solutions for open, distributed, interoperable environments supporting front-end, multipurpose data warehousing and nternet-publishing on the basis of CB, UL and business-process reengineering (BP) technologies. 2. elated Works Papers [1,3,9-12] provide rigorous mathematics foundation and solid theoretical background for notations used. elational B and weak-structured document solutions are cross-examined. Papers [13,14] provide a mathematically rigorous overview of object-oriented systems development and describe a number of promising implementations. P generation history is discussed in [8,15]. 3. rchitecture and nterface equirements pecific features of the problem domain require support of dynamic multi-level personnel restructuring process with multi-alternative assignment-based comprehensive estimation of enterprise activity. nterface requirements set should allow dynamic variation of mandatory input fields, flexible access rights differentiation and noninterruptible data integrity support. n respect of architecture, the system should provide interoperability, expandability, flexible adjustment to problem domain Workshop on Computer cience and nformation echnologies C 2002, Patras, Greece, 2002

2 changes and easy data/metadata correction (including rollback mechanisms). 4. he ntegrated ata and etadata odel 4.1 he ata bject odel athematical formalisms existing for problem domains are not fully adequate to dynamics and statics semantics. oreover, current methods of C-and- enterprise applications development do not result in solutions of a wide application range; commercial P implementations do not provide sufficient flexibility of heterogeneous data handling. ccording to research results of personnel management problem domain, a computational based on object calculus has been built. he model is an innovative synthesis of finite sequence, category and semantic network theories. ata objects () of the can be represented as follows: = < concept, individual, state >, where a concept is understood as a collection of functions with the same definition area and the same value range. n individual implies an essence selected by a problem domain expert, who indicates the identi fying properties. tate changes simulate dynamics of problem domain individuals. Compared to research results known as yet, the suggested features more adequate dynamics mapping for heterogeneous problem domains. he also benefits better support for problem-oriented integrated data management. n architecture and interface aspects the provides straightforward iterative design of open, distributed, interoperable H based on UL and BP methodologies. s far as implementation part is concerned, multi-repository information processing of heterogeneous problem domains is supported. hus, frontend data access is provided which is based on eventdriven procedures and dynamic QL technologies. he computational model suggested is based on the twolevel conceptualization scheme [13]. Conceptualization implies a process of establishing relationship between problem domain concepts. ndividuals h, according to the assigned types, are united in assignment-dependent collections, thus forming variables of sor t H () = {h h : }. his formalism is used to simulate problem domain dynamics. When fixing data model individuals, uniqueness of individualization of data object d from problem domain by means of the formula Φ is required: x Φ (x) i = d {d} = {d Φ(d) i = 1}. 4.2 he etadata bject odel Let us introduce a compression principle for the computational data object model C = y: [] x : (y(x) Φ ) = {x : Φ} that allows to apply the model to concepts, individuals and states separately, as well as to data objects as a whole. he suggested computational metadata model expands traditional -model [3] by adding the following compression principle: x j+1 z j+1 : [ [] ] x j : [ [] ] (z j+1 (x j ) Φ j ), where z j+1, x j+1 metadata predicate characters in relation to level j, x j individual and Φ j definition language construction of level j. he integrated model for objects of data, metadata and states is characterized by scalability, expandability, metadata encapsulation and transparent visualiza tion. xpandability, adequacy, neutrality and semantic correctness of the formalism introduced provide problemoriented software design with adequacy maintenance at every stage of the implementation process. emantics of computational model of objects of data, metadata and states can be adequately and uniformly formalized by means of typed λ-calculus, combinatory logic, and semantic network-based scenario description. 4.3 odel pplication for Web ite nhancement nhancing ntegrated P with ata View epresentation odel Let us consider the following parameters of nternet user appearance and behavior: data access rights, personal preferences (fonts, color settings, etc.), Web browser settings (links, cache, history, etc.) and data access device (Web V, P, mobile phone, terminal, etc.) profile. Let us assume that and B are sets. Let В А stand for the mapping from to B: В А = {f f : B}. Let us match В А with the evaluation function: = {f f B B}. hus, = (<f, x>) = f(x), so <f, x&g t; = f(x). ow let us build the semantics network language model. Let us consider an ordered pair of of the form L=<,C>, where ={ 1, 2, } is a dyadic predicate symbols set and C={C 1,C 2, } is a set of constants. herewith, the atomic formulae of the model suggested correspond to simple frames, and terms denote problem domain individuals. Let us construct a frame evaluation he ntegrated pproach to P: mbracing the Web

3 procedure using the introduced evaluation function. ow let us consider an example of user profile evaluation procedure based on the suggested data model. Let the F functional denote the most general class of users. Let the assignment s={high resolution graphics, multimedia} account for user customary settings. Let F(s) stand for the set of users, for whom the customary settings are restricted to high resolution graphics and multimedia. Let the assignment p={registered, unregistered, corporate} account for user registration status. Let F(s)(p) designate the set of users with high graphics and multimedia preferences for whom a registration status is assigned, i.e., those who have already visited the Web site. For the sake of simplicity and without loss of generality, let us consider that in general the site visitors set referred above as functional F, is dependent upon browser settings (v), data access device type (e), personal preferences and access rights: F=F((v), (e), ). n thi s case, the formula F=F((v), (e), ) indicates a formal procedure that evaluates parameterized functional, the expression F=F((v), (e), ) (s) evaluates users with given customary settings (s), and the formal procedure F=F((v), (e), ) (s) (p) evaluates users with given customary settings (s) and registration status (p). he introduced functional F can be considered an illustration of computational formalism for parameterized procedure of profile evaluation of certain visitor categories. Let us demonstrate that two-level conceptualization scheme is sufficient for the model adequacy. Let us introduce the following denotations: r = {r c.s., r r.s. } specific costs; z = {z c.s., z r.s. } segmentation degree (i.e., possibility of splitting users into stable and independent groups); q i =q i overheads; l i =l i duration of the request processing stage (download, dynamic form or report creation, etc.); n i =n i number of request processing stages. valuated values are generalized, i.e., there is no uniqueness of value choice for specific costs and segmentation degree. Generalization level decrease is achieved by considering an assignment point s: z (higraph) = z higraph, z ( s ) = z (mmedia) = z mmedia ; r (higraph) = r higraph, r ( s ) = r (mmedia) = r mmedia. oreover, further generalization level decrease by introducing the second assignment p does not succeed: z ( s )( p ) = z ( s ); r ( s )( p ) = r ( s ). he result obtained can be explained by the fact that the evaluation procedure involves visitor position in the structure. However, it is obvious that overheads q i are dependent both on customary settings functions and on registration status, i.e. we should let q i = {q i higraph, q i mmedia }. he equality q i = q i implies that q i higraph. = q i mmedia = q i. 5. he ntegrated P mplementation 5.1 Customizing the mplementation cheme uring design process, P specification is transformed from problem domain concepts to data model entities, then, further, to B scheme (w ith PL\QL as manipulation language), and, finally, to target P description with required architecture and interface (see fig. 1). s a result of problem domain analysis, computational and generalized scheme of P development [16] have been customized to satisfy the required problem domain conditions. Web applications implementation is finalizing P implementation stage. racle nternet eveloper uite serves a gateway between data warehouse and corporate nternet site. When content-critical warehouse updates occur, content is automatically updated accordingly. Periodical and manual data updates and retrievals are also supported options. 5.2 Problem-riented nterface and vent- riven rchitecture ccording to the detailed P design sequence, a comprehensive heterogeneous repository processing scheme is introduced that allows users to interact with distributed database in a certain state depending on dynamically activated (i.e., assigned) scripts. hus, scripts (in a form of data access profiles an d stored objectoriented program language procedures) are initiated depending on user-triggered events. cripts provide transparent and intellectual client/server front-end user-todatabase connection. ynamically adjustable database access profiles provide high fault tolerance and data security both for ordinary and privileged system users in heterogeneous environment. he profiles are implemented using CB technology as an intellectual media between end-user and heterogeneous data warehouses. epending on semantics-oriented user profile structure, Workshop on Computer cience and nformation echnologies C 2002, Patras, Greece, 2002

4 certain database connection and access level profiles are dynamically assigned. he profiles are valid only until the end of data exchange session. ccording to the hierarchy, users access data under one of the basic scenario profiles. ccess is granted not only to data, but also to metadata (i.e., data object dimensions, integrity constraints, access rights, browser parameters, user preferences, etc.). dministrative users have extended access to metadata. hus, under the model introduced, data and metadata objects are manipulated uniformly. his makes system interface a problem-oriented, straightforward and uniform one and significantly increases system performance. nternet-published warehouse data processing scheme under conditions of event-driven architecture is presented in fig.2. Client-side web page object states can change depending on event script execution. hough the warehouse data remains unchanged, user can request data update, or produce a query. oreover, front-end interf ace itself is also client profile dependent. ptions include personal preferences (color schemes, screen resolution etc.), data access device and Web browser settings. he essential benefit of event-orientation for the P global network extension is that corporate users get additional access to ntranet resources of corporate nternet site. egistered (and/or extranet) users access some extra data compared to ordinary ones. Warehouse data access is also dependent on user profile. t the upper level of data access hierarchy, clients can be divided into administrators, managers and ordinary users (see fig.3). Judging by the profile, data and metadata object states (i.e., system interface) are changed. For instance, a web designer rights assignment provides full access to interface elements database, while content manager can get full access to another warehouse instantiation by means of a different interface instance. cenario-based end-user interface results in higher degree of interactivity , user-friendliness and security. User profiles (i.e. assignments) can be stored in metadata base of visitors, and, depending on their properties, data access and representation level could be customized. ets of Web pages accessible and their content are dependent on user profile. Client-side profile also accounts for preferable information layout depending both on personal preferences and on the device type for which the data is customized. User profile and preferences are also vital for client analysis, and resulting statistic reports serve a foundation for performance and interface optimization. 5.3 mplementation escription he introduced methodology has been practically approved during H P system component improvement at nternational Group of Figure 1 Generalized P implementation chart racle Financials racle Human esources racle nterffice ata Center WB VLP nnual eport Web erver Corporate Profile Contact nfo Vacancy List PL- -G P Web V obile Phone emote PC L Figure 2 vent-driven W eb-wired P data publishing and access scheme he ntegrated pproach to P: mbracing the Web

5 Companies. ll of the H component modules are cemented together by a unified interface and integrated into P environment of racle pplications family of financial, commodity and document management systems. From the system architecture viewpoint, the integrated H component provides certain level of data input, correction, analysis and output depending on front-end position (i.e., assignment) in user hierarchy. nteractive interface is represented by problem-oriented form designer, report generator, on-line documentation and administration tools. he enhanced P system database supports the integrated storage for data (i.e., information for on-line users) and metadata (i.e., data object dimensions, integrity constraints and business process parameters). uring the enhanced P design process, problem domain specification (in a semantic network representation) is transformed into use-case UL diagrams, then, by means of racle eveloper/2000 and integrated C-tool, - into -scheme and, finally , into the attributes of target databases. racle nternet eveloper uite has been used to transform the P into a Web-wired application. n the basis of the information model developed, an architecture and interface solution for integrated H management software has been designed. o prove adequacy of the model developed and component integration algorithm suggested, software prototype has been designed. o provide required levels of industrial scalability and fault tolerance, judging by the results of exhaustive C-and- software analysis, racle eveloper/2000 toolkit has been chosen as a solution supporting UL and BP methodologies. ccording to specification requirements developed by the author, implementation had been significantly improved by extracting essential information from integrated P and publishing it in the corporate nternet site. H information subsystem provides a number of significant data items for corporation profile Web page including total establishment, number of countries and companies that represent the corporation. ince the integrated solution implemented includes vacancy module as a part of its H subsystem, dynamically updated vacancy data nternet page can be easily produced. imilarly, financial components could provide data for a number of periodical or privileged user-triggered financial reports. ata examples include revenues, profits, production dynamics, stock values, etc. ame as with H subsystem, dynamics tracking is important. Corporation development plans based on deferred charges could be published. Production manufacturing module can provide productivity and capacity data for executive summaries and company profile Web and ntranet pages. ddress book from document control subsystem serves for contact information and provides automatic feedback routing G U ystem dministrator ecurity dministrator Content anager Web-designer Webmaster egistered User ntranet User xtranet User Unregistered User Workshop on Computer cience and nformation echnol ogies C 2002, Patras, Greece, 2002 U F F C H F G udit Journal ccess ights B Password B Content B nterface B B nstance B nstance Financials aterials Projects Figure 3 Front-end interface and data access profiling G V

6 through corporate organizational structure. For further enhancement of corporate Web site performance and interface, data published in the Web page is dynamically updated by an event-driven software agent. he integrated P data warehouse (fig.3) is stored in the data center. Group warehouse is based on racle Financials, Human esources and nterffice. he full-scale implementation is based on hardware platform of an B /6000 two-server high availability cluster running under B X operating system. he system has passed a three-year test in a large corporation. 6. esults and Conclusion computational data model has been introduced that provides integrated manipulation of data and metadata objects, especially in rapidly changing heterogeneous problem domains. he model is an alloy of methods of finite sequences, categories and semantic networks. n the basis of the formal model, an original and comprehensive iterative scheme for Web-integrated P design and implementation has been proposed. he scheme includes an algorithm for new component integration into existing P environment. he integration algorithm provides adequacy, consistency and data integrity; algorithm details are presented in [16]. ccording to the approach suggested, a comprehensive Web-enabled P interface has been designed. he P interface is based on an open and extendable architecture. s a first step towards implementing the enterprise resource management solution, a fast event-driven software prototype has been developed on the basis of the designed UL-based interface and architecture scheme. Using the prototype testing results, a full-scale objectoriented P application has been designed. he fullscale enterprise-level software has been customized for corporate resource management and implemented at an enterprise with more than 1000 employees. Web-integrated P solution have proved significant decrease in time and costs of implementation. ther major benefits include growth of portability, expandability, s calability and ergonomics levels in comparison with existing commercial software of the kind. terative multilevel software design scheme is based on formal model unifying object-oriented methods of data (data objects) and knowledge (metadata objects) management. ndustrial implementation of the nternet-embracing P H component has been carried out using integrated Cand- and Web publishing toolkits. xperience of implementation support has proved importance, urgency, originality and efficiency of the approach suggested. heoretical and practical statements outlined in the paper have been approved by successful implementation of the Web-wired enterprise-level P software at nternational Group of Companies. he author is going to continue research in order to turn the enhanced P into an integrated corporate ntranet and e-commerce solution. eferences 1. K.Baclawski,.imovici, W.White. " categorical approach to database semantics". athematical tructures in Computer ci., vol.4, p.p , Calvanese, G.Giacomo,.Lenzerini,.ardi,.osati. "ource ntegration in ata Warehousing". X Workshop 1998, p.p F.Codd. "elational Completeness of ata Base ublanguages ata Base ystems". n: ustin. ds.,.- ew York; Prentice Hall, 1972 (Courant Computer cience ymposia eries o.6) 4..Florescu,.Y.Levy. "ecent Progress in ata ntegration - utorial". B 1998, p.p Y.Kambayashi. "esearch and evelopment of dvanced atabase ystems for ntegration of edia and User nvironments". F 1999, p.p Linthicum. "nterprise pplication ntegration". ddison Wesley Longman. B ov p. 7. nformation about racle software mentioned in this paper available at WWW: http: // 8..rfali,.Harkey, J.dwards. "he ssential Client/erver urvival Guide", 2 nd dition. Wiley Computer Publishing, 1996, 678p. 9...oussopulos. " semantic network model of data bases", oronto Univ., cott. he lattice of flow diagrams, n: ymposium on emantics of lgorithmic Languages.- Berlin, Heidelberg, ew York: pringer-verlag, 1971, p.p (Lecture otes in athematics o. 188) 11... cott. "elating theories of the λ-calculus". n J.Hinhley and J.eldin, eds., o H.B.Curry: ssays on combinatory logic, lambda calculus and formalism, p.p cademic Press, cott. "omains for denotational semantics". n.ielsen and..chmidt, eds., Lecture otes in Computer cience, vol.140, p.p , arhus, enmark, July pringer 13. V..Wolfengagen. "bject-oriented solutions. B 1996, p.p pringer 14. V..Wolfengagen. "vent riven bjects". C'99. oscow, ussia, 1999 p.p V.Zykov. "ntegrated Human esources nformation ystems: nvolving xtra ata ources Centered around Groupware". C 99. oscow: Ph, 1999, p.p V.Zykov. "nterprise esource Planning ystems: the ntegrated pproach". C 2001, Ufa, ussia. ept.21-26, 2001 he ntegrated pproach to P: mbracing the Web

Source: The Integrated Approach to ERP: Embracing the Web