Automatizacija fizičkog projektovanja skladišta podataka proširenog Data Vault pristupa

Abstract

Skladište podataka predstavlja podlogu za implementaciju sistema poslovne inteligencije, integraciju podataka i podršku poslovnom odlučivanju. Za skladišta podataka uglavnom se izgrađuje dimenzioni model kojim se dobija bolja mogućnost vizualizacije podataka. Arhitektura skladišta podataka podrazumijeva definicije osnovnih elemenata skladišta podataka i opisuje kako su ti elementi izgrađeni i povezani. Kod arhitekture skladišta podataka danas postoje dva dominantna pristupa. Jedan pristup, poznat je pod nazivom CIF (Corporate Information Factory) arhitektura, zastupa Bill Inmon koji smatra da skladište podataka treba da bude normalizovano i da sadrži atomske podatke u trećoj normalnoj formi. Drugi pristup, poznat poda nazivom BUS arhitektura, zastupa Ralph Kimbal koji tvrdi da i samo skladište treba da ima višedimenzionalnu strukturu. Data Vault koncept skladišta podataka koji je osmislio Daniel Linstedt, predstavlja detaljno orijentisani, istorijski praćen i jedinstveno povezani skup normalizovanih tabela koji podržavaju jedan ili više funkcionalnih područja poslovanja. To je hibridni pristup koji obuhvata najbolje osobine između 3. normalne forme (3NF) i zvijezda šeme. Data Vault koncept naglašava potrebu da se ostavi trag odakle i kad su podaci došli u bazu podataka, te omogućuje lake promjene strukture skladišta podataka u skladu sa promjenama u poslovnom okruženju. Skladišta podataka se izrađuju različitim metodološkim pristupima. Međutim, problem je što proces dizajniranja skladišta podataka i izrade dimenzionog modela nije automatizovan. U disertaciji je sagledan problem dizajniranja skladišta podataka na osnovu šire literature, te istražena mogućnost automatizacije fizičkog projektovanja skladišta podataka baziranog na Data Vault konceptu uz minimalnu interakciju sa korisnikom...

Data warehouse is a basis for the implementation of business intelligence systems, data integration and decision support. Usually, a dimensional model is built for data warehouse in order to get better data visualization. Data warehouse architecture represents definitions of basic data warehouse elements and describes the connection and structure of the elements. There are two dominant approaches in data warehouse architecture. First one is known as CIF (Corporate Information Factory) architecture and it is represented by Bill Inmon who believes that data warehouse needs to be normalized and needs to contain atomic data in third normal form. Second approach, known as BUS architecture, is represented by Ralph Kimball who believes that warehouse should also have multi-dimensional structure. Data Vault concept of data warehouse which was designed by Daniel Linstedt and represents detail oriented, historically followed and uniquely connected group of normalized tables which support one or more functional areas of a business. That is a hybrid approach which covers the best features from third normal form (3NF) and star scheme. Data Vault concept emphasizes the need to leave a trace from where and when data was sent into database, and also enables easy changes in data warehouse structure in accordance with business environment changes. Data warehouses are built in different methodological approaches. However, there is a problem with the fact that the process of data warehouse design and related dimensional model is not automatized. In this dissertation a problem of data warehouse design process was accessed, based on broad analysis of literature, and a possibility of automation of physical data warehouse design based on Data Vault concept, with a minimal interaction with users, researched. The work presents simple algorithm for automation of data warehouse physical design, based on Data Vault concept, using data model with metadata transactional data sources and rules...