COMPARISON OF DATA WAREHOUSING ARCHITECTURES

V raziskavi smo primerjali arhitekture, ki so primerne za podatkovno skladiščenje operaterja mobilne telefonije. Upoštevali smo realne količine podatkov in njihovo vsebino. Izdelali smo tri primerljive prototipne rešitve na treh arhitekturah podatkovnega skladiščenja: zvezdna shema, snežinkasta shema in OLAP-kocka. Vse prototipne rešitve smo implementirali na istem podatkovnem strežniku, ki je uporabljal sistem za upravljanje baze podatkov Oracle 10g. Prototipne rešitve smo kvantitativno primerjali z merjenjem časov izvajanja poizvedb. Med primerjalnim testiranjem prototipnih rešitev smo z analizo izvajalnih načrtov in interpretacijo doseženih časov izvajanj poizvedb prikazali prednosti in slabosti izbora posamezne arhitekture podatkovnega skladiščenja. Raziskava je osvetlila problematiko izbora arhitekture podatkovnega skladiščenja in je uporabna v konkretnem primeru operaterja mobilne telefonije pri odločanju o razvoju informacijskega sistema. Z vidika odzivnosti sistema in možnosti širitve je eksperiment pokazal, da sta za obravnavano problematiko bolj primerni arhitekturi OLAP-kocke in relacijske zvezdne sheme ter manj snežinkasta shema.This research compared data warehousing architectures suitable for operators of mobile telecommunications. Tests included near real volume and content of source data from actual business environment. We built three comparable prototype solutions on three basic data warehousing architectures: a star schema, a snowflake schema and an OLAP cube. All prototype solutions were built on the same data server with underlying Oracle 10g Relational Database Management System. Prototype solutions were compared by quantitative measurement of query execution times for particular architectures. During comparative testing of prototype solutions, we demonstrated the benefits and weaknesses of particular data warehousing architectures by analyzing query execution times and interpreting achieved query execution times. The research gave us deep insights into the problems of tested data warehousing architectures. The research findings can be used in decision-making on future development of information systems. Results of the experiments indicate that the system performance and growth possibilities of the OLAP cube and relational star schema are more acceptable. The relational snowflake schema gave less suitable results