Geospatial Publish/Subscribe System for Efficient Sensor Data Stream Filtering

Cilj ovog rada je proizvesti učinkoviti geoprostorni sustav objavi-pretplati te provesti evaluaciju najučinkovitijeg načina filtriranja za kreirani sustav. Arhitektura objavi-pretplati sastoji se od skupa objavitelja koji šalju poruke sustavu bez poznavanja primatelja i skupa pretplatnika koji se pretplaćuju na određene kategorije i sadržaj objava bez poznavanja objavitelja. U okviru rada objavitelji su senzori koji izvode mjerenja za određeno geografsko područje. Mjerenja koja senzori izvode mogu biti izražena kroz tipove podataka i mjerne jedinice koji su proširivi u sustavu. Pretplatnici se pretplaćuju na interesno geografsko područje te mogu postavljati dodatni uvjet nad izmjerenom vrijednosti objave. Prilikom dolaska nove objave potrebno je profiltrirati pretplatnike s obzirom na definirane uvjete preferiranog geografskog područja i dodatne uvjete nad mjernom vrijednosti. Kako bi se filtriranje učinkovito provelo koriste se razne strategije indeksiranja koje se na kraju rada evaluiraju i donosi se odluka o optimalnom procesu filtriranja za dani slučaj.The goal of this paper is to implement efficient geospatial publish/subscribe system and evaluate which is the most effective filtering method in the implemented system. In publish/subscribe architecture there is a set of publishers who send messages to system without knowing who the recipients of the messages are and a set of subscribers who subscribe to categories and certain message content without knowing who the publishers are. In the system which was developed as part of the paper, publishers are sensors which measure data for certain geographic area. Sensors can perform measuring in custom data types and measuring units which can be added to the system. Subscribers subscribe to a certain geographic area and can add additional conditions on measured value inside the publication. When the new publication arrives, it is necessary to filter subscribers depending on their preferred geographic area as well as additional conditions on measured value. In order to perform filtering as efficiently as possible, indexing strategies were used and evaluated to decide which is the optimal filtering process for the implemented publish/subscribe system