Towards an editable live ETL system for the eventwarehouse

Nowadays we are witnesses of the proliferations of different sensor devices that can be connected to the network and are able to produce heterogeneous data (textual, visual, audio, and other rich multimedia formats). They can be profitable used for detecting, handling and advising people of the verification of emergency events such as disasters due to natural phenomena (like flooding, storming, extreme temperatures). Beside the physical sensors, able to detect data about physical phenomena (like temperature, humidity, wind, rain), social sensors able to collect data from people (like twitter data, traffic information) are becoming more and more relevant [5]. The identification of real word events by combining data gathered from these sensors is a main research challenge [1]. The EventWarehouse (EvWH) [2] has been designed to cope with emerging requirements of understanding and managing big and heterogeneous data in order to bring to human not only information but also knowledge. Differently from traditional data-warehouse that control only data, EvWH is designed to manage complex events and presents three main internal components: Unique-Data-Format Standardization (UDFS), Warehouse (WH), and EventManagement (EM). The EM is in charge of mining information and knowledge from things stored in WH (e.g. correlation discovery, data mining, or event detection). Moreover, the EM can connect to external applications/systems to export things stored in WH (both in streaming and non-streaming modes), or import event models to detect complex events. An external component -- User-Defined-Sensor (UDS) – has been developed, by which users can create a new sensor that can crawl data in cyber world and stream them to EvWH.