ADS+ builds a minimal tree during the indexing phase. The tree contains only iSAX representations, which are sufficient to build the index tree. The actual data series are only necessary during query time, i.e., in order to give a correct answer. In addition, not all data series are needed to answer a particular set of queries. In this way, ADS+ first creates all necessary iSAX representations and builds the index tree without inserting any data series.

When a query arrives (in the form of a data series), it is first converted to an iSAX representation. The index tree is traversed searching for a leaf with an iSAX representation similar to that of the query. Then, this leaf it is split until it becomes small enough. In the case that the leaf node where the search finally ends contains only iSAX representations but not any data series, then all missing data series are fetched from the raw file.

ADS+ achieves both faster indexing times and faster query answering times compared to the state-of-the-art iSAX 2.0 index.

The overall 3-tiered architecture of the RINSE demonstration system can be seen in the Figure below. It is composed of a server, a web interface and a middleware that connects the web interface to the server. The RINSE web interface is developed as a single page application in HTML5, JavaScript and CSS. This allows for an intuitive and responsive experience where users can draw queries and watch the results appear on screen in near real-time.

RINSE allows users to index data and issue nearest neighbor queries. Three different access methods have been implemented. A simple serial scan, which reads the complete raw data file for every query; also employing a simple early abandoning technique for avoiding useless computations. A complete iSAX 2.0 index, which is built before the users can start posing queries, and an adaptive ADS+ index, which takes considerably less time to construct. In all the three cases, the distance measure used is the Euclidean Distance. Users can issue queries using a web based graphical query definition tool (they draw data series on screen using the mouse). They can also generate random queries, or choose queries from a list. Each query can be re-run using a different data structure such that users can see the differences. Furthermore, users have access to various statistics that measure the performance of each access method, the amount of data currently ingested by the adaptive index and various memory footprint information.