Searching and analyzing on-line high volume measurement streams
|Uppsala universitet - Institutionen för informationsteknologi
|Funding from Vinnova
|SEK 9 840 000
|January 2008 - December 2011
Purpose and goal
A prototype of a data stream management system is developed for searching among streaming measurement data from built in units in engineering systems such as vehicles, engines, instruments, and tools. The DSMS is scalable to handle very large data volumes. The architecture is distributed with automatically parallelized operations. The system will support domain dependent filtering and data mining from streams.
Results and expected effects
The technology is expected to improve efficiency and quality in Swedish engineering, mining, and vehicle industries. It enables better quality of delivered products with lower maintenance costs. Novel data collection services can be established. It will enable products to be reconstructed and material in hardware reused.
Approach and implementation
First a pre study is made on state-of-the-art in DSMSs and wireless communication systems. Based on the requirements the functionality of the DSMS is designed. A prototype system is designed based on possible design alternatives and existing system components and research systems. Domain specific search and mining operators are designed based on the industrial needs. To evaluate performance a standard benchmark for DSMS is used. A new benchmark is developed based on simulated industrial data that include chosen search and mining operators. The prototype is implemented and evaluated in academic environment using the benchmarks. Based on the result from the analyses an improved prototype is developed.