Institutional Repository
Technical University of Crete
EN  |  EL

Search

Browse

My Space

High throughput “on-the-fly” stream data processing on hybrid FPGA-basedplatform

Rousopoulos Christos

Full record


URI: http://purl.tuc.gr/dl/dias/A4E41FAF-038A-46B7-A7C1-0639898AE859
Year 2017
Type of Item Master Thesis
License
Details
Appears in Collections

Summary

With the dawn of social networks and the growth of the Internet of Things market, huge volumes of data are generated every second in various fields, such as banking, agriculture, finance, health care, bioinformatics etc. This data congestion drives the need for systems that offer real-time fast processing of this high-volume and velocity data, while looking for low-cost and power efficient solutions. The stream join is a fundamental and computationally expensive operation for relating information from different data streams, usually applied over a specific time-based window due to the unbounded size of the data streams. This thesis presents an innovative high performance FPGA-based system for accelerating stream join processing. The implemented system achieves at least one order of magnitude better processing throughput in comparison to other state-of-the-art software-based solutions and totally outperforms all lately proposed hardware-based solutions. Finally, the proposed hardware- based architecture is fully configurable and can be used as generic template for mapping stream processing algorithms on reconfigurable logic, taking into consideration real-world challenges and restrictions.

Available Files

Services

Statistics