Development of a data stream management system using special hardware assistance

2016 Fiscal Year Final Research Report

• Project/Area Number: 26330061
• Research Category: Grant-in-Aid for Scientific Research (C)
• Allocation Type: Multi-year Fund
• Section: 一般
• Research Field: Computer system
• Research Institution: The University of Electro-Communications
• Principal Investigator: Yoshinaga Tsutomu 電気通信大学, 大学院情報理工学研究科, 教授 (60210738)
• Project Period (FY): 2014-04-01 – 2017-03-31
• Keywords: データストリーム処理 / リコンフィギャラブル計算 / FPGA / 専用ハードウェア / 結合演算 / 集約演算 / スライディングウィンドウ

Outline of Final Research Achievements

We designed special hardware modules which accelerate aggregation and join operations for stream data process in big data analytics. We also developed a query compiler which generates configuration information for FPGAs from SQL-based query plans as well as distributed system software to utilize a PC cluster with interconnected FPGAs.
We have obtained approximately 9.2 times faster join throughput and 58.8 times performance per energy, compared to conventional software-based join operation, on an FPGA-enhanced 4-node PC cluster when a sliding window size is 4096 tuples per node.

Free Research Field

リコンフィギャラブルシステム