Resource Allocation for High-Volume Streaming Data in Data Centers

点击翻译按钮获取中文摘要

Almost all chip vendors are producing new hardware accelerators by combining several units into a single main-board, and therefore making the execution of parallel and distributed run-time primitives not efficient/scalable. This project aims to develop innovative ways to building incremental and iterative computations over massive data sets in a cluster of heterogeneous systems. This will provide a significant reduction of performance bottlenecks when running heavily distributed data-driven applications. Expected outcomes will include resource management algorithms that optimise performance at large scale. The project will benefit many areas, including running stateful iterative stream-based data-analysis applications in data centres.