GStream: A General Purpose Data-Streaming Framework on GPU Clusters


Project summary

Emerging accelerating architectures, such as GPUs, have proved successful in providing significant performance gains to various application domains. However, their viability to operate on general streaming data is still unknown. We propose GStream, a general-purpose, scalable data-streaming framework on GPUs. The objectives of GStream are as follows: (1) To provide powerful, yet concise language abstractions suitable to describe conventional algorithms as streaming problems. (2) To project these abstractions onto GPUs to fully exploit their inherent massive data-parallelism. (3) To show the viability of streaming on accelerators in experiments to assess flexibility, programmability and performance gains for various benchmarks from a variety of domains, including but not limited to data streaming, data parallel problems, numerical codes and text search.

The proposed work will benefit stream-based and high-end data-intensive computing for GPUs, specifically in the area of massively data-parallel processing to support scalability, and to adapt to changing environments.

Participants:

Publications: