FineStream: Fine-Grained Window-Based Stream Processing on CPU-GPU Integrated Architectures

Accelerating SQL queries on stream processing by utilizing heterogeneous coprocessors, such as GPUs, has shown to be an effective approach. Most works show that heterogeneous coprocessors bring significant performance improvement because of their high parallelism and computation capacity. However, the discrete memory architectures with relatively low PCI-e bandwidth and high latency have dragged down the benefits of heterogeneous coprocessors. Recently, hardware vendors propose CPU-GPU integrated architectures that integrate CPU and GPU on the same chip. This integration provides new opportunities for fine-grained cooperation between CPU and GPU for optimizing SQL queries on stream processing. In this paper, we propose a data stream system, called FineStream, for efficient window-based stream processing on integrated architectures. Particularly, FineStream performs fine-grained workload scheduling between CPU and GPU to take advantage of both architectures, and it also provides efficient mechanism for handling dynamic stream queries. Our experimental results show that 1) on integrated architectures, FineStream achieves an average 52% throughput improvement and 36% lower latency over the state-of-the-art stream processing engine; 2) compared to the stream processing engine on the discrete architecture, FineStream on the integrated architecture achieves 10.4x price-throughput ratio, 1.8x energy efficiency, and can enjoy lower latency benefits.