Title: GPU Acceleration of SQL Analytics on Compressed Data Show Chinese title

Title: GPU加速的压缩数据SQL分析

Abstract: GPUs are uniquely suited to accelerate (SQL) analytics workloads thanks to their massive compute parallelism and High Bandwidth Memory (HBM) -- when datasets fit in the GPU HBM, performance is unparalleled. Unfortunately, GPU HBMs remain typically small when compared with lower-bandwidth CPU main memory. Besides brute-force scaling across many GPUs, current solutions to accelerate queries on large datasets include leveraging data partitioning and loading smaller data batches in GPU HBM, and hybrid execution with a connected device (e.g., CPUs). Unfortunately, these approaches are exposed to the limitations of lower main memory and host-to-device interconnect bandwidths, introduce additional I/O overheads, or incur higher costs. This is a substantial problem when trying to scale adoption of GPUs on larger datasets. Data compression can alleviate this bottleneck, but to avoid paying for costly decompression/decoding, an ideal solution must include computation primitives to operate directly on data in compressed form. This is the focus of our paper: a set of new methods for running queries directly on light-weight compressed data using schemes such as Run-Length Encoding (RLE), index encoding, bit-width reductions, and dictionary encoding. Our novelty includes operating on multiple RLE columns without decompression, handling heterogeneous column encodings, and leveraging PyTorch tensor operations for portability across devices. Experimental evaluations show speedups of an order of magnitude compared to state-of-the-art commercial CPU-only analytics systems, for real-world queries on a production dataset that would not fit into GPU memory uncompressed. This work paves the road for GPU adoption in a much broader set of use cases, and it is complementary to most other scale-out or fallback mechanisms. Show Chinese abstract

Abstract: GPU由于其巨大的计算并行性和高带宽内存（HBM）而特别适合加速（SQL）分析工作负载——当数据集适合GPU HBM时，性能是无与伦比的。 不幸的是，与低带宽的CPU主存相比，GPU HBM通常仍然较小。 除了在多个GPU上进行暴力扩展外，当前加速大型数据集查询的解决方案包括利用数据分区和将较小的数据批次加载到GPU HBM中，以及与连接设备（例如CPU）的混合执行。 不幸的是，这些方法受到较低主存和主机到设备互连带宽的限制，会引入额外的I/O开销或增加更高的成本。 当试图在更大的数据集上扩展GPU的采用时，这是一个重大问题。 数据压缩可以缓解这一瓶颈，但为了避免支付昂贵的解压/解码费用，理想解决方案必须包括可以直接在压缩数据上运行的计算原语。 这就是我们论文的重点：一种新的方法集合，使用如行程长度编码（RLE）、索引编码、位宽减少和字典编码等方案，在轻量级压缩数据上直接运行查询。 我们的创新包括在不进行解压缩的情况下操作多个RLE列，处理异构列编码，并利用PyTorch张量操作实现跨设备的可移植性。 实验评估显示，与最先进的商业CPU-only分析系统相比，对于无法放入GPU内存的生产数据集上的实际查询，速度提升了数量级。 这项工作为在更广泛的应用场景中采用GPU铺平了道路，并且与其他大多数扩展或回退机制是互补的。