Query co-processing on graphics processors (GPUs) has become an effective means to improve the performance of main memory databases. However, this co-processing requires the data transfer between the main memory and the GPU memory via a lowbandwidth PCI-E bus. The overhead of such data transfer becomes an important factor, even a bottleneck, for query co-processing performance on the GPU. In this paper, we propose to use compression to alleviate this performance problem. Specifically, we implement nine lightweight compression schemes on the GPU and further study the combinations of these schemes for a better compression ratio. We design a compression planner to find the optimal combination. Our experiments demonstrate that the GPU-based compression and decompression achieved a processing speed up to 45 and 56 GB/s respectively. Using partial decompression, we were able to significantly improve GPU-based query co-processing performance. As a side product, we have integrated our GPUbased compression into MonetDB, an open source column-oriented DBMS, and demonstrated the feasibility of offloading compression and decompression to the GPU. 1.

In recent years, column stores (or C-stores for short) have emerged as a novel approach to deal with read-mostly data warehousing applications. Experimental evidence suggests that, for certain types of queries, the new features of C-stores result in orders of magnitude improvement over traditional relational engines. At the same time, some C-store proponents argue that C-stores are fundamentally different from traditional engines, and therefore their benefits cannot be incorporated into a relational engine short of a complete rewrite. In this paper we challenge this claim and show that many of the benefits of C-stores can indeed be simulated in traditional engines with no changes whatsoever. We then identify some limitations of our “pure-simulation ” approach for the case of more complex queries. Finally, we predict that traditional relational engines will eventually leverage most of the benefits of C-stores natively, as is currently happening in other domains such as XML data. 1.

The database storage hierarchy has been heavily optimized for the performance characteristics of disks. Storage managers typically employ row- or column-oriented storage layouts, or a combination, to improve the I/O performance of different query workloads with disks. The recent rise of flash memory-based solid-state drives (SSDs) significantly change the performance characteristics of storage: these drives provide an order of magnitude lower read/access latencies, significantly higher read bandwidths, and most importantly, negligible seek overheads. In light of these differences, we analyze major storage optimizations for read-optimized databases. We examine the benefits of row and column-oriented storage layouts on flash SSDs. Our measurments span through different workload variations, including selectivity, projectivity and concurrency that affect query processing on flash. Further, we also investigate the cost and benefits of a set of database optimizations, including data compression, prefetching, and indexes on flash SSDs. Our analytical models back our experimental evaluation of the performance tradeoffs of these optimizations. Three of our key findings are: (1) SSDs scale up linearly with concurrent execution of database queries and outperform disks by up to a factor of two, (2) the low seek cost on SSDs makes columnstorage a better choice for laying out data on a variety of flash devices, (3) and that while data compression is useful to further leverage the bandwidth of flash, database prefetching has less benefit for flash storage. Finally, we present a list of design implications of our findings on future database and operating systems for effectively embracing flash storage.

The database storage hierarchy has been heavily optimized for the performance characteristics of disks. Storage managers typically employ row- or column-oriented storage layouts, or a combination, to improve the I/O performance of different query workloads with disks. The recent rise of flash memory-based solid-state drives (SSDs) significantly change the performance characteristics of storage: these drives provide an order of magnitude lower read/access latencies, significantly higher read bandwidths, and most importantly, negligible seek overheads. In light of these differences, we analyze major storage optimizations for read-optimized databases. We examine the benefits of row and column-oriented storage layouts on flash SSDs. Our measurements span through different workload variations, including selectivity, projectivity and concurrency that affect query processing on flash. Further, we also investigate the cost and benefits of a set of database optimizations, including data compression, prefetching, and indexes on flash SSDs. We back our experimental evaluation with analytical models of the performance tradeoffs of these optimizations. Three of our key findings are: (1) SSDs scale up linearly with concurrent execution of database queries and outperform disks by up to a factor of two, (2) the low seek cost on SSDs makes columnstorage a better choice for laying out data on a variety of flash devices, (3) and that while data compression is useful to further leverage the bandwidth of flash, database prefetching has less benefit for flash storage. Finally, we present a list of design implications of our findings on future database and operating systems for effectively embracing flash storage.