String processing tasks are common in analytical queries powering business intelligence. Besides substring matching, provided in SQL by the like operator, popular DBMSs also support regular expressions as selective filters. Substring matching can be optimized by using specialized SIMD instructions on mainstream CPUs, reaching the performance of numeric column scans. However, generic regular expressions are harder to evaluate, being dependent on both the DFA size and the irregularity of the input. Here, we optimize matching string columns against regular expressions using SIMD-vectorized code. Our approach avoids accessing the strings in lockstep without branching, to exploit cases when some strings are accepted or rejected early by looking at the first few characters. On common string lengths, our implementation is up to 2X faster than scalar code on a mainstream CPU and up to 5X faster on the Xeon Phi co-processor, improving regular expression support in DBMSs.

Today's exponentially increasing data volumes and the high cost of storage make compression essential for the Big Data industry. Although research has concentrated on efficient compression, fast decompression is critical for analytics queries that repeatedly read compressed data. While decompression can be parallelized somewhat by assigning each data block to a different process, break-through speed-ups require exploiting the massive parallelism of modern multi-core processors and GPUs for data decompression within a block. We propose two new techniques to increase the degree of parallelism during decompression. The first technique exploits the massive parallelism of GPU and SIMD architectures. The second sacrifices some compression efficiency to eliminate data dependencies that limit parallelism during decompression. We evaluate these techniques on the decompressor of the DEFLATE scheme, called Inflate, which is based on LZ77 compression and Huffman encoding. We achieve a 2× speed-up in a head-to-head comparison with several multi core CPU-based libraries, while achieving a 17% energy saving with comparable compression ratios.

Implementations of data processing operators on GPU processors have achieved significant performance improvements over their multicore CPU counterparts. To achieve maximum performance, database operator implementations must take into consideration special features of GPU architectures. A crucial difference is that the unit of execution is a group ("warp") of threads, 32 threads in our target architecture, as opposed to a single thread for CPUs. In the presence of branches, threads in a warp have to follow the same execution path; if some threads diverge then different paths are serialized. Additionally, similarly to CPUs, branches degrade the efficiency of instruction scheduling. Here, we study conjunctive selection queries where branching hurts performance. We compute the optimal execution plan for a conjunctive query, taking branch penalties into account and consider both single-kernel and multi-kernel plans. Our evaluation suggests that divergence affects performance significantly and that our techniques reduce resource underutilization and improve operator performance.

Implementations of database operators on GPU processors have shown dramatic performance improvement compared to multicore-CPU implementations. GPU threads can cooperate using shared memory, which is organized in interleaved banks and is fast only when threads read and modify addresses belonging to distinct memory banks. Therefore, data processing operators implemented on a GPU, in addition to contention caused by popular values, have to deal with a new performance limiting factor: thread serialization when accessing values belonging to the same bank.

Modern frameworks, such as Hadoop, combined with abundance of computing resources from the cloud, offer a significant opportunity to address long standing challenges in distributed processing. Infrastructure-as-a-Service clouds reduce the investment cost of renting a large data center while distributed processing frameworks are capable of efficiently harvesting the rented physical resources. Yet, the performance users get out of these resources varies greatly because the cloud hardware is shared by all users. The value for money cloud consumers achieve renders resource sharing policies a key player in both cloud performance and user satisfaction. In this paper, we employ microeconomics to direct the allotment of cloud resources for consumption in highly scalable master-worker virtual infrastructures. Our approach is developed on two premises: the cloud-consumer always has a budget and cloud physical resources are limited. Using our approach, the cloud administration is able to maximize per-user financial profit. We show that there is an equilibrium point at which our method achieves resource sharing proportional to each user's budget. Ultimately, this approach allows us to answer the question of how many resources a consumer should request from the seemingly endless pool provided by the cloud.

Scheduling data processing workflows (dataflows) on the cloud is a very complex and challenging task. It is essentially an optimization problem, very similar to query optimization, that is characteristically different from traditional problems in two aspects: Its space of alternative schedules is very rich, due to various optimization opportunities that cloud computing offers; its optimization criterion is at least two-dimensional, with monetary cost of using the cloud being at least as important as query completion time. In this paper, we study scheduling of dataflows that involve arbitrary data processing operators in the context of three different problems: 1) minimize completion time given a fixed budget, 2) minimize monetary cost given a deadline, and 3) find trade-offs between completion time and monetary cost without any a-priori constraints. We formulate these problems and present an approximate optimization framework to address them that uses resource elasticity in the cloud. To investigate the effectiveness of our approach, we incorporate the devised framework into a prototype system for dataflow evaluation and instantiate it with several greedy, probabilistic, and exhaustive search algorithms. Finally, through several experiments that we have conducted with the prototype elastic optimizer on numerous scientific and synthetic dataflows, we identify several interesting general characteristics of the space of alternative schedules as well as the advantages and disadvantages of the various search algorithms. The overall results are quite promising and indicate the effectiveness of our approach.