Shared-nothing并行数据库系统查询优化技术

查询优化是并行数据库系统的核心技术。该文介绍作者自行研制的一个Ｓｈａｒｅｄ－ｎｏｔｈｉｎｇ并行数据库系统ＰＢＡＳＥ／２中独特的两阶段优化策略。为了缩减并行相称优化庞大的搜索空间,ＰＢＡＳＥ／２将并行查询优化划分为顺序优化和并行化两个在阶段。在顺序优化阶段对并行化后的通信代价进行预先估算,将通信开销加入顺序优化的代价模型,同时对动态规划搜索算法进行了修正和扩展,保证了顺序优化阶段得到的最小代价计划在     

一个并行查询优化器的设计与实现

并行查询优化器的目标是缩减庞大的计划搜索空间,获得优化的查询规划。为此,并行实时数据库PRTD-BASE查询优化器针对无共享结构（SN）,充分考虑通信开销,采用两阶段 优化方法,依据代价估计模型先对查询树进行基于代价估计的顺序优化,然后利用启发式规则对顺序优化的查询计划进行并行化,充分利用了多处理机的并行性,获得了较快的查 询响应时间。     

基于CUDA的并行K-近邻连接算法实现

针对大规模空间数据的K-近邻连接查询问题,设计了一种CUDA编程模型下K-近邻连接算法的并行优化方法。将K-近邻连接算法的并行过程分两个阶段:1)对参与查询的数据集P和Q分别建立R-Tree索引;2)基于R-Tree索引进行KNNJ查询。首先根据结点所在位置划分最小外包框,在CUDA下基于递归网格排序算法创建R-Tree索引。然后在CUDA下基于R-Tree索引进行KNNJ查询,其中涉及并行求距离和并行距离排序两个阶段:求距离阶段利用每一个线程计算任意两点之间的距离,点与点之间距离的求取无依赖并行;排序阶段将快速排序基于CUDA以实现并行化。实验结果表明,随着样本量的不断增大,基于R-Tree索引的并行K-近邻连接算法的优势更加明显,具有高效性和可扩展性。     

列存储数据查询中的连接策略选择方法

列的连接策略优化是列存储数据查询中的重要问题。现有的列存储系统中,列的连接存在策略单一,缺少优化处理,无法满足复杂查询等缺陷。针对这些问题,提出一种连接策略选择方法。该方法首先定义简单规则过滤代价过大的查询计划,生成候选查询计划树。进而根据动态Huffman树原理提出动态优化树算法,对候选查询计划树中的查询执行顺序进行改进。根据列存储数据的特点,候选计划中每个连接节点的执行策略被归纳为两种:串行连接和并行连接。在此基础上构建代价估计模型,集中针对这两种连接策略进行代价估计和策略选择,从而以较小的时间复杂度获得优化的查询执行策略。     

基于Greenplum数据库的查询优化

针对分布式数据库查询效率随着数据规模的增大而降低的问题,以Greenplum分布式数据库为研究对象,从优化查询路径的角度提出一个基于代价的最优查询计划生成方法。首先,该方法设计一种有效的代价模型来估算查询代价;然后,采用并行最大最小蚁群算法来搜索具有最小查询代价的连接顺序,即最优连接顺序;最后,根据Greenplum数据库对查询计划中不同操作的默认最优选择得到最优查询计划。采用该方法在自主生成的数据集与事务处理性能理事会测试基准（TPC-H）的标准数据集上进行了多组实验。实验结果表明,所提出的优化方法能有效地搜索出最优解,获得最优的查询计划,从而提升Greenplum数据库的查询效率。     

列存储数据查询中的连接策略优化方法

列存储数据查询优化的重点是列的连接策略.现有的列存储系统通过存储的改变来简化列的连接,致使列的连接缺少查询优化处理,策略单一且无法满足复杂查询.在剖析现有连接选择策略的基础上,提出一种新的连接策略优化方法,即首先利用基于规则的优化方法为列存储数据查询制定优化规则,过滤不可能产生最优计划的候选计划;然后设计了基于代价的优化算法,根据动态Huffman树和左深连接树原理对查询执行顺序进行改进,进一步减少候选计划的规模;根据列存储数据的特点将候选计划中每个连接节点的执行策略归纳为串行连接和并行连接两类,并在此基础上提出代价估计模型,进而可针对这两种连接策略进行代价估计和策略选择.最后在SSB数据集上通过实验证明了方法在列存储数据查询中的有效性.     

基于PQR-tree的空间查询代价模型

空间信息处理和地理信息系统等领域的数据管理涉及到海量、高维空间数据对象的处理。本文针对传统数据索引结构在处理这类空间数据时所存在的内存使用过大、I/O消耗过多等问题,通过改进选择查询的代价模型,给出了基于PQR-tree的查询和代价模型,以提高空间数据查询的性能。提出了基于PQR-tree的三阶段并行查询的方法,分别在任务创建、分配、执行阶段进行优化。提出在任务创建和任务分配阶段应用于空间查询中过滤和精炼阶段的有效算法。测试表明,本文算法在处理各种不同分布类型数据集过程中有效降低了空间数据处理对时间和空间的代价和需求,并且并行机制下的代价模型在预测和评估方面也具有较好的精确度。     

基于BSP的SPARQL基本图模式查询算法

随着语义网的不断发展,发布在互联网上的资源描述框架(RDF)数据达到百亿级三元组规模,并且呈现几何增长趋势,针对RDF数据的单机SPARQL查询方法已经不再适用。为此,提出一种基于整体同步并行(BSP)模型的SPARQL基本图模式查询算法。根据RDF有向图数据特性及基本图模式定义,将整个查询过程分成匹配和迭代2个阶段,在匹配出所需查询的三元组模式后,通过迭代使部分解逐步逼近完全解,得到最终查询结果。利用HAMA分布式计算框架进行算法实现,实验结果表明,与基于MapReduce的SPARQL查询算法相比,该算法具有较高的查询效率,能为大规模RDF数据的快速SPARQL查询提供支持。     

并行数据库系统PBASE/2的查询优化机制

并行执行计划的搜索空间呈指数级增长,如何高效地裁剪搜索空间是并行查询优化的关键所在,ＰＢＡＳＥ／２以流机制为基础,采用基于代价估算和启发式规则的改良的两阶段优化,有交地解决了这一问题。     

Extensible Parallel Query Processing for Exploratory Geoscientific Data Mining

Exploratory data mining and analysis requires a computing environment which provides facilities for the user-friendly expression and rapid execution of scientific queries. In this paper, we address research issues in the parallelization of scientific queries containing complex user-defined operations. In a parallel query execution environment, parallelizing a query execution plan involves determining how input data streams to evaluators implementing logical operations can be divided to be processed by clones of the same evaluator in parallel. We introduced the concept of relevance window that characterizes data lineage and data partitioning opportunities available for an user-defined evaluator. In addition, we developed a query parallelization framework by extending relational parallel query optimization algorithms to allow the parallelization characteristics of user-defined evaluators to guide the process of query parallelization in an extensible query processing environment. We demonstrated the utility of our system by performing experiments mining cyclonic activity, blocking events, and the upward wave-energy propagation features from several observational and model simulation datasets.     

A survey of parallel execution strategies for transitive closure and logic programs

An important feature of database technology of the nineties is the use of parallelism for speeding up the execution of complex queries. This technology is being tested in several experimental database architectures and a few commercial systems for conventional select-project-join queries. In particular, hash-based fragmentation is used to distribute data to disks under the control of different processors in order to perform selections and joins in parallel. With the development of new query languages, and in particular with the definition of transitive closure queries and of more general logic programming queries, the new dimension of recursion has been added to query processing. Recursive queries are complex; at the same time, their regular structure is particularly suited for parallel execution, and parallelism may give a high efficiency gain. We survey the approaches to parallel execution of recursive queries that have been presented in the recent literature. We observe that research on parallel execution of recursive queries is separated into two distinct subareas, one focused on the transitive closure of Relational Algebra expressions, the other one focused on optimization of more general Datalog queries. Though the subareas seem radically different because of the approach and formalism used, they have many common features. This is not surprising, because most typical Datalog queries can be solved by means of the transitive closure of simple algebraic expressions. We first analyze the relationship between the transitive closure of expressions in Relational Algebra and Datalog programs. We then review sequential methods for evaluating transitive closure, distinguishing iterative and direct methods. We address the parallelization of these methods, by discussing various forms of parallelization. Data fragmentation plays an important role in obtaining parallel execution; we describe hash-based and semantic fragmentation. Finally, we consider Datalog queries, and present general methods for parallel rule execution; we recognize the similarities between these methods and the methods reviewed previously, when the former are applied to linear Datalog queries. We also provide a quantitative analysis that shows the impact of the initial data distribution on the performance of methods. Recommended by: Patrick Valduriez     

An introduction to compilation issues for parallel machines

The exploitation of today's high-performance computer systems requires the effective use of parallelism in many forms and at numerous levels. This survey article discusses program analysis and restructuring techniques that target parallel architectures. We first describe various categories of architectures that are oriented toward parallel computation models: vector architectures, shared-memory multiprocessors, massively parallel machines, message-passing architectures, VLIWs, and multithreaded architectures. We then describe a variety of optimization techniques that can be applied to sequential programs to effectively utilize the vector and parallel processing units. After an overview of basic dependence analysis, we present restructuring transformations on DO loops targeted both to vectorization and to concurrent execution, interprocedural and pointer analysis, task scheduling, instruction-level parallelization, and compiler-assisted data placement. We conclude that although tremendous advances have been made in dependence theory and in the development of a toolkit of transformations, parallel systems are used most effectively when the programmer interacts in the optimization process.     

A software architecture for user transparent parallel image processing

This paper describes a software architecture that allows image processing researchers to develop parallel applications in a transparent manner. The architecture's main component is an extensive library of data parallel low level image operations capable of running on homogeneous distributed memory MIMD-style multicomputers. Since the library has an application programming interface identical to that of an existing sequential library, all parallelism is completely hidden from the user.The first part of the paper discusses implementation aspects of the parallel library, and shows how sequential as well as parallel operations are implemented on the basis of so-called parallelizable patterns. A library built in this manner is easily maintainable, as extensive code redundancy is avoided. The second part of the paper describes the application of performance models to ensure efficiency of execution on all target platforms. Experiments show that for a realistic application performance predictions are highly accurate. These results indicate that the core of the architecture forms a powerful basis for automatic parallelization and optimization of a wide range of imaging software.     

网络处理器同步问题的编译优化方法

网络处理器具有并行体系结构,而其高级语言往往具有串行语义。对串行程序进行并行化编译要求引入同步,而同步的优劣又影响生成代码的执行效率。针对网络处理器上的程序,提出一个对同步进行优化的程序划分算法以增加程序的并行性。实验数据表明,在一些有代表性的网络应用上,该算法可提高程序的并行性,并提升性能。     

An algorithmic framework for solving large-scale multistage stochastic mixed 0–1 problems with nonsymmetric scenario trees. Part II: Parallelization

This note is a sequel of paper (Escudero et al. (2012) [1]), in which the sequential Branch-and-Fix Coordination referred to as the BFC-MS algorithm was introduced for solving large-scale multistage mixed 0–1 optimization problems up to optimality under uncertainty. The aim of the note is to present the parallelization version of the BFC algorithm, referred to as PC-BFCMS, so that the elapsed time reduction on problem solving is analyzed. The parallelization is performed at two levels. The inner level parallelizes the optimization of the MIP submodels attached to the set of scenario clusters that have been created by the modeler-defined break stage to decompose the original problem, where the nonanticipativity constraints are partially relaxed. Several strategies are presented for analyzing the performance of the inner parallel computation based on MPI (Message Passing Interface) threads to solve scenario cluster submodels versus the sequential version of the BFC-MS methodology. The outer level of parallelization defines a set of 0–1 variables, the combinations of whose 0–1 values, referred to as paths (one for each combination), allow independent models to be optimized in parallel, such that each one can itself be internally optimized with the inner parallelization approach. The results of using the outer–inner parallelization are remarkable: the larger the number of paths and MPI threads (in addition to the number of threads that the MIP solver allows to be used), the smaller the elapsed time to obtain the optimal solution. This new approach allows problems to be solved faster, and, can thus solve very large scale problems that could not otherwise be solved by plain use of a state-of the-art MIP solver, or could not be solved by the sequential version of the decomposition algorithm in acceptable elapsed time.     

GPU acceleration of the stochastic grid bundling method for early-exercise options

In this work, a parallel graphics processing units (GPU) version of the Monte Carlo stochastic grid bundling method (SGBM) for pricing multi-dimensional early-exercise options is presented. To extend the method's applicability, the problem dimensions and the number of bundles will be increased drastically. This makes SGBM very expensive in terms of computational costs on conventional hardware systems based on central processing units. A parallelization strategy of the method is developed and the general purpose computing on graphics processing units paradigm is used to reduce the execution time. An improved technique for bundling asset paths, which is more efficient on parallel hardware is introduced. Thanks to the performance of the GPU version of SGBM, a general approach for computing the early-exercise policy is proposed. Comparisons between sequential and GPU parallel versions are presented.     

基于多核阵列体系结构的嵌套循环并行优化

多核处理器已广泛应用于高性能计算领域,如何有效地将传统串行程序转换为并行代码并减少程序中嵌套循环所占用时间仍是该领域的挑战性问题。本文首先基于多面体模型对嵌套循环进行依赖特征分析并实现瓦片分割,据此自动生成粗粒度并行代码。针对多核阵列处理器的结构特点,采用遗传算法生成通信优化的瓦片任务序列,在此基础上建立了有效的任务调度模型。最后将上述方法应用于LU分解,结果表明该方法与传统调度算法相比,在增加数据局部性、实现负载平衡方面具有更好效果。