Incremental Design of a Data Warehouse

A data warehouse (DW) can be seen as a set of materialized views defined over remote base relations. When a query is posed, it is evaluated locally, using the materialized views, without accessing the original information sources. The DWs are dynamic entities that evolve continuously over time. As time passes, new queries need to be answered by them. Some of these queries can be answered using exclusively the materialized views. In general though new views need to be added to the DW.In this paper we investigate the problem of incrementally designing a DW when new queries need to be answered and possibly extra space is allocated for view materialization. Based on an AND/OR dag representation of multiple queries, we model the problem as a state space search problem. We design incremental algorithms for selecting a set of new views to additionally materialize in the DW that: (a) fits in the extra space, (b) allows a complete rewriting of the new queries over the materialized views, and (c) minimizes the combined new query evaluation and new view maintenance cost. Finally, we discuss methods for pruning the search space so that efficiency is improved.     

View selection for designing the global data warehouse

A global data warehouse (DW) integrates data from multiple distributed heterogeneous databases and other information sources. A global DW can be abstractly seen as a set of materialized views. The selection of views for materialization in a DW is an important decision in the design of a DW. Current commercial products do not provide tools for automatic DW design. We provide a general method that, given a set of select-project-join queries to be satisfied by the DW, generates sets of materialized views that satisfy all the input queries. This process is complex since ‘common subexpressions' between the queries need to be detected and exploited. Our method is then applied to solve the problem of selecting such a materialized view set that fits in the space allocated to the DW for materialization and minimizes the combined overall query evaluation and view maintenance cost. We design algorithms which are implemented and we report on their experimental evaluation.     

Detecting redundant materialized views in data warehouse evolution

A data warehouse (DW) can be abstractly seen as a set of materialized views defined over a set of remote data sources. A DW is intended to satisfy a set of queries. The views materialized in a DW relate to each other in a complex manner, through common subexpressions, in order to guarantee high query performance and low view maintenance cost. DWs are time varying. As time passes new materialized views are added in order to satisfy new queries, or for performance reasons, while old queries are dropped. The evolution of a DW can result in a redundant set of materialized views. In this paper, we address the problem of detecting redundant materialized views in a given DW view selection, that is, materialized views that can be removed from DW without negatively affecting the query evaluation or the view maintenance process. Using an AND/OR dag representation for multiple queries and views, we first formalize the process of propagating source relation changes to the materialized views by exploiting common subexpressions between views and by using other materialized views that are not affected by these changes. Then, we provide an algorithm for detecting materialized views that are not needed in the process of propagating source relation changes to the DW. We also show how trivially redundant views can be identified in this process. Finally, we use these results to provide a procedure for detecting materialized views that are redundant in a DW. Our approach considers a broad class of views that includes grouping/aggregation views and is not dependent on a specific cost model.     

Answering queries using materialized views with minimum size

In this paper, we study the following problem. Given a database and a set of queries, we want to find a set of views that can compute the answers to the queries, such that the amount of space, in bytes, required to store the viewset is minimum on the given database. (We also handle problem instances where the input has a set of database instances, as described by an oracle that returns the sizes of view relations for given view definitions.) This problem is important for applications such as distributed databases, data warehousing, and data integration. We explore the decidability and complexity of the problem for workloads of conjunctive queries. We show that results differ significantly depending on whether the workload queries have self-joins. Further, for queries without self-joins we describe a very compact search space of views, which contains all views in at least one optimal viewset. We present techniques for finding a minimum-size viewset for a single query without self-joins by using the shape of the query and its constraints, and validate the approach by extensive experiments. Part of this article was published elsewhere [Chirkova, R., Li, C.: Materializing views with minimal size to answer queries. PODS (2003)]. In addition to the prior materials, this article contains new theoretical results, as well as new results on how to efficiently implement the proposed techniques (Sects. 5 and 5.4)     

Web数据集成系统基于QC模型的物化视图选择

在Web数据集成系统中,物化视图能够有效地减少网络传输代价,提高系统的查询效率．如何选择查询进行物化,使得选中的查询满足集成层的空间限制,同时获取最大物化收益,成为集成系统中一个迫切需要解决的问题．传统方法没有考虑到海量XML查询之间的包含关系,其选择的物化视图中可能包含冗余的信息．针对上述问题,提出了①Web数据集成系统中海量查询集合的QC(query containment)模型,该模型能够捕捉查询之间最常见的包含关系;②基于QC模型的物化视图选择算法,算法考虑了物化视图选择相关的主要因素,包括查询提交的频率、空间代价、查询重写能力和查询结果的完备性,提出了查询位图的物化视图组织方式,从而获取更加合理的物化视图选择方案．实验结果证明了该方法的有效性．     

Selection of views to materialize in a data warehouse

A data warehouse stores materialized views of data from one or more sources, with the purpose of efficiently implementing decision-support or OLAP queries. One of the most important decisions in designing a data warehouse is the selection of materialized views to be maintained at the warehouse. The goal is to select an appropriate set of views that minimizes total query response time and the cost of maintaining the selected views, given a limited amount of resource, e.g., materialization time, storage space, etc. In This work, we have developed a theoretical framework for the general problem of selection of views in a data warehouse. We present polynomial-time heuristics for a selection of views to optimize total query response time under a disk-space constraint, for some important special cases of the general data warehouse scenario, viz.: 1) an AND view graph, where each query/view has a unique evaluation, e.g., when a multiple-query optimizer can be used to general a global evaluation plan for the queries, and 2) an OR view graph, in which any view can be computed from any one of its related views, e.g., data cubes. We present proofs showing that the algorithms are guaranteed to provide a solution that is fairly close to (within a constant factor ratio of) the optimal solution. We extend our heuristic to the general AND-OR view graphs. Finally, we address in detail the view-selection problem under the maintenance cost constraint and present provably competitive heuristics.     

数据仓库中物化视图选择策略

为了提高决策支持和OLAP查询的响应效率,数据仓库多采用物化视图的思想.因此,物化视图的选择策略是数据仓库研究的重要问题之一.其目标是选出一组存储、维护代价与查询代价的总和为最小的物化视图.提出一个以MVPP(multi-view processing plan)为视图选择的搜索空间的物化视图选择新算法--VSMF(views selection base on multi-factor)算法.该算法在存储空间约束下同时实现多查询最优化和视图维护最优化.     

Answering Queries Using Limited External Query Processors

When answering queries using external information sources, the contents of the queries can be described by views. To answer a query, we must rewrite it using the set of views presented by the sources. When the external information sources also have the ability to answer some (perhaps limited) sets of queries that require performing operations on their data, the set of views presented by the source may be infinite (albeit encoded in some finite fashion). Previous work on answering queries using views has only considered the case where the set of views is finite. In order to exploit the ability of information sources to answer more complex queries, we consider the problem of answering conjunctive queries using infinite sets of conjunctive views. Our first result is that an infinite set of conjunctive views can be partitioned into a finite number of equivalence classes, such that picking one view from every nonempty class is sufficient to determine whether the query can be answered using the views. Second, we show how to compute the set of equivalence classes for sets of conjunctive views encoded by a datalog program. Furthermore, we extend our results to the case when the query and the views use the built-in predicates <, ⩽, =, and ≠, and they are interpreted over a dense domain. Finally, we extend our results to conjunctive queries and views with the built-in predicates <, ⩽, and = interpreted over the integers. In doing so we present a result of independent interest, namely, an algorithm to minimize such queries.     

Efficient approaches for materialized views selection in a data warehouse

View materialization is an effective method to increase query efficiency in a data warehouse and improve OLAP query performance. However, one encounters the problem of space insufficiency if all possible views are materialized in advance. Reducing query time by means of selecting a proper set of materialized views with a lower cost is crucial for efficient data warehousing. In addition, the costs of data warehouse creation, query, and maintenance have to be taken into account while views are materialized. In this paper, we propose efficient algorithms to select a proper set of materialized views, constrained by storage and cost considerations, to help speed up the entire data warehousing process. We derive a cost model for data warehouse query and maintenance as well as efficient view selection algorithms that effectively exploit the gain and loss metrics. The main contribution of our paper is to speed up the selection process of materialized views. Concurrently, this will greatly reduce the overall cost of data warehouse query and maintenance.     

Optimization in Data Cube System Design

The design of an OLAP system for supporting real-time queries is one of the major research issues. One approach is to use data cubes, which are materialized precomputed multidimensional views of data in a data warehouse. We can derive a set of data cubes to answer each frequently asked query directly. However, there are two practical problems: (1) the maintenance cost of the data cubes, and (2) the query cost to answer those queries. Maintaining a data cube requires disk storage and CPU computation, so the maintenance cost is related to the total size as well as the total number of data cubes materialized. In most cases, materializing all data cubes is impractical. The maintenance cost may be reduced by merging some data cubes. However, the resulting larger data cubes will increase the query cost of answering some queries. If the bounds on the maintenance cost and the query cost are too strict, we help the user decide which queries to be sacrificed and not taken into consideration. We have defined an optimization problem in data cube system design. Given a maintenance-cost bound, a query-cost bound and a set of frequently asked queries, it is necessary to determine a set of data cubes such that the system can answer a largest subset of the queries without violating the two bounds. This is an NP-hard problem. We propose approximate Greedy algorithms GR, 2GM and 2GMM, which are shown to be both effective and efficient by experiments done on a census data set and a forest-cover-type data set.     

Parallel multisource view maintenance

In a distributed environment, materialized views are used to integrate data from different information sources and then store them in some centralized location. In order to maintain such materialized views, maintenance queries need to be sent to information sources by the data warehouse management system. Due to the independence of the information sources and the data warehouse, concurrency issues are raised between the maintenance queries and the local update transactions at each information source. Recent solutions such as ECA and Strobe tackle such concurrent maintenance, however with the requirement of quiescence of the information sources. SWEEP and POSSE overcome this limitation by decomposing the global maintenance query into smaller subqueries to be sent to every information source and then performing conflict correction locally at the data warehouse. Note that all these previous approaches handle the data updates one at a time. Hence either some of the information sources or the data warehouse is likely to be idle during most of the maintenance process. In this paper, we propose that a set of updates should be maintained in parallel by several concurrent maintenance processes so that both the information sources as well as the warehouse would be utilized more fully throughout the maintenance process. This parallelism should then improve the overall maintenance performance. For this we have developed a parallel view maintenance algorithm, called PVM, that substantially improves upon the performance of previous maintenance approaches by handling a set of data updates at the same time. The parallel handling of a set of updates is orthogonal to the particular maintenance algorithm applied to the handling of each individual update. In order to perform parallel view maintenance, we have identified two critical issues that must be overcome: (1) detecting maintenance-concurrent data updates in a parallel mode and (2) correcting the problem that the data warehouse commit order may not correspond to the data warehouse update processing order due to parallel maintenance handling. In this work, we provide solutions to both issues. For the former, we insert a middle-layer timestamp assignment module for detecting maintenance-concurrent data updates without requiring any global clock synchronization. For the latter, we introduce the negative counter concept to solve the problem of variant orders of committing effects of data updates to the data warehouse. We provide a proof of the correctness of PVM that guarantees that our strategy indeed generates the correct final data warehouse state. We have implemented both SWEEP and PVM in our EVE data warehousing system. Our performance study demonstrates that a manyfold performance improvement is achieved by PVM over SWEEP.Received: 12 November 2001, Accepted: 18 December 2002, Published online: 31 July 2003This work was supported in part by the NSF NYI grant IIS-979624 and NSF CISE Instrumentation grant IRIS 97-29878 and NSF grant IIS-9988776.     

Incremental join view maintenance on distributed log-structured storage

Modern database systems desperate for the ability to support highly scalable transactions and efficient queries simultaneously for real-time applications. One solution is to utilize query optimization techniques on the on-line transaction processing (OLTP) systems. The materialized view is considered as a panacea to decrease query latency. However, it also involves the significant cost of maintenance which trades away transaction performance. In this paper, we examine the design space and conclude several design features for the implementation of a view on a distributed log-structured merge-tree (LSMtree), which is a well-known structure for improving data write performance. As a result, we develop two incremental view maintenance (IVM) approaches on LSM-tree. One avoids join computation in view maintenance transactions. Another with two optimizations is proposed to decouple the view maintenance with the transaction process. Under the asynchronous update, we also provide consistency queries for views. Experiments on TPC-H benchmark show our methods achieve better performance than straightforward methods on different workloads.     

Selecting and using views to compute aggregate queries

We consider a workload of aggregate queries and investigate the problem of selecting materialized views that (1) provide equivalent rewritings for all the queries, and (2) are optimal, in that the cost of evaluating the query workload is minimized. We consider conjunctive views and rewritings, with or without aggregation; in each rewriting, only one view contributes to computing the aggregated query output. We look at query rewriting using existing views and at view selection. In the query-rewriting problem, we give sufficient and necessary conditions for a rewriting to exist. For view selection, we prove complexity results. Finally, we give algorithms for obtaining rewritings and selecting views.     

A Genetic Selection Algorithm for OLAP Data Cubes

Multidimensional data analysis, as supported by OLAP (online analytical processing) systems, requires the computation of many aggregate functions over a large volume of historically collected data. To decrease the query time and to provide various viewpoints for the analysts, these data are usually organized as a multidimensional data model, called data cubes. Each cell in a data cube corresponds to a unique set of values for the different dimensions and contains the metric of interest. The data cube selection problem is, given the set of user queries and a storage space constraint, to select a set of materialized cubes from the data cubes to minimize the query cost and/or the maintenance cost. This problem is known to be an NP-hard problem. In this study, we examined the application of genetic algorithms to the cube selection problem. We proposed a greedy-repaired genetic algorithm, called the genetic greedy method. According to our experiments, the solution obtained by our genetic greedy method is superior to that found using the traditional greedy method. That is, within the same storage constraint, the solution can greatly reduce the amount of query cost as well as the cube maintenance cost.     

A formal perspective on the view selection problem

The view selection problem is to choose a set of views to materialize over a database schema, such that the cost of evaluating a set of workload queries is minimized and such that the views fit into a prespecified storage constraint. The two main applications of the view selection problem are materializing views in a database to speed up query processing, and selecting views to materialize in a data warehouse to answer decision support queries. In addition, view selection is a core problem for intelligent data placement over a wide-area network for data integration applications and data management for ubiquitous computing. We describe several fundamental results concerning the view selection problem. We consider the problem for views and workloads that consist of equality-selection, project and join queries, and show that the complexity of the problem depends crucially on the quality of the estimates that a query optimizer has on the size of the views it is considering to materialize. When a query optimizer has good estimates of the sizes of the views, we show a somewhat surprising result, namely, that an optimal choice of views may involve a number of views that is exponential in the size of the database schema. On the other hand, when an optimizer uses standard estimation heuristics, we show that the number of necessary views and the expression size of each view are polynomially bounded. Received: November 20, 1001 / Accepted: May 30, 2002 / Published online: September 25, 2002     

On solving efficiently the view selection problem under bag and bag-set semantics

In this paper, we investigate the problem of view selection for workloads of conjunctive queries under bag and bag-set semantics. In particular, for both semantics we aim to limit the search space of candidate viewsets. We also start delineating boundaries between query workloads for which certain even more restricted search spaces suffice. They suffice in the sense that they do not compromise optimality in that they contain at least one of the optimal solutions. We start with the general case for both bag and bag-set semantics, where we give a tight condition that candidate views can satisfy and still the search space (thus limited) does contain at least one optimal solution. We show that these results, for both semantics, reduce the size of the search space significantly. Further on, due to this analysis for both semantics, a delineation of the space of viewsets and the space of the corresponding equivalent rewritings for a certain conjunctive query workload is given. We show that for chain query workloads under both bag and bag-set semantics, taking only chain views may miss optimal solutions, whereas, if we further limit the queries to be path-queries (i.e., chain queries over a single binary relation), then, under bag semantics, path-views suffice. Concentrating to bag-set semantics, we show that the path-viewsets do not suffice for every path-query workload.     

一种空间数据仓库动态物化视图选择算法

物化视图能够有效地提高空间数据仓库的查询效率,但由于空间操作的复杂性,传统数据仓库中物化视图的选择算法不能很好地应用于空间数据仓库。为了在存储空间约束下选择查询进行物化,并动态调整物化视图集,以适应用户查询的时变性和即席查询,提出了空间物化视图选择算法SMVS。实验结果表明该算法是有效可行的,不仅能够提高查询性能,而且解决了查询响应性能随用户查询分布变化而下降的问题。     

基于多维护策略的物化视图选择方法

物化视图是数据仓库环境中提高OLAP查询效率的重要手段，因此，物化视图的选择是数据仓库设计中重要的决策之一。本文提出的物化视图选择方法目标是选择合适的视图进行物化，使得查询处理的总代价和物化视图的维护代价最低，提出了物化视图收益模型，并在此基础上基于视图的多维护策略提出了物化视图选择的方法：基于增量和重计算的物化视图选择算法IRMVS、基于增量策略的物化视图选择算法IMVS和基于重计算策略的物化视图选择算法RMVs和基于增量策略的物化后代视图选择算法IMDVS，理论分析和实验表明这些算法是有效可行的。     

数据方体系统设计中的优化问题

支持实时查询的联机分析处理系统的设计是当前一个很重要的研究问题。其中常用的方法是使用数据方体来实现。对于出现频率较高的查询,可以给出对应的数据方体集,使得每个查询都可以直接得到回答。但是在设计基于方体的系统时,需要考虑以下两个问题:(1)数据方体的维护成本,(2)回答频繁查询的响应时间。在用户给出了维护成本上限和响应时间上限后,需要对数据方体集进行优化,使得系统能够满足用户的要求,并回答尽可能多的查询。文章给出了数据方体系统设计优化问题的定义,这是一个NP完全问题,并提出了贪心删除和贪心合并的近似算法。实验表明了算法的有效性。