数据仓库系统中层次式Cube存储结构

区域查询是数据仓库上支持联机分析处理(on-line analytical processing,简称OLAP)的重要操作.近几年,人们提出了一些支持区域查询和数据更新的Cube存储结构.然而这些存储结构的空间复杂性和时间复杂性都很高,难以在实际中使用.为此,提出了一种层次式Cube存储结构HDC(hierarchical data cube)及其上的相关算法.HDC上区域查询的代价和数据更新代价均为O(log^dn),综合性能为O((logn)^2d)(使用C_qC_u模型)或O(K(logn)^d)(使用C_qn_q+C_un_u模型).理论分析与实验表明,HDC的区域查询代价、数据更新代价、空间代价以及综合性能都优于目前所有的Cube存储结构.     

High Performance OLAP and Data Mining on Parallel Computers

On-Line Analytical Processing (OLAP) techniques are increasingly being used in decision support systems to provide analysis of data. Queries posed on such systems are quite complex and require different views of data. Analytical models need to capture the multidimensionality of the underlying data, a task for which multidimensional databases are well suited. Multidimensional OLAP systems store data in multidimensional arrays on which analytical operations are performed. Knowledge discovery and data mining requires complex operations on the underlying data which can be very expensive in terms of computation time. High performance parallel systems can reduce this analysis time. Precomputed aggregate calculations in a Data Cube can provide efficient query processing for OLAP applications. In this article, we present algorithms for construction of data cubes on distributed-memory parallel computers. Data is loaded from a relational database into a multidimensional array. We present two methods, sort-based and hash-based for loading the base cube and compare their performances. Data cubes are used to perform consolidation queries used in roll-up operations using dimension hierarchies. Finally, we show how data cubes are used for data mining using Attribute Focusing techniques. We present results for these on the IBM-SP2 parallel machine. Results show that our algorithms and techniques for OLAP and data mining on parallel systems are scalable to a large number of processors, providing a high performance platform for such applications.     

Time Lattice: A Data Structure for the Interactive Visual Analysis of Large Time Series

Advances in technology coupled with the availability of low‐cost sensors have resulted in the continuous generation of large time series from several sources. In order to visually explore and compare these time series at different scales, analysts need to execute online analytical processing (OLAP) queries that include constraints and group‐by's at multiple temporal hierarchies. Effective visual analysis requires these queries to be interactive. However, while existing OLAP cube‐based structures can support interactive query rates, the exponential memory requirement to materialize the data cube is often unsuitable for large data sets. Moreover, none of the recent space‐efficient cube data structures allow for updates. Thus, the cube must be re‐computed whenever there is new data, making them impractical in a streaming scenario. We propose Time Lattice, a memory‐efficient data structure that makes use of the implicit temporal hierarchy to enable interactive OLAP queries over large time series. Time Lattice is a subset of a fully materialized cube and is designed to handle fast updates and streaming data. We perform an experimental evaluation which shows that the space efficiency of the data structure does not hamper its performance when compared to the state of the art. In collaboration with signal processing and acoustics research scientists, we use the Time Lattice data structure to design the Noise Profiler, a web‐based visualization framework that supports the analysis of noise from cities. We demonstrate the utility of Noise Profiler through a set of case studies.     

Graph-Based Parallel Query Processing and Optimization Strategies for Object-Oriented Databases

Much work has been accomplished in the past on the subject of parallel query processing and optimization in parallel relational database systems; however, little work on the same subject has been done in parallel object-oriented database systems. Since the object-oriented view of a database and its processing are quite different from those of a relational system, it can be expected that techniques of parallel query processing and optimization for the latter can be different from the former. In this paper, we present a general framework for parallel object-oriented database systems and several implemented query processing and optimization strategies together with some performance evaluation results. In this work, multiwavefront algorithms are used in query processing to allow a higher degree of parallelism than the traditional tree-based query processing. Four optimization strategies, which are designed specifically for the multiwavefront algorithms and for the optimization of single as well as multiple queries, are introduced. The query processing algorithms and optimization strategies have been implemented on a parallel computer, nCUBE2; and the results of a performance evaluation are presented in this paper. The main emphases and the intended contributions of this paper are (1) data partitioning, query processing and optimization strategies suitable for parallel OODBMSs, (2) the implementation of the multiwavefront algorithms and optimization strategies, and (3) the performance evaluation results.     

Performance analysis of parallel object-oriented query processing algorithms

Advanced application domains such as computer-aided design, computer-aided software engineering, and office automation are characterized by their need to store, retrieve, and manage large quantities of data having complex structures. A number of object-oriented database management systems (OODBMS) are currently available that can effectively capture and process the complex data. The existing implementations of OODBMS outperform relational systems by maintaining and querying cross-references among related objects. However, the existing OODBMS still do not meet the efficiency requirements of advanced applications that require the execution of complex queries involving the retrieval of a large number of data objects and relationships among them. Parallel execution can significantly improve the performance of complex OO queries. In this paper, we analyze the performance of parallel OO query processing algorithms for various benchmark application domains. The application domains are characterized by specific mixes of queries of different semantic complexities. The performance of the application domains has been analyzed for various system and data parameters by running parallel programs on a 32-node transputer based parallel machine developed at the IBM Research Center at Yorktown Heights. The parallel processing algorithms, data routing techniques, and query management and control strategies have been implemented to obtain accurate estimation of controlling and processing overheads. However, generation of large complex databases for the study was impractical. Hence, the data used in the simulation have been parameterized. The parallel OO query processing algorithms analyzed in this study are based on a query graph approach rather than the traditional query tree approach. Using the query graph approach, a query is processed by simultaneously initiating the execution at several object classes, thereby, improving the parallelism. During processing, the algorithms avoid the execution of time-consuming join operations by making use of the object references among the objects. Further, the algorithms do not generate any temporary data, thereby, reducing disk accesses. This is accomplished by marking the selected objects and by employing a two-phase query processing strategy.     

基于集合运算的数据立方体结构

提出一种新的数据立方体结构,通过索引和集合的交并运算来获得查询结果,特别是在进行区域查询时,避免了将区域分解为点后再依次进行点查询的方式,从而在保持较少的磁盘空间和较好的点查询响应速度的情况下,改善区域查询的性能;同时给出其生成和查询算法,并使用合成数据和实际数据进行了实验验证.     

PARSIMONY: An Infrastructure for Parallel Multidimensional Analysis and Data Mining

Multidimensional analysis and online analytical processing (OLAP) operations require summary information on multidimensional data sets. Most common are aggregate operations along one or more dimensions of numerical data values. Simultaneous calculation of multidimensional aggregates are provided by the Data Cube operator, used to calculate and store summary information on a number of dimensions. This is computed only partially if the number of dimensions is large. Query processing for these applications requires different views of data to gain insight and for effective decision support. Queries may either be answered from a materialized cube in the data cube or calculated on the fly.  The multidimensionality of the underlying problem can be represented both in relational and in multidimensional databases, the latter being a better fit when query performance is the criteria for judgment. Relational databases are scalable in size for OLAP and multidimensional analysis and efforts are on to make their performance acceptable. On the other hand multidimensional databases have proven to provide good performance for such queries, although they are not very scalable. In this article we address (1) scalability in multidimensional systems for OLAP and multidimensional analysis and (2) integration of data mining with the OLAP framework. We describe our system PARSIMONY, parallel and scalable infrastructure for multidimensional online analytical processing, used for both OLAP and data mining. Sparsity of data sets is handled by using chunks to store data either as a dense block using multidimensional arrays or as sparse representation using a bit encoded sparse structure. Chunks provide a multidimensional index structure for efficient dimension oriented data accesses much the same as multidimensional arrays do. Operations within chunks and between chunks are a combination of relational and multidimensional operations depending on whether the chunk is sparse or dense. Further, we develop parallel algorithms for data mining on the multidimensional cube structure for attribute-oriented association rules and decision-tree-based classification. These take advantage of the data organization provided by the multidimensional data model.  Performance results for high dimensional data sets on a distributed memory parallel machine (IBM SP-2) show good speedup and scalability.     

Parallel evaluation of the transitive closure of a database relation

Parallelism is a promising approach to high performance data management. In a highly parallel data server with declustered data placement, an important issue is to exploit parallelism in processing complex queries such as recursive queries. In this paper, we consider the transitive closure of a database relation as a paradigm to study parallel recursive query processing. And we propose two new parallel algorithms for evaluating the transitive closure of a relation in a parallel data server. Performance comparisons based on an analytical model indicate the superior response time of the parallel algorithms over their centralized version. With one hundred nodes, performance gain is between one and two orders of magnitude. One parallel algorithm provides superior response time while the other exhibits better response time/total time trade-off.This work had been done within the Advanced Computer Architecture Program, Microelectronics and Computer Technology Corporation, Austin, Texas. The current affiliation of Setrag Khoshafian is Ashton-Tate, Walnut Creek, California.     

Efficient parallel processing of range queries through replicated declustering

A common technique used to minimize I/O in data intensive applications is data declustering over parallel servers. This technique involves distributing data among several disks so as to parallelize query retrieval and thus, improve performance. We focus on optimizing access to large spatial data, and the most common type of queries on such data, i.e., range queries. An optimal declustering scheme is one in which the processing for all range queries is balanced uniformly among the available disks. It has been shown that single copy based declustering schemes are non-optimal for range queries. In this paper, we integrate replication in conjunction with parallel disk declustering for efficient processing of range queries. We note that replication is largely used in database applications for several purposes like load balancing, fault tolerance and availability of data. We propose theoretical foundations for replicated declustering and propose a class of replicated declustering schemes, periodic allocations, which are shown to be strictly optimal for a number of disks. We propose a framework for replicated declustering, using a limited amount of replication and provide extensions to apply it on real data, which include arbitrary grids and a large number of disks. Our framework also provides an effective indexing scheme that enables fast identification of data of interest in parallel servers. In addition to optimal processing of single queries, we show that this framework is effective for parallel processing of multiple queries. We present experimental results comparing the proposed replication scheme to other techniques for both single queries and multiple queries, on synthetic and real data sets. Recommended by: Ahmed Elmagarmid Supported by U.S. Department of Energy (DOE) Award No. DE-FG02-03ER25573, and National Science Foundation (NSF) grant CNS-0403342.     

Indexing range sum queries in spatio-temporal databases

Although spatio-temporal databases have received considerable attention recently, there has been little work on processing range sum queries on the historical records of moving objects despite their importance. Since the direct access to a huge amount of data to answer range sum queries incurs prohibitive computation cost, materialization techniques based on existing index structures are suggested. A simple but effective solution is to apply the materialization technique to the MVR-tree known as the most efficient structure for window queries with spatio-temporal conditions. Aggregate structures based on other index structures such as the HR-tree and the 3DR-tree do not provide satisfactory query performance. In this paper, we propose a new index structure called the Adaptively Partitioned Aggregate R-Tree (APART) and query processing algorithms to efficiently process range sum queries in many situations. Our experimental results show that the performance of the APART is typically 1.3 times better than that of its competitor for a wide range of scenarios.     

Algorithms for asynchronous parallel processing of object-orienteddatabases

Management of large quantities of complex data is essential in many advanced application areas. Object-oriented (OO) database management system have been developed to effectively model and process the complex domain knowledge. They have been shown to outperform some existing relational systems. The existing implementations of OO database management systems attempt to improve the efficiency of OO queries by explicitly capturing the relationships among objects. However, the execution of complex queries involving the retrieval of objects from many classes and relationships among them causes the existing system to operate inefficiently. In this paper, we present parallel algorithms for the processing of queries against a large OO database. The algorithms are based on a closed model of query processing pattern-based access instead of the conventional value-based access. During processing, the algorithms avoid the execution of time-consuming join operations by making use of the explicitly stored object associations. Generation of large quantities of temporary data is avoided by marking objects using their identifiers and by employing a two-phase query processing strategy. A query is processed by concurrent multiple waves, thereby improving parallelism avoiding the complexities introduced in their sequential implementation. The correctness and the performance of the parallel algorithms have been tested and analyzed by running parallel programs on a 32-node transputer based parallel machine designed and developed at the IBM Research Center at Yorktown Heights, New York. Benchmark queries of different semantic complexities are generated, and their performance is analyzed for various data and query parameters     

Revisiting the cube lifecycle in the presence of hierarchies

On-line analytical processing (OLAP) typically involves complex aggregate queries over large datasets. The data cube has been proposed as a structure that materializes the results of such queries in order to accelerate OLAP. A significant fraction of the related work has been on Relational-OLAP (ROLAP) techniques, which are based on relational technology. Existing ROLAP cubing solutions mainly focus on “flat” datasets, which do not include hierarchies in their dimensions. Nevertheless, as shown in this paper, the nature of hierarchies introduces several complications into the entire lifecycle of a data cube including the operations of construction, storage, indexing, query processing, and incremental maintenance. This fact renders existing techniques essentially inapplicable in a significant number of real-world applications and mandates revisiting the entire cube lifecycle under the new perspective. In order to overcome this problem, the CURE algorithm has been recently proposed as an efficient mechanism to construct complete cubes over large datasets with arbitrary hierarchies and store them in a highly compressed format, compatible with the relational model. In this paper, we study the remaining phases in the cube lifecycle and introduce query-processing and incremental-maintenance algorithms for CURE cubes. These are significantly different from earlier approaches, which have been proposed for flat cubes constructed by other techniques and are inadequate for CURE due to its high compression rate and the presence of hierarchies. Our methods address issues such as cube indexing, query optimization, and lazy update policies. Especially regarding updates, such lazy approaches are applied for the first time on cubes. We demonstrate the effectiveness of CURE in all phases of the cube lifecycle through experiments on both real-world and synthetic datasets. Among the experimental results, we distinguish those that have made CURE the first ROLAP technique to complete the construction and usage of the cube of the highest-density dataset in the APB-1 benchmark (12 GB). CURE was in fact quite efficient on this, showing great promise with respect to the potential of the technique overall.     

无须附加空间的数据立方体联机聚集

以往在数据立方体上实现的联机聚集往往需要附加空间来存储联机聚集估算所需要的信息,极大地影响了数据立方体的存储和维护性能.提出了基于QC-Tree的用于范围查询处理的联机聚集PE(progressively estimate)算法以及它与简单聚集算法相结合的混合聚集算法HPE(hybrid progressively estimate);还提出了一种能够同时处理多个范围查询的联机聚集算法MPE(multiple progressively estimate).与以往联机聚集算法不同,这些算法不需要任何附加空间,而是利用QC-Tree自身保存的聚集数据和语义关系来估算聚集结果.由于QC-Tree是一种极为高效的数据立方体存储结构,因此能够以较理想的性能实现数据立方体上的联机聚集.对算法的分析和实验结果表明,所提出的算法具有较好的性能.     

Distributed snapshot maintenance in wide-column NoSQL databases using partitioned incremental ETL pipelines

Wide-column NoSQL databases are an important class of NoSQL (Not only SQL) databases which scale horizontally and feature high access performance on sparse tables. With current trends towards big Data Warehouses (DWs), it is attractive to run existing business intelligence/data warehousing applications on higher volumes of data in wide-column NoSQL databases for low latency by mapping multidimensional models to wide-column NoSQL models or using additional SQL add-ons. For examples, applications like retail management can run over integrated data sets stored in big DWs or in the cloud to capture current item-selling trends. Many of these systems also employ Snapshot Isolation (SI) as a concurrency control mechanism to achieve high throughput for read-heavy workloads. SI works well in a DW environment, as analytical queries can now work on (consistent) snapshots and are not impacted by concurrent update jobs performed by online incremental Extract-Transform-Load (ETL) flows that refresh fact/dimension tables. However, the snapshot made available in the DW is often stale, since at the moment when an analytical query is issued, the source updates (e.g. in a remote retail store) may not have been extracted and processed by the ETL process in time due to high input data volume or slow processing speed. This staleness may cause incorrect results for time-critical decision support queries. To address this problem, snapshots which are supposed to be accessed by analytical queries need to be first maintained by corresponding ETL flows to reflect source updates based on given freshness needs. Snapshot maintenance in this work means maintaining the distributed data partitions that are required by a query. Since most NoSQL databases are not ACID compliant and do not provide full-fledged distributed transaction support, snapshot may be inconsistently derived when its data partitions are updated by different ETL maintenance jobs.This paper describes an extended version of HBelt system [1] which tightly integrates the wide-column NoSQL database HBase with a clustered & pipelined ETL engine. Our objective is to efficiently refresh HBase tables with remote source updates while a consistent snapshot is guaranteed across distributed partitions for each scan request in analytical queries. A consistency model is defined and implemented to address so-called distributed snapshot maintenance. To achieve this, ETL jobs and analytical queries are scheduled in a distributed processing environment. In addition, a partitioned, incremental ETL pipeline is introduced to increase the performance of ETL (update) jobs. We validate the efficiency gain in terms of data pipelining and data partitioning using the TPC-DS benchmark, which simulates a modern decision support system for a retail product supplier. Experimental results show that high query throughput can be achieved in HBelt when distributed, refreshed snapshots are demanded.     

Efficient aggregation algorithms for compressed data warehouses

Aggregation and cube are important operations for online analytical processing (OLAP). Many efficient algorithms to compute aggregation and cube for relational OLAP have been developed. Some work has been done on efficiently computing cube for multidimensional data warehouses that store data sets in multidimensional arrays rather than in tables. However, to our knowledge, there is nothing to date in the literature describing aggregation algorithms on compressed data warehouses for multidimensional OLAP. This paper presents a set of aggregation algorithms on compressed data warehouses for multidimensional OLAP. These algorithms operate directly on compressed data sets, which are compressed by the mapping-complete compression methods, without the need to first decompress them. The algorithms have different performance behaviors as a function of the data set parameters, sizes of outputs and main memory availability. The algorithms are described and the I/O and CPU cost functions are presented in this paper. A decision procedure to select the most efficient algorithm for a given aggregation request is also proposed. The analysis and experimental results show that the algorithms have better performance on sparse data than the previous aggregation algorithms     

Identification- and Elimination-Based Parallel Query Processing Techniques for Object-Oriented Databases

The newly developed object-oriented database management systems provide rich facilities for the modeling and processing of structural as well as behavioral properties of complex application objects. However, due to their inherent generality, new functionalities to be added to these systems as they continue to evolve, and high performance demand in many application domains, efficient parallel algorithms and architectures would be needed to meet the performance requirement for processing large OODBs. In our previous work, we have shown that processing OODBs can be viewed as the manipulation of patterns of object associations. In this paper, we present several parallel, multiwavefront algorithms based on two approaches, i.e., identification and elimination approaches, to verify association patterns specified in queries. Both approaches allow more processors to operate concurrently on a query than the traditional tree-structured query processing approach, thus introducing a higher degree of parallelism in query processing. We present a graph model to transform the query processing problem into a graph problem. Based on the graph model, proofs of correctness of both approaches for tree-structured queries are given, and a combined approach for solving cyclic queries is also provided. We present a new data structure to represent associations between objects, parallel algorithms based on these approaches, and some evaluation results obtained from an actual implementation of these algorithms on an nCUBE 2 parallel computer.     

Tuning a parallel database algorithm on a shared-memory multiprocessor

Database query processing can benefit significantly from parallelism. Parallel database algorithms combine substantial CPU and I/O activity, memory requirements, and massive data exchange between processes, all of which must be considered to obtain optimal performance. Since parallel external sorting is a very typical example, we have focused on sorting to tune Volcano, a new query processing system. The purpose of the Volcano project is to provide efficient, extensible tools for query and request processing in novel application domains, particularly in object-oriented and scientific database systems, and for experimental database performance research. It includes all query processing algorithms conventionally used in relational database systems as well as several new ones, and can execute all of them in parallel. In this article, we present Volcano's parallel external sorting algorithm and a sequence of enhancements to improve its performance. We obtained very good absolute performance, 84 seconds for 100 MB of data, as well as near-linear speedup with sixteen CPUs and disks. Furthermore, these results were achieved on a shared-memory machine despite the common belief that parallel query processing is best implemented on distributed-memory systems. We detail our tuning measures and report on their effectiveness.     

一种并行处理Skyline查询的有效方法

Skyline查询是近年来数据库领域的一个研究重点和热点, 这主要是因为Skyline查询在许多领域有着广泛的应用. 现有的工作大都集中于单处理机环境, 然而, 由于Skyline查询是CPU敏感的, 因此,在实际应用中, 现有的方法具有很大的局限性. 基于此, 提出一种有效降低处理Skyline查询时间开销的并行算法PAPSQ (Parallel algorithm for processing skyline queries). 算法有机结合多维数据对象的自身特性和通用多处理机系统的实施优点, 以Skyline查询搜索偏序格为底层结构, 利用多维数据对象的同胚评估值和偏序格加权技术来有效提高并行处理Skyline查询的效率. 实验评估表明, PAPSQ算法具有有效性和实用性.     

Star join revisited: Performance internals for cluster architectures

Data warehouse workloads are crucial for the support of on-line analytical processing (OLAP). The strategy to cope with OLAP queries on such huge amounts of data calls for the use of large parallel computers. The trend today is to use cluster architectures that show a reasonable balance between cost and performance. In such cases, it is necessary to tune the applications in order to minimize the amount of I/O and communication, such that the global execution time is reduced as much as possible.

In this paper, we model and analyze the most up-to-date strategies for ad hoc star join query processing in a cluster of computers. We show that, for ad hoc query processing and assuming a limited amount of resources available, these strategies still have room for improvement both in terms of I/O and inter-node data traffic communication. Our analysis concludes with the proposal of a hybrid solution that improves these two aspects compared to the previous techniques, and shows near optimal results in a broad spectrum of cases.     

Measuring and modelling the performance of a parallel ODMG compliant object database server

Object database management systems (ODBMSs) are now established as the database management technology of choice for a range of challenging data intensive applications. Furthermore, the applications associated with object databases typically have stringent performance requirements, and some are associated with very large data sets. An important feature for the performance of object databases is the speed at which relationships can be explored. In queries, this depends on the effectiveness of different join algorithms into which queries that follow relationships can be compiled. This paper presents a performance evaluation of the Polar parallel object database system, focusing in particular on the performance of parallel join algorithms. Polar is a parallel, shared‐nothing implementation of the Object Database Management Group (ODMG) standard for object databases. The paper presents an empirical evaluation of queries expressed in the ODMG Query Language (OQL), as well as a cost model for the parallel algebra that is used to evaluate OQL queries. The cost model is validated against the empirical results for a collection of queries using four different join algorithms, one that is value based and three that are pointer based. Copyright © 2005 John Wiley & Sons, Ltd.