The convergence behavior of many distributed machine learning (ML) algorithms can be sensitive to the number of machines being used or to changes in the computing environment. As a result, scaling to a large number of machines can be challenging. In this paper, we describe a new scalable coordinate descent (SCD) algorithm for generalized linear models whose convergence behavior is always the same, regardless of how much SCD is scaled out and regardless of the computing environment. This makes SCD highly robust and enables it to scale to massive datasets on low-cost commodity servers. Experimental results on a real advertising dataset in Google are used to demonstrate SCD's cost effectiveness and scalability. Using Google's internal cloud, we show that SCD can provide near linear scaling using thousands of cores for 1 trillion training examples on a petabyte of compressed data. This represents 10,000x more training examples than the 'large-scale' Netflix prize dataset. We also show that SCD can learn a model for 20 billion training examples in two hours for about $10.

MapReduce has emerged as a promising architecture for large scale data analytics on commodity clusters. The rapid adoption of Hive, a SQL-like data processing language on Hadoop (an open source implementation of MapReduce), shows the increasing importance of processing structured data on MapReduce platforms. MapReduce offers several attractive properties such as the use of low-cost hardware, fault-tolerance, scalability, and elasticity. However, these advantages have required a substantial performance sacrifice.

Users of MapReduce often run into performance problems when they scale up their workloads. Many of the problems they encounter can be overcome by applying techniques learned from over three decades of research on parallel DBMSs. However, translating these techniques to a Map-Reduce implementation such as Hadoop presents unique challenges that can lead to new design choices. This paper describes how column-oriented storage techniques can be incorporated in Hadoop in a way that preserves its popular programming APIs. We show that simply using binary storage formats in Hadoop can provide a 3x performance boost over the naive use of text files. We then introduce a column-oriented storage format that is compatible with the replication and scheduling constraints of Hadoop and show that it can speed up MapReduce jobs on real workloads by an order of magnitude. We also show that dealing with complex column types such as arrays, maps, and nested records, which are common in MapReduce jobs, can incur significant CPU overhead. Finally, we introduce a novel skip list column format and lazy record construction strategy that avoids deserializing unwanted records to provide an additional 1.5x performance boost. Experiments on a real intranet crawl are used to show that our column-oriented storage techniques can improve the performance of the map phase in Hadoop by as much as two orders of magnitude.

Spinnaker is an experimental datastore that is designed to run on a large cluster of commodity servers in a single datacenter. It features key-based range partitioning, 3-way replication, and a transactional get-put API with the option to choose either strong or timeline consistency on reads. This paper describes Spinnaker's Paxos-based replication protocol. The use of Paxos ensures that a data partition in Spinnaker will be available for reads and writes as long a majority of its replicas are alive. Unlike traditional master-slave replication, this is true regardless of the failure sequence that occurs. We show that Paxos replication can be competitive with alternatives that provide weaker consistency guarantees. Compared to an eventually consistent datastore, we show that Spinnaker can be as fast or even faster on reads and only 5% to 10% slower on writes.

The MapReduce framework is increasingly being used to analyze large volumes of data. One important type of data analysis done with MapReduce is log processing, in which a click-stream or an event log is filtered, aggregated, or mined for patterns. As part of this analysis, the log often needs to be joined with reference data such as information about users. Although there have been many studies examining join algorithms in parallel and distributed DBMSs, the MapReduce framework is cumbersome for joins. MapReduce programmers often use simple but inefficient algorithms to perform joins. In this paper, we describe crucial implementation details of a number of well-known join strategies in MapReduce, and present a comprehensive experimental comparison of these join techniques on a 100-node Hadoop cluster. Our results provide insights that are unique to the MapReduce platform and offer guidance on when to use a particular join algorithm on this platform.

Modern enterprises must manage uncertain data for purposes of risk assessment and decisionmaking under uncertainty. The Monte Carlo approach embodied in the MCDB system of Jampani et al. is well suited for such a task. MCDB can support industrial strength business-intelligence queries over uncertain warehouse data. Moreover, MCDB's extensible approach to specifying uncertainty can also capture complex stochastic prediction models, allowing sophisticated ``what-if'' analyses within the DBMS. The MCDB computations can be highly CPU intensive, but offer the potential for massive parallelization. To realize this potential, we provide a new system, called MC3 (Monte Carlo Computation on a Cluster), that extends the MCDB approach to the map-reduce processing framework. MC3 can exploit the robustness and scalability of map-reduce, and can handle data stored in non-relational formats. We show how MCDB query plans over ``tuple bundles'' can be translated to sequences of map-reduce operations over nested data, and describe different parallelization schemes. We also provide and analyze several novel distributed algorithms for adding pseudorandom number seeds to tuple bundles. These algorithms ensure statistical correctness of the Monte-Carlo computations while minimizing the seed length. Our experiments show that MC3 can scale well for a variety of workloads.

Inverted indexes have become the standard indexing method for supporting search queries in a variety of content-based applications. Examples of such applications include enterprise document management, e-mail, web search, and social networks. One shortcoming in current inverted index designs is that they support only document-level updates, forcing a full document to be reindexed even if just part of it changes. This paper describes a new inverted index design that enables applications to break a document into semantically meaningful sub-documents or "sections". Each section of a document can be updated separately, but search queries can still work seamlessly across sections. Our index design is motivated by applications where there is metadata associated with each document that tends to be smaller and more frequently updated than the document's content, but at the same time, it is desireable to search the metadata and content with the same index structure. A novel self-optimizing query execution algorithm is described to efficiently join the sections of a document in the inverted index. Experimental results on TREC and patent data are provided, showing that sections can dramatically improve overall system throughput on a mixed workload of updates and queries.

This paper extends traditional faceted search to support richer information discovery tasks over more complex data models. Our first extension adds exible, dynamic business intelligence aggregations to the faceted application, enabling users to gain insight into their data that is far richer than just knowing the quantities of documents belonging to each facet. We see this capability as a step toward bringing OLAP capabilities, traditionally supported by databases over relational data, to the domain of free-text queries over metadata-rich content. Our second extension shows how one can efficiently extend a faceted search engine to support correlated facets - a more complex information model in which the values associated with a document across multiple facets are not independent. We show that by reducing the problem to a recently solved tree-indexing scenario, data with correlated facets can be efficiently indexed and retrieved

A major difference between corporate intranets and the Internet is that in intranets the barrier for users to create web pages is much higher. This limits the amount and quality of anchor text, one of the major factors used by Internet search engines, making intranet search more difficult. The social phenomenon at play also means that spam is relatively rare. Both on the Internet and in intranets, users are often willing to cooperate with the search engine in improving the search experience. These characteristics naturally lead to considering using user feedback to improve search quality in intranets. In this paper we show how a particular form of feedback, namely user annotations, can be used to improve the quality of intranet search. An annotation is a short description of the contents of a web page, which can be considered a substitute for anchor text. We propose two ways to obtain user annotations, using explicit and implicit feedback, and show how they can be integrated into a search engine. Preliminary experiments on the IBM intranet demonstrate that using annotations improves the search quality.

Maintaining strict static score order of inverted lists is a heuristic used by search engines to improve the quality of query results when the entire inverted lists cannot be processed. This heuristic, however, increases the cost of index generation and requires complex index build algorithms. In this paper, we study a new index organization based on static score bucketing. We show that this new technique significantly improves in index build performance while having minimal impact on the quality of search results.

Holistic twig join algorithms represent the state of the art for evaluating path expressions in XML queries. Using inverted indexes on XML elements, holistic twig joins move a set of index cursors in a coordinated way to quickly find structural matches. Because each cursor move can trigger I/O, the performance of a holistic twig join is largely determined by how many cursor moves it makes, yet, surprisingly, existing join algorithms have not been optimized along these lines. In this paper, we describe TwigOptimal, a new holistic twig join algorithm with optimal cursor movement. We sketch the proof of TwigOptimal's optimality, and describe how TwigOptimal can use information in the return clause of XQuery to boost its performance. Finally, experimental results are presented, showing TwigOptimal's superiority over existing holistic twig join algorithms.

We propose a new ranking paradigm for relational databases called Structured Value Ranking (SVR). SVR uses structured data values to score (rank) the results of keyword search queries over text columns. Our main contribution is a new family of inverted list indices and associated query algorithms that can support SVR efficiently in update-intensive databases, where the structured data values (and hence the scores of documents) change frequently. Our experimental results on real and synthetic data sets using BerkeleyDB show that we can support SVR efficiently in relational databases.

Structural joins are a fundamental operation in XML query processing and a large body of work has focused on index-based algorithms for executing them. In this paper, we describe how two well-known index features -- path indices and ancestor information -- can be combined in a novel way to replace one or more of the physical index cursors in a structural join with virtual cursors. The position of a virtual cursor is derived from the path and ancestor information of a physical cursor. Implementation results are provided to show that, by eliminating index I/O, virtual cursors can improve the performance of structural joins by an order of magnitude or more.

Personal webservers have proven to be a popular means of sharing files and peer collaboration. Unfortunately, the transient availability and rapidly evolving content on such hosts render centralized, crawl-based search indices stale and incomplete. To address this problem, we propose YouSearch, a distributed search application for personal webservers operating within a shared context (e.g., a corporate intranet). With YouSearch, search results are always fast, fresh and complete -- properties we show arise from an architecture that exploits both the extensive distributed resources available at the peer webservers in addition to a centralized repository of summarized network state. YouSearch extends the concept of a shared context within web communities by enabling peers to aggregate into groups and users to search over specific groups. In this paper, we describe the challenges, design, implementation and experiences with a successful intranet deployment of YouSearch.

XML is quickly becoming the de facto standard for data exchange over the Internet. This is creating a new set of data management requirements involving XML, such as the need to store and query XML documents. Researchers have proposed using relational database systems to satisfy these requirements by devising ways to "shred" XML documents into relations, and translate XML queries into SQL queries over these relations. However, a key issue with such an approach, which has largely been ignored in the research literature, is how (and whether) the ordered XML data model can be efficiently supported by the unordered relational data model. This paper shows that XML's ordered data model can indeed be efficiently supported by a relational database system. This is accomplished by encoding order as a data value. We propose three order encoding methods that can be used to represent XML order in the relational data model, and also propose algorithms for translating ordered XPath expressions into SQL using these encoding methods. Finally, we report the results of an experimental study that investigates the performance of the proposed order encoding methods on a workload of ordered XML queries and updates.

Many commercial database systems maintain histograms to summarize the contents of relations and permit efficient estimation of query result sizes and access plan costs. Although several types of histograms have been proposed in the past, there has never been a systematic study of all histogram aspects, the available choices for each aspect, and the impact of such choices on histogram effectiveness. In this paper, we provide a taxonomy of histograms that captures all previously proposed histogram types and indicates many new possibilities. We introduce novel choices for several of the taxonomy dimensions, and derive new histogram types by combining choices in effective ways. We also show how sampling techniques can be used to reduce the cost of histogram construction. Finally, we present results from an empirical study of the proposed histogram types used in selectivity estimation of range predicates and identify the histogram types that have the best overall performance. 1 Introduction...

Decision support applications are growing in popularity as more business data is kept on-line. Such applications typically include complex SQL queries that can test a query optimizer's ability to produce an efficient access plan. Many access plan strategies exploit the physical ordering of data provided by indexes or sorting. Sorting is an expensive operation, however. Therefore, it is imperative that sorting is optimized in some way or avoided all together. Toward that goal, this paper describes novel optimization techniques for pushing down sorts in joins, minimizing the number of sorting columns, and detecting when sorting can be avoided because of predicates, keys, or indexes. A set of fundamental operations is described that provide the foundation for implementing such techniques. The operations exploit data properties that arise from predicate application, uniqueness, and functional dependencies. These operations and techniques have been implemented in IBM's DB2/CS.

Modern, high-performance processors employ techniques such as superscalar execution and super-pipelining to increase their instruction issue rate. As the instruction issue rate of processors increases, however, the negative impact of branches on performance also increases. This paper describes an instruction cache (I-cache) prefetch mechanism to improve processor performance on the taken branches. Under perfect conditions, the mechanism allows the target of a taken branch to be prefetched early enough to completely hide the memory latency of an I-cache miss.>

As part of the Starburst extensible database project, the authors have designed and implemented a memory resident storage component that can coexist along side traditional disk-oriented storage components. The memory resident storage component shares the code of Starburst's common services, such as query optimization, plan generation, query evaluation, record manipulation, and transaction management. The design of Starburst's memory resident storage component is discussed, contrasted with Starburst's default disk-oriented storage component, and compared to the performance of the two storage components using the Wisconsin Benchmarks. The results show that a memory resident storage component can perform significantly better than a disk-oriented storage component, even when the disk-oriented storage component has all of its data cached in memory. The benchmark results show that, by using memory resident techniques, overall query execution can be improved by up to a factor of four.>

article Free Access Share on Data caching tradeoffs in client-server DBMS architectures Authors: Michael J. Carey Computer Sciences Department, University of Wisconsin-Madison Computer Sciences Department, University of Wisconsin-MadisonView Profile , Michael J. Franklin Computer Sciences Department, University of Wisconsin-Madison Computer Sciences Department, University of Wisconsin-MadisonView Profile , Miron Livny Computer Sciences Department, University of Wisconsin-Madison Computer Sciences Department, University of Wisconsin-MadisonView Profile , Eugene J. Shekita IBM Almaden Research Center and Computer Sciences Department, University of Wisconsin-Madison IBM Almaden Research Center and Computer Sciences Department, University of Wisconsin-MadisonView Profile Authors Info & Claims ACM SIGMOD RecordVolume 20Issue 2June 1991 pp 357–366https://doi.org/10.1145/119995.115854Published:01 April 1991Publication History 151citation1,074DownloadsMetricsTotal Citations151Total Downloads1,074Last 12 Months100Last 6 weeks4 Get Citation AlertsNew Citation Alert added!This alert has been successfully added and will be sent to:You will be notified whenever a record that you have chosen has been cited.To manage your alert preferences, click on the button below.Manage my AlertsNew Citation Alert!Please log in to your account Save to BinderSave to BinderCreate a New BinderNameCancelCreateExport CitationPublisher SiteeReaderPDF

In this paper we describe three pointer-based join algorithms that are simple variants of the nested-loops, sort-merge, and hybrid-hash join algorithms used in relational database systems. Each join algorithm is described and an analysis is carried out to compare the performance of the pointer-based algorithms to their standard, non-pointer-based counterparts. The results of the analysis show that the pointer-based algorithms can provide significant performance gains in many situations. The results also show that the pointer-based nested-loops join algorithm, which is perhaps the most natural pointer-based join algorithm to consider using in an object-oriented database system, performs quite poorly on most medium to large joins.

The purpose of the Starburst project is to improve the design of relational database management systems and enhance their performance, while building an extensible system to better support nontraditional applications and to serve as a testbed for future improvements in database technology. The design and implementation of the Starburst system to date are considered. Some key design decisions and how they affect the goal of improved structure and performance are examined. How well the goal of extensibility has been met is examined: what aspects of the system are extensible, how extensions can be done, and how easy it is to add extensions. Some actual extensions to the system, including the experiences of the first real customizers, are discussed.>

In this paper we describe how replicated data can be used to speedup query processing in an object-oriented database system. The general idea is to use replicated data to eliminate some of the functional joins that would otherwise be required for query processing. We refer to our technique for replicating data as field replication because it allows individual data fields to be selectively replicated. In the paper we describe how field replication can be specified at the data model level and we present storage-level mechanisms to efficiently support it. We also present an analytical cost model to give some feel for how beneficial field replication can be and the circumstances under which it breaks down. While field replication is a relatively simple notion, the analysis shows that it can provide significant performance gains in many situations.