Differences in fundamental frequencies (F0s) are an important cue for segregating multiple speakers. However, the ability to avail this cue for identification varies with sound levels. For different-and same-F0 conditions, the identification scores of both vowels increased from low- to mid-levels and then reduced at higher levels for younger adults with normal hearing (YNH). These subjects benefited from the F0 difference; however, this benefit varied across the levels. The current study aims to develop a deep-neural-network (DNN) model that can predict the level-dependent changes in concurrent-vowel scores for YNH subjects. This DNN-model includes two-dimensional convolutional neural networks (2D-CNNs) in parallel, with the same architecture, to predict the concurrent-vowel scores. The input layer was neural responses of the auditory-nerve model to concurrent vowels. The 2D-CNN models were trained and validated using the subsets of the concurrent vowels from 50, 65, and 75 dB SPL for both F0 conditions, using the batch gradient descent algorithm. The trained model was then fine-tuned with a single epoch. The 2D-CNN models were evaluated against vowel levels (25 to 85 dB SPL) and F0 conditions. Compared with previous models, the current model accurately predicts the level-dependent changes in concurrent vowel scores for different-and same-F0 conditions.

Modern commercial database systems are increasingly evolving into a hybrid distributed system model where a primary database host system enlists the services of a loosely coupled secondary system that acts as an accelerator. Often the secondary system is a distributed system that can perform specific tasks massively parallelized with results fed back to the host database. Similar models can also be seen in architectures that separate compute from storage. As the scale of the system grows, failures of nodes become common, and the architectural goal is to recover the system with minimal disruption to the workload as seen by the user. This paper introduces a new framework that allows a host database to efficiently manage the availability of a massive secondary distributed system and describes a mechanism to achieve query fault tolerance at the primary database by transparently re-executing query (sub)plans on the secondary distributed system. The focus is on improving two important aspects of disruption ? downtime and transparency to the user. The proposed mechanisms achieve quick recovery, reduced duration of downtime and isolation of errors during query execution, thus improving execution transparency for the users.

Machine learning (ML) is at the forefront of the rising popularity of data-driven software applications. The resulting rapid proliferation of ML technology, explosive data growth, and shortage of data science expertise have caused the industry to face increasingly challenging demands to keep up with fast-paced develop-and-deploy model lifecycles. Recent academic and industrial research efforts have started to address this problem through automated machine learning (AutoML) pipelines and have focused on model performance as the first-order design objective. We present Oracle AutoML, a novel iteration-free AutoML pipeline designed to not only provide accurate models, but also in a shorter runtime. We are able to achieve these objectives by eliminating the need to continuously iterate over various pipeline configurations. In our feed-forward approach, each pipeline stage makes decisions based on metalearned proxy models that can predict candidate pipeline configuration performances before building the full final model. Our approach, which builds and tunes only the best candidate pipeline, achieves better scores at a fraction of the time compared to state-of-the-art open source AutoML tools, such as H2O and Auto-sklearn. This makes Oracle AutoML a prime candidate for addressing current industry challenges.

Today, an ever increasing amount of transistors are packed into processor designs with extra features to support a broad range of applications. As a consequence, processors are becoming more and more complex and power hungry. At the same time, they only sustain an average performance for a wide variety of applications while not providing the best performance for specific applications. In this paper, we demonstrate through a carefully designed modern data processing system called RAPID and a simple, low-power processor specially tailored for data processing that at least an order of magnitude performance/power improvement in SQL processing can be achieved over a modern system running on today's complex processors. RAPID is designed from the ground up with hardware/software co-design in mind to provide architecture-conscious extreme performance while consuming less power in comparison to the modern database systems. The paper presents in detail the design and implementation of RAPID, a relational, columnar, in-memory query processing engine supporting analytical query workloads.

For many years, the highest energy cost in processing has been data movement rather than computation, and energy is the limiting factor in processor design [21]. As the data needed for a single application grows to exabytes [56], there is clearly an opportunity to design a bandwidth-optimized architecture for big data computation by specializing hardware for data movement. We present the Data Processing Unit or DPU, a shared memory many-core that is specifically designed for high bandwidth analytics workloads. The DPU contains a unique Data Movement System (DMS), which provides hardware acceleration for data movement and partitioning operations at the memory controller that is sufficient to keep up with DDR bandwidth. The DPU also provides acceleration for core to core communication via a unique hardware RPC mechanism called the Atomic Transaction Engine. Comparison of a DPU chip fabricated in 40nm with a Xeon processor on a variety of data processing applications shows a 3× - 15× performance per watt advantage.

Oracle RDBMS has supported XML data management for more than six years since version 9i. Prior to 11g, text-centric XML documents can be stored as-is in a CLOB column and schema-based data-centric documents can be shredded and stored in object-relational (OR) tables mapped from their XML Schema. However, both storage formats have intrinsic limitations---XML/CLOB has unacceptable query and update performance, and XML/OR requires XML schema. To tackle this problem, Oracle 11g introduces a native Binary XML storage format and a complete stack of data management operations. Binary XML was designed to address a wide range of real application problems encountered in XML data management---schema flexibility, amenability to XML indexes, update performance, schema evolution, just to name a few. In this paper, we introduce the Binary XML storage format based on Oracle SecureFiles System[21]. We propose a lightweight navigational index on top of the storage and an NFA-based navigational algorithm to provide efficient streaming processing. We further optimize query processing by exploiting XML structural and schema information that are collected in database dictionary. We conducted extensive experiments to demonstrate high performance of the native Binary XML in query processing, update, and space consumption.

XML is being increasingly used in diverse domains ranging from data and application integration to content management. Oracle provides an enterprise wide platform for managing all types of XML content. Within the Oracle database and the application server, the XML content can be efficiently stored using a variety of storage and indexing methods and it can be processed using multiple standard languages within different programmatic environments. 1.

Most commercial SQL database systems support user-defined functions that can be used in WHERE clause filters, SELECT list items, or in sorting/grouping clauses. Often, user-defined functions are used as inexact search filters and then the filtered rows are sorted by a relevance measure. This is commonplace in Web search engines, multimedia, and personalization applications. We refer to the values, such as relevance measure, associated with the filtered rows as ancillary values, and address the problem of efficiently and expressively supporting queries involving them in Oracle. In our approach, the filtering operator is designated as the primary operator, and the associated ancillary values are modeled by additional operators that are declared to be ancillary to the primary operator. An ancillary operator can represent any auxiliary value for the filtered rows, including relevance values (e.g. a score which describes how well a document matches the text search query) and additional properties (e.g. the nature of spatial relationship for objects that overlap a given region). The query execution is optimized by allowing the primary and ancillary operator invocations to share computations via a shared context. Also, queries involving ancillary values can exploit user defined indexes and their capability to return results in the order of ancillary values. We present the key concepts, describes our implementation scheme and optimization techniques, and discusses alternative approaches for supporting ancillary values. Finally, we provide an experimental study that illustrates the scalability and effectiveness of our approach.

Extensible indexing is a SQL-based framework that allows users to define domain-specific indexing schemes, and integrate them into the Oracle8i server. Users register a new indexing scheme, the set of related operators, and additional properties through SQL data definition language extensions. The implementation for an indexing scheme is provided as a set of Oracle Data Cartridge Interface (ODCIIndex) routines for index-definition, index-maintenance, and index-scan operations. An index created using the new indexing scheme, referred to as domain index, behaves and performs analogous to those built natively by the database system. The Oracle8i server implicitly invokes user-supplied index implementation code when domain index operations are performed, and executes user-supplied index scan routines for efficient evaluation of domain-specific operators. This paper provides an overview of the framework and describes the steps needed to implement an indexing scheme. The paper also presents a case study of Oracle Cartridges (intermedia text, spatial, and visual information retrieval), and Daylight (Chemical compound searching) Cartridge, which have implemented new indexing schemes using this framework and discusses the benefits and limitations.

One of the distinctive features of multimedia systems is the wide range of applications they intend to cover, stretching the gamut from entertainment to life-critical applications such as real-time remote surgery. In the face of such a wide spectrum of applications, protocols used to deal with various issues in multimedia systems should be adaptable to the application. Synchronization is one of the key characteristics of a multimedia system. In this paper, we propose a mechanism for synchronization of distributed media data, the accuracy of which can be tailored to the intended application. The model is extended for both stored and live multimedia data. The synchronization mechanism we propose ensures synchronization for all possible temporal relations. The performance of our scheme is evaluated and some implementation issues are discussed.