Amr El Abbadi

An efficient, fault-tolerant protocol for replicated data management

A data management protocol for executing transactions on a replicated database is presented. The protocol ensures one-copy serializability. i.e., the concurrent execution of transactions on a replicated database is equivalent to some serial execution of the same transactions on a non-replicated database. The protocol tolerates a large class of failures, including: processor and communication link crashes, partitioning of the communication network, lost messages, and slow responses of processors and communication links. Processor and link recoveries are also handled. The protocol implements the reading of a replicated object efficiently by reading the nearest available copy of the object. When reads outnumber writes, the protocol performs better than other known protocols.

Storage and Retrieval of Moving Objects

We investigate the problem and provide a data model storing, indexing, and retrieving future locations of moving objects in an efficient manner. Each moving object has four independent variables which allow us to predict its future location: a starting location, a destination, a starting time, and an initial velocity. To understand the underlying complexity of the problem, we investigate and categorize the configurations where two variables can vary. Based on that understanding, we choose a configuration which is to some extent restrictive, but still can be used in a wide variety of realistic settings. A performance study shows that our model has much less overhead in processing range queries compared to other proposed approaches.

The Dynamic Data Cube

Range sum queries on data cubes are a powerful tool for analysis. A range sum query applies an aggregation operation (e.g., SUM, AVERAGE) over all selected cells in a data cube, where the selection is specified by providing ranges of values for numeric dimensions. We present the Dynamic Data Cube, a new approach to range sum queries which provides efficient performance for both queries and updates, which handles clustered and sparse data gracefully, and which allows for the dynamic expansion of the data cube in any direction.

Flexible Data Cubes for Online Aggregation

Applications like Online Analytical Processing depend heavily on the ability to quickly summarize large amounts of information. Techniques were proposed recently that speed up aggregate range queries on MOLAP data cubes by storing pre-computed aggregates. These approaches try to handle data cubes of any dimensionality by dealing with all dimensions at the same time and treat the different dimensions uniformly. The algorithms are typically complex, and it is difficult to prove their correctness and to analyze their performance. We present a new technique to generate Iterative Data Cubes (IDC) that addresses these problems. The proposed approach provides a modular framework for combining one-dimensional aggregation techniques to create space-optimal high-dimensional data cubes. A large variety of cost tradeoffs for high-dimensional IDC can be generated, making it easy to find the right configuration based on the application requirements.

Selectivity Estimation for Spatial Joins with Geometric Selections

Spatial join is an expensive operation that is commonly used in spatial database systems. In order to generate efficient query plans for the queries involving spatial join operations, it is crucial to obtain accurate selectivity estimates for these operations. In this paper we introduce a framework for estimating the selectivity of spatial joins constrained by geometric selections. The center piece of the framework is Euler Histogram, which decomposes the estimation process into estimations on vertices, edges and faces. Based on the characteristics of different datasets, different probabilistic models can be plugged into the framework to provide better estimation results. To demonstrate the effectiveness of this framework, we implement it by incorporating two existing probabilistic models, and compare the performance with the Geometric Histogram [1] and the algorithm recently proposed by Mamoulis and Papadias [2].

FATES: Finding A Time dEpendent Shortest path

We model a moving object as a sizable physical entity equipped with GPS, wireless communication capability, and a computer. Based on a grid model, we develop a distributed system, FATES, to manage data for moving objects in a two-dimensional space. The system is used to provide time-dependent shortest paths for moving objects. The performance study shows that FATES yields shorter average trip time when there is a more congested route than any other routes in the domain space.

Space-Efficient Data Cubes for Dynamic Environments

Data cubes provide aggregate information to support the analysis of the contents of data warehouses and databases. An important tool to analyze data in data cubes is the range query. For range queries that summarize large regions of massive data cubes, computing the query result on-the-fly can result in non-interactive response times. To speed up range queries, values that summarize regions of the data cube are pre-computed and stored. This faster response time results in more expensive updates and/or space overhead. While the emphasis is typically on low query and update costs, growing data collections increase the demand for space-efficient approaches. In this paper two techniques are presented that have the same update and query costs as earlier approaches, without introducing any space overhead.

Accessing Scientific Data: Simpler is Better

A variety of index structures has been proposed for supporting fast access and summarization of large multidimensional data sets. Some of these indices are fairly involved, hence few are used in practice. In this paper we examine how to reduce the I/O cost by taking full advantage of recent trends in hard disk development which favor reading large chunks of consecutive disk blocks over seeking and searching. We present the Multiresolution File Scan (MFS) approach which is based on a surprisingly simple and flexible data structure which outperforms sophisticated multidimensional indices, even if they are bulk-loaded and hence optimized for query processing. Our approach also has the advantage that it can incorporate a priori knowledge about the query workload. It readily supports summarization using distributive (e.g., count, sum, max, min) and algebraic (e.g., avg) aggregate operators.

Optimal allocation of two-dimensional data (Extended abstract)

Efficient browsing and retrieval of geographically referenced information requires the allocation of data on different storage devices for concurrent retrieval. By dividing a two dimensional space into tiles, a system can allow users to specify regions of interest using a query rectangle and then retrieving all information related to tiles overlapping with the query. In this paper, we derive the necessary and sufficient conditions for strictly optimal allocations of two-dimensional data. These methods, when they exist, guarantee that for any query, the minimum number of tiles are assigned the same storage device, and hence ensures maximal retrieval concurrency.

A Java-Based Framework for Processing Distributed Objects

The Alexandria Digital Library Project at UC Santa Barbara has been building an information retrieval system for geographically referenced information and datasets. To meet these requirements, we have designed a distributed Data Store to store its holdings. The librarys map, image and geographical data are viewed as a collection of objects with evolving roles. Developed in the Java programming language and the HORB distributed object system, the Data Store manages these objects for flexible and scalable processing. To implement the Data Store we provide a messaging layer that allows applications to distribute processing between the Data Store and the local host. We define a data model for Data Store repositories that provide Client access to Data Store objects. We finally provide support for specialized views of these Data Store items.