Ahcene Boukorca

Eco-DMW: Eco-Design Methodology for Data warehouses

In the Big Data Era, the management of energy consumption by servers and data centers has become a challenging issue for companies, institutions, and countries. In data-centric applications, DBMS are one of the major energy consumers when executing complex queries involving very large databases. Some research has been devoted to this issue, covering both the hardware and software dimensions. Regarding software, several proposals have been outlined, focusing either on analytical cost models to predict energy when executing queries or techniques to save energy. To this date, no research has taken account of energy at the physical design level, a crucial phase in database design. In this paper, we propose a methodology, called Eco-DMW, that integrates the energy dimension into the physical design. To show this integration, we study the case of materialized views, a redundant optimization structure. We first show the place that energy takes throughout this stage of design. A multi-objective formalization of the problem of materialized view selection is given. A genetic algorithm is developed to solve the problem. Intensive experiments are conducted using a mathematical cost model and a real measurement tool dedicated to computing energy. Results show the interest of this proposal to save energy and optimize queries in the presence of the selected materialized views.

SONIC: Scalable Multi-query OptimizatioN through Integrated Circuits

In the first generation of databases, query optimizers were designed to optimize individual queries. Due to the predefined number of tables of a given database, the probability to have interaction between queries is high. As a consequence, optimizers propose solutions for multi-queries optimization. Getting this optimization is known as NP-hard problem. To ensure a scalable solution, we borrow techniques used in the electronic design automation EDA domain. In this paper, we first make an analogy between the multi-query optimization problem and the EDA domain. Secondly, we propose to model our problem with hypergraphs massively used to design and test integrated circuits. Thirdly, we use our results to materialize views. Finally, experiments are conducted to show the scalability of our approach.

Coupling Materialized View Selection to Multi Query Optimization: Hyper Graph Approach

Materialized views are queries whose results are stored and maintained in order to facilitate access to data in their underlying base tables of extremely large databases. Selecting the best materialized views for a given query workload is a hard problem. Studies on view selection have considered sharing common sub expressions and other multi-query optimization techniques. Multi-Query Optimization is a well-studied domain in traditional and advanced databases. It aims at optimizing a workload of queries by finding and reusing common sub-expression between queries. Finding the best shared expression is known as a NP-hard problem. The shared expressions usually identified by graph structure have been used to be candidate for materialized views. This shows the strong interdependency between the problems of materialized view selection (PVS) and multi query optimization (PMQO), since the PVS uses the graph structure of the PMQO. Exploring the existing works on PVS considering the interaction between PVS and PMQO figures two main categories of studies: (i) those considering the PMQO as a black box where the output is the graph and (ii) those preparing the graph to guide the materialized view selection process. In this category, the graph generation is based on individual query plans, an approach that does not scale, especially with the explosion of Big Data applications requiring large number of complex queries with high interaction. To ensure a scalable solution, this work proposes a new technique to generate a global processing plan without using individual plans by borrowing techniques used in the electronic design automation (EDA) domain. This paper first presents a rich state of art regarding the PVS and a classification of the most important existing work. Secondly, an analogy between the MQO problem and the EDA domain, in which large circuits are manipulated, is established. Thirdly, it proposes to model the problem with hypergraphs which are massively used to design and test integrated circuits. Fourthly, it proposes a deterministic algorithm to select materialized views using the global processing plan. Finally, experiments are conducted to show the scalability of our approach.

MapReduce-DBMS: An Integration Model for Big Data Management and Optimization

The data volume and the multitude of sources have an exponential number of technical and application challenges. In the past, Big Data solutions have been presented as a replacement for the Parallel Database Management Systems. However, Big Data solutions can be seen as a complement to a RDBMS for analytical applications, because different problems require complex analysis capabilities provided by both technologies. The aim of his work is to integrate a Big Data solution and a classic DBMS, in a goal of queries optimization. We propose a model for OLAP queries process. Then, we valid the proposed optimized model through experiments showing the gain of the execution cost saved up.

What-if Physical Design for Multiple Query Plan Generation

The multiple query optimization problem (MQO) has been largely studied in traditional and advanced databases. In scientific and statistical databases, queries are complex, recurrent and share common sub-expressions. As a consequence, the MQO problem re-emerges in this context. An important characteristic of the MQO problem is that its result (usually represented by a unified plan merging isolated plans of the workload queries) may be used to select optimization structures (OS) such as materialized views. By examining the literature, we discover that the interconnection between these two problems is often neglected. Ignoring what-if questions about selecting the unified plan can result in disastrous consequences when used as a basis for selecting OS. In this paper, we first exhibit the link between global plans and optimization structures. Secondly, we give a formalization of the OS-oriented unified plan generation problem. Thirdly, a generic approach for plan generation is given. Finally, we instantiate our formalization to deal with the problems of selecting materialized views and horizontal data partitioning and we show its effectiveness and efficiency through intensive experiments.

HYPAD: Hyper-Graph-Driven Approach for Parallel Data Warehouse Design

Small, medium and large companies all face three well-identified problems, precisely: (i) the data deluge, (ii) the large number of interacted exploratory queries and (iii) the economic crisis. Hence, it becomes a real necessity to consider those problems and develop low-cost database deployment solutions. Data parallel architectures are one of the relevant deployment platforms that may manage efficiently this deluge of data. The process of designing such architecture has to integrate the interaction that may exist between queries. Although, the state-of-art on parallel data warehouses is quite rich, to the best of our knowledge, the query interaction is not highlighted. Amazingly, the queries are in the core of the parallel design. Ignoring their interaction may impact the quality of the final design. In this paper, we propose a new scalable hyper-graph approach, called HYPAD, for designing cluster warehouses by considering concurrent analytical highly interacted queries. Our approach is validated through a data warehouse cluster simulator. The obtained results show the effectiveness and efficiency of our proposal.

BINOS4DNA: Bitmap Indexes and NoSQL for Identifying Species with DNA Signatures through Metagenomics Samples

The advancement of next generation sequencing (NGS) and shotgun sequencing technologies produced massive amounts of genomics data. Metagenomics, a powerful technique to study genetic material of uncultivable microorganisms received directly from their natural environment, is dealing with high throughput sequencing read data sets. Assembling, binning and alignment of short reads in order to identify microorganisms of a Metagenomics sample are expensive and time- consuming, regardless of other restrictions. DNA signature is a short nucleotide sequence fragment which is used to distinguish species across all other species. It can be a basis for identifying microorganisms both in environmental and clinical samples directly from the short reads, without assembling and alignment processes. In this paper, we propose a scalable method in which we use optimization techniques borrowed from database technology, namely bitmap indexes. They are used to speed up searching and matching of billions of DNA signatures in the short reads of thousands of different microorganisms, using commodity High Performance Computing, such as Hadoop MapReduce, Hive and Hbase.