The Second International Conference on Dependability DEPEND 2009 June 18-23, 2009 - Athens/Glyfada, Greece

Tutorials

Using the AMBER Data Repository to Analyze, Share and Cross-exploit Dependability Data

by Marco Vieira and Henrique Madeira, University of Coimbra, Portugal

The AMBER Data Repository (ADR) is an infrastructure based in well-known business intelligence technology that provides powerful tools and techniques to analyze, share and cross-exploit results from resilience evaluation experiments. The analysis of resilience data is a complex and time-consuming task and the approach used by most research teams is far from being satisfactory. In fact, research teams typically use spreadsheets and other specific tools to analyze data, which are normally quite limited and strongly coupled to a very specific environment.

The AMBER data repository (ADR) is an infrastructure based in well-known business intelligence techniques and tools. The central idea is to collect data from resilience evaluation experiments and store it in a multidimensional data structure (data warehouse) supported by a relational database engine. The data analysis is carried out using three types of data analysis techniques, depending on the raw data nature and on the specific analysis needs: On-Line Analytical Processing (OLAP), Data Mining (DM), and Information Retrieval (IR). OLAP tools are traditionally used in business decision support analysis and are the basic type of tools for the analysis of multidimensional raw data. Data mining algorithms are employed to automatically identify which infrastructural factors were really relevant for the results obtained. Information too rich to be represented by relational databases is represented as text. This information can be explored using information retrieval technologies.

The goal of the AMBER Data Repository (http://www.amber-project.eu) is to become a worldwide repository for resilience assessment related data. It is a powerful tool in three different scenarios: 1) at research team level to perform the analysis of data in a very efficient way; 2) at project level to share and cross-exploit results obtained by the different teams; and 3) world wide in the form of a common repository to store and share experimental resilience evaluation data.

The goal of this tutorial, intended for researchers working in experimental dependability evaluation, is to demonstrate the power of business intelligence technologies for the analysis of dependability related data and discuss how researchers can successfully use the ADR to increase the impact of their research work. The major steps required to use the repository will be presented and demonstrated by means of an actual example.

The intended audience is researchers working in experimental dependability evaluation interested in learning new technologies to analyze and share field data and results from dependability evaluation experiments. The preliminary outline is as follows:

1. Introduction
2. Concepts on Business Intelligence
3. Using Business Intelligence Technology to Analyze Dependability Related Data
4. The AMBER Data Repository
5. A Case Study on Using the AMBER Data Repository for Experimental Data Analysis
6. Closure