Mobile access to grid knowledge discovery services

Abstract

This master's thesis presents the concept and implementation of the prototype MobileGridMiner that extends the Grid knowledge discovery system GridMiner in terms of mobile access. A number of technologies are addressed in this work. The GridMiner is an infrastructure based on Grid technologies, like the Open Grid Service Architecture (OGSA). It combines data mining and On-Line Analytical Processing (OLAP) technologies to analyze scientific data. Software development frameworks for mobile devices are further important technologies in this work. To address robustness and situated-ness context-awareness is introduced and used in particular to achieve fault tolerance. This work gives an overview of recent mobile Grid projects. We further present various integration modes of mobile devices into Grid systems.<br />These modes classify mobile Grid systems and we investigate the integration modes feasible for Grid knowledge discovery systems like the GridMiner. Hereby, potentials of mobile computing are increased by a concept that deals with different contexts of mobile users. Parameters that specify the context of mobile users are mapped to specific scenarios, e.g.<br />working at the office or cycling. A rule engine uses the current sensed values of the context parameters to derive the scenarios. For evaluation purpose, we define five scenarios and discuss the usage of the prototype in these contexts. The prototype itself is a mobile client extending the GridMiner. The program is based on the Microsoft .NET Compact Framework and written in C#. It runs on a Windows Mobile 2003 Pocket PC. The prototype accesses the GridMiner services and its data over a C# Web service adapter. The MobileGridMiner allows mobile users to access basic GridMiner functionality as well as special new mobile functionalities. Mobile users can be notified about service data changes on the GridMiner by audio feedback or by e-mail. Other functionalities are requesting a GridMiner status, viewing OLAP data, and displaying graphics of GridMiner workflows. Context sensitive functionality is added to the prototype by observing the connection quality of the wireless network.<br />All these features constantly consider the mobile nature of the user and enable him or her to stay informed about GridMiner services.