Roodyn, Neil; (1998) An architecture for integrating multiple real time data feeds. Doctoral thesis (Ph.D), UCL (University College London).

Text An_architecture_for_integratin.pdf Download (10MB)

Abstract

This thesis investigates 'industry strength' solutions for integrating multiple real time data feeds. It addresses one of the major problems challenging computing namely 'information overload'. The thesis focuses on those problems faced by a user of multiple data sources, either real time data feeds or databases. The thesis' goal is a system architecture providing a generic interface which allows the connection of one or more data sources to a unified collection point. On entering the system the data is screened and filtered for information which is pertinent to the end user. This data is then available for other programs that use the data, through a standard interface. The thesis comprises four parts: a critique of existing commercial real time products, and then the invention of three experimental real time systems, constituting the core of this research. These systems have been created in collaboration with industry, providing unique end user feed back, and also ensuring a degree of compliance with industry standards. The three experimental systems comprise: (i) a 16 bit real time system for providing private investors with financial information; (ii) a 32-bit system for tracking buses in real time; (iii) a 32 bit 'generic' real time system which caters for multiple feeds. The three systems have all been designed and implemented by the author, and the systems are now being tested in a commercial environment. The critique examines several commercial real time systems and analyses their key features relating to information overload. It studies the fundamental aspects of real lime data processing with specific reference to the industry standard Microsoft Windows in the PC environment. The second part describes the design and implementation of a 16-bit real time system for providing end users with stock exchange data from one or more data sources. To provide experimental data on the use of the system it was built in collaboration with Updata Software. It utilises 16 bit DLLs for multiple process information sharing and providing data in a 'soft' real time manner. The third part describes the design and implementation of a 32-bit real time system. This system provides feedback on vehicle positions and with the aid of Hampshire County Council is being tested for passenger information systems on buses. It utilises shared global memory blocks to share information between processes. The final part presents the new 'generic' system architecture, whose design draws on experience gained from the previous two systems and the feedback, provided from users. Lessons learnt from previous systems include the need for integration of filtering into the system and providing an open interface to ease data input from new data feeds. This system also addresses the need for compliance with industry standards. It makes strong use of object architectures and, being WOSA compliant, provides COM interfaces to satisfy the need to share information and provide an open interface. The two chapters covering this work contain the design of the system and the strategy employed to implement the design, and then test the system. The thesis and the software implemented make three contributions to science: i) it reduces information overload by integrating the data from multiple sources and providing a single interface for accessing that data, ii) it provides techniques for bringing real time data to the personal computer desktop, and iii) it helps to personalise the data by providing a simple filtering mechanism and an interface for more complex filtering systems. An innovation of this thesis is that the implementations of the designs have been taken through to industry products, in order to provide feedback.

Download activity - last month

Export as CSVXMLJSON

Download activity - last 12 months

Export as JSONXMLCSV

Downloads by country - last 12 months

Export as CSVXMLJSON

Archive Staff Only

View Item