Scalability of Topic Map Systems

Hoyer, Marcel
Abstract in English: 
The purpose of this thesis was to find approaches solving major performance and scalability issues for Topic Maps-related data access and the merging process. Especially regarding the management of multiple, heterogeneous topic maps with different sizes and structures. Hence the scope of the research was mainly focused on the Maiana web application with its underlying MaJorToM and TMQL4J back-end. In the first instance the actual problems were determined by profiling the application runtime, creating benchmarks and discussing the current architecture of the Maiana stack. By presenting different distribution technologies afterwards the issues around a single-process instance, slow data access and concurrent request handling were investigated to determine possible solutions. Next to technological aspects (i. e. frameworks or applications) this discussion included fundamental reflection of design patterns for distributed environments that indicated requirements for changes in the use of the Topic Maps API and data flow between components. With the development of the JSON Topic Maps Query Result format and simple query-focused interfaces the essential concept for an prototypical implementation was established. To concentrate on scalability for query processing basic principles and benefits of message-oriented middleware were presented. Those were used in combination with previous results to create a distributed Topic Maps query service and to present ideas about optimizing virtual merging of topic maps. Finally this work gave multiple insights to improve the architecture and performance of Topic Maps-related applications by depicting concrete bottlenecks and providing prototypical implementations that show the feasibility of the approaches. But it also pointed out remaining performance issues in the persisting data layer.
Pubdate / Erscheinungsdatum: 
Pages / Seitenanzahl: 
index.pdf3.05 MB