Component-based digital library scalability using clusters

Omar, Muammar Zamir

Component-based digital library scalability using clusters

Thesis / Dissertation

2011

Abstract

Digital Libraries (DLs) are systems to manage information or data. They range from monolithic systems to loosely coupled component-based ones. DLs provide the services to manage, retrieve and access this information. Where they have fallen short is providing methods to manage huge volumes of information quickly and effectively. While the services provided by these systems work correctly, the time it takes to provide a response is unacceptable. In many cases, this is due to the underlying architecture of the DL system and other factors which influence resources available to the system owners for upgrades or maintenance. This dissertation documents an alternate approach to the normal one of using more powerful machines to overcome the problem. Instead, a cluster of computers is used to provide increased performance. It presents a DL system in which loosely-coupled components can migrate and replicate across machines in order to meet the demands of the system. These components provide user services such as searching and browsing. Over time, the system adjusts itself automatically to provide better service times as the number of incoming requests increase. These adjustments include migrating components or services to machines with more resources and replicating those which are being queried constantly. The architecture introduced is one which can be created from most component-based DLs and is easily replicated. Initial analysis and evaluation indicate that this system provides better performance under conditions of heavy load while maintaining good response times under minimal loads. This approach has thus proven to be a viable one for addressing performance degradation in an experimental environment and is ready for testing in a live environment.

Keywords

Computer Science

Reference:

Collections

Masters

Full item page