|Home \ Graduation Activities \ Post-Graduation Page||Login|
Exploring Data-Access Patterns for Optimizing Large-Scale Applications
Large-scale applications are characterized by elaborate logic and considerable volume of data to be processed. They may be distributed, both logically and physically. Their development and maintenance are complex activities, demanding the attention of teams of qualified individuals. Good data locality is essential if an application is to perform well, but achieving it requires a deep knowledge about how the system operates. When done properly, this may allow the proper distribution, sharing, and loading/unloading of data, leading to an optimized system performance with minimal overhead. The scale, complexity, and constant evolution of applications make it practically impossible to identify manually the appropriate course of action for achieving good data locality. The principal objective of this dissertation is the development of automatic techniques capable of identifying, with a high degree of confidence, the data access patterns performed by object-oriented applications and, subsequently, of taking suitable measures for improving the performance and efficiency displayed by those applications.
Post-Graduation Student / Researcher / Professor: