Title: Smart prefetching for mobile users under volatile network conditions
Language: English
Authors: Borkowski, Michael 
Qualification level: Diploma
Keywords: Prefetching; Mobile Netzwerke; Caching; Optimierung; Service-Qualität
prefetching; mobile networks; caching; optimization; quality of service
Advisor: Dustdar, Schahram 
Assisting Advisor: Schulte, Stefan  
Issue Date: 2015
Number of Pages: 119
Qualification level: Diploma
Abstract: 
In the field of mobile distributed computing, the term prefetching is used to describe the practice of retrieving data before its actual usage. Doing so results in several beneficial effects for the user, ranging from reduced perceived delay, the possibility to handling lower connection speeds to masking network outages completely. This thesis proposes a prefetch scheduling algorithm. The described algorithm is capable of scheduling requests so that, in a best-case scenario, the results are fetched exactly on-time and no delay at all is experienced by the user (or client code). After discussing some formal concepts necessary as a foundation for the proposed algorithm, a basic idea on how to approach a scenario of volatile network connectivity with a given context (predicted future network quality) is presented. The idea is first stated as a problem of mathematical analysis, and subsequently transformed into a concrete algorithm, of which a reference implementation stub in Java is shown. The second main section of this thesis is a simulation environment capable of reproducibly evaluating the prefetching scenarios by running simulations of mobile units with fluctuating network quality. This thesis presents the architecture of this simulation environment in detail. Following this description, a series of simulation configurations is shown and the resulting simulation outcomes are presented. The results show that the algorithm indeed increases the overall quality of user experience significantly, in the typical use case modelled after a real-world situation by up to 70 %, while maintaining context constraints like limited energy usage. Concluding, the algorithm is put into context of real-world applications and its feasibility is discussed, alongside with possible future fields of research and concepts of enhancement.
URI: https://resolver.obvsg.at/urn:nbn:at:at-ubtuw:1-78105
http://hdl.handle.net/20.500.12708/7103
Library ID: AC12652596
Organisation: E184 - Institut für Informationssysteme 
Publication Type: Thesis
Hochschulschrift
Appears in Collections:Thesis

Files in this item:

Show full item record

Page view(s)

19
checked on Feb 26, 2021

Download(s)

61
checked on Feb 26, 2021

Google ScholarTM

Check


Items in reposiTUm are protected by copyright, with all rights reserved, unless otherwise indicated.