|
The proliferation of mobile devices and the fact that high-bandwidth
and continuous connectivity is not available everywhere, has led to
the creation of hoarding algorithms that attempt to mitigate the
problems related with disconnected operation and with the operation
in areas where bandwidth is either scarce or expensive. In this
dissertation, a generic hoarding approach for semi-structured
information is introduced that has been specialized and evaluated
for Web page requests. The approach is based on an infrastructure of
infostations that provide comparatively cheap high-speed access to
the Web. It relies on clustering of semantically related data items
by classifying pages into content and transit pages, where this
classification is used to guide the clustering algorithm. It could
be shown shown by means of experimental evaluation that the proposed
hoarding scheme outperforms existing ones by a factor of more than
three in terms of content hit ratio. Moreover, an analysis of energy
consumption of wireless interfaces could show that the proposed
hoarding scheme also allows for substantial energy savings of mobile
devices when accessing data. The performance evaluation of Web
applications usually requires the analysis of sequences of user
requests for specific Web pages. These sequences can be obtained,
for example, by applying empirical methods (recording the real
sequence of requests), or by applying a formal model for generating
synthetic results. In this dissertation, a novel Web Browsing Model
is proposed. By taking into account the hyperlink structure as well
as the different user behavior on the Web, the implementation of
this model allows us to generate accurate synthetic data that can be
used instead of empirically obtained requests. Additionally, it
could be shown using empirical data that the probability of choosing
some hyperlink from a given page as well as the probability of a
user leaving a page without following a hyperlink is best
characterized by a power-law. In order to validate the approach, I
could show the flexibility and applicability of the Web Browsing
Model by performing the required correlations to empirical data.
|