In the recent past, there have been rapid advances in the technology
of processors, storage and networks, leading to technologies like
cloud computing. However, amid all these advances, performance of
clouds and cloud services continues to present challenges. Access
latencies to the information on the cloud due to variable bandwidth
continues to be a serious problem of research; more so in environments
requiring mobile devices to stay connected to the cloud. One way to
smooth out bumps in bandwidth available is to use anticipatory
retrieval of data, and to cache data that is likely to be requested
later. The proposed anticipatory retrieval and caching system is a
solution that takes this path. It offers a better experience to those
mobile users who are connected to a cloud and make frequent access to
the cloud's datastore. The proposed method aims to provide ubiquitous
access to data on clouds regardless of the bandwidth levels. This is
done by locally caching all the one-hop related item-sets $I_1, I_2,
\ldots, I_k$ semantically belonging to (or semantically linked to) a
particular item-set $I'$. Caching is done asynchronously in the
background during times of high bandwidth. The proposed algorithms
assess the semantic relevance of the data using semantic distances
along with user priorities and availability of bandwidth, and then
prioritizes anticipatory data downloads on to the cloud's storage
based on the relevance quotient.