Fundamental Techniques for Incentive-aware, Efficient,
and Reliable Cloudlet Management and Services
The goal of this work is to develop some fundamental
techniques that can facilitate efficient management of cloudlet resources for
high cloud service performance while minimizing the system cost. These
techniques are to enable efficient, reliable and low cost mobile cloud
services, which will in turn support high-performance wireless applications in
an energy- and resource efficient manner. With this research, a cloud service
provider can exploit sparse signal processing technique to remotely monitor one
or multiple cloudlet clusters running in different location sites with high
accuracy but low cost. To reduce the service initialization delay and
operational cost, the system can cache application components, and a user can
exploit the proposed rate dependent Bloom filter (RDBF) to quickly determine of
some cloudlets have its required application components cached. Rather than
being constrained by a limited number of options for virtual machine (VM)
configurations, the system will be equipped with multi-dimensional pricing and
resource negotiation capabilities for a user to flexibly negotiate multiple
types of resources to maximize the benefits of both users and cloudlets while
taking into account user application preference, performance requirements, and
the system resource conditions.
Complementary to the research agenda, the project will carry out a broad range
of education and outreach activities and facilitate technology transfer through
the industrial partners and industry affiliate programs. The proposed research
is part of a global effort in exploiting (mobile) cloud computing for advancing
some critical fields for national economy and security, including health-care,
transportation, smart-grid, homeland security, and education.
.