Topics: Network analysis and design; Optimization techniques; Traffic engineering
Authors: Mirza Mohd Shahriar Maswood (University of Missouri-Kansas City, USA); Chris Develder (Ghent University - iMinds, Belgium); Edmundo Madeira (State University of Campinas, Brazil); Deep Medhi (University of Missouri-Kansas City, USA)
Presenter bio: Mirza Mohd Shahriar Maswood is a PhD student at University of
Missouri-Kansas City (UMKC) in the Department of Computer Science
Electrical Engineering. His major is Telecommunications and Computer
Networking and co-major is Electrical and Computer Engineering.
Currently, he is doing research in the energy optimization of
data-center network under the supervision of Dr. Deep Medhi. Prior to
beginning the PhD program, he worked as a lecturer in Khulna University
of Engineering and Technology, Khulna, Bangladesh. He achieved his
Master of Science in Electrical Engineering from University of
Missouri-Kansas City in May, 2015 and Bachelor of Science in Electronics
and Communication Engineering in April, 2010. He recently got School of
Graduate Studies Research Grant, Preparing Future Faculty Award and
Interdisciplinary Applied Math Fellowship.
Abstract: For cloud enterprise customers that require services on demand, data
centers must
allocate and partition data center resources in a dynamic fashion. We
consider the problem in which a request from an enterprise customer is
mapped to a virtual network (VN) that is allocated requiring both
bandwidth and compute resources by connecting it from an entry point of
the datacenter to one or more servers, should this data center be
selected from multiple geographically distributed data centers.
We present a dynamic traffic engineering framework, for which we
develop an optimization model based on mixed-integer linear programming
(MIP) formulation
that a data center operator can use is at each review point to optimally
assign VN customers.
Through a series of studies, we then present results on how different VN
customers are treated in terms of request acceptance when each VN class
have different resource requirement. We found that a VN class with low
resource requirement has a low blocking even in heavy traffic , while
the VN class with high resource requirement faces high service denial.
On the other hand, cost for the VN with the highest resource requirement
is not always the highest in the heavy traffic because of very high
service denial faced by this VN class.
