Topics: Optimization techniques; Green networking; Robust networking
Authors: Antonio Marotta and Andreas J. Kassler (Karlstad University, Sweden)
Presenter bio: Antonio Marotta received the M.Sc. and Ph.D. degrees in Computer Science
And Control Systems Engineering from the University of Napoli "Federico
II" in 2011 and 2015 respectively. He is currently a Post Doc
researcher at the Computer Science Department of the University of
Karlstad (Sweden). His research interests mainly focus on Cloud
Computing Infrastructure as a Service technologies, Green Computing
optimization models and Software Defined Networks.
Abstract: Reducing the CAPEX and OPEX is a major concern for Telecom Operators
(TOs): to this extent, Network Function Virtualization (NFV) has been
considered a key aspect to virtualize network functions and push them to
the NFV Infrastructure. Virtual Network Functions (VNFs) can be
deployed as a set of components running on several cooperating Virtual
Machines (VMs) inside modern data centers. As a consequence, it becomes
crucial for network operators to minimize the power consumption of their
NFV infrastructure, by using the minimum set of physical servers and
networking equipment subject to the constraints that VNFs impose on the
infrastructure in terms of compute, memory, disk and network resources
requirements. In this work, we present a joint resources and flow
routing assignment problem for VNFs placement, with the objective of
minimizing both the power consumption of the servers and switches needed
to deploy the overall virtualized infrastructure and routing graph. In
contrast to many existing works assuming perfect knowledge on input
parameters, such as traffic demands between the VNFs, which is difficult
to predict, we propose a novel mathematical model based on the Robust
Optimisation (RO) theory to deal with uncertainty on data. Our numerical
evaluation applies focuses on a specific use-case, that is the
deployment of a set of VNFs, typical of a virtualized Evolved Packet
Core (vEPC), namely the core for next generation mobile networks. We
demonstrate that with our model, a vEPC operator can trade-off between
two important aspects: the power consumption minimization on one side,
and the protection from severe deviations of the input parameters, such
as traffic demands and compute resources needed by the individual VNFI
components.
