Column generation based-approach for IaaS aware Networked Edge Data-Centers

Abstract

Abstract-In this paper, our focus is on proposing a resource allocation approach for Infrastructure as Cloud Service (IaaS). Networked Edge Data-Centers (DCs) were selected as a physical infrastructure to handle IaaS Cloud requests. Doing so, we hold the potential of providing both a relatively high degree of independence between physical data centers' outages and an opportunity to reach economically IaaS cloud service users with low latency. An IaaS request is defined through two main requirements: Hosting and Networking resources. In literature, a number of works have proposed IaaS mapping approaches, however their focus was mainly on the cloud hosting requirements. Consequently, this may result in a high blocking of IaaS requests and a low utilization of resources inside the DCs due to stranding and fragmentation. To handle these drawbacks, we propose a joint optimization model of DCs and networking IaaS requirements that makes use of Column Generation technique coupled with a Branch and Bound approach in order to solve it efficiently.

Software Defined Network-Based Edge Cloud Resource Allocation Framework

“…Using the Column Generation (CG) technique means reformulating the link provisioning problem in terms of Independent provisioning Configurations (IECs) [31].…”

“…The link provisioning problem can be formulated with respect to the variables (λ c ), c ∈ C. Under the new formulation, the problem then chooses a maximum of W IECs, where W is the number of available wavelengths per optical fiber link l ∈ L s . The resulting configuration corresponds to the so-called master problem in the Column Generation technique [31], while each configuration design corresponds to the so-called pricing problem. We denote by cost c the cost of an IEC c. It corresponds to the costs of link bandwidth and OEO conversion ROADM ports (used to groom IaaSs bandwidth requirement on the same wavelength) used by IEC c. It is defined according to Equation (1) as follows:…”