Resource availability–aware adaptive provisioning of virtual data center networks

Abstract

Optimal provisioning of virtual networks with varying topologies on their physical counterpart is a major challenge in cloud data centers. Embedding the virtual network on the shared physical infrastructure poses a multitude of challenges to the infrastructure provider. This paper proposes a hybrid approach which adapts to the availability and fragmentation level of residual resources for near-optimal mapping of multiple virtual data center networks on a substrate data center network. The proposed technique uses coordinated static and dynamic embedding phases, adaptively, to maximize the density of embedding which increases the revenue of the cloud service provider. Also, it judiciously modifies the existing mappings for efficient utilization of resources. Empirical results show that the proposed hybrid approach can attain ∼10% higher acceptance ratio compared with the existing exact and heuristic techniques in the literature. With the proposed mixed integer programming model for static embedding, the average path length is reduced by 20% compared with the existing techniques.Int J Network Mgmt. 2019;29:e2066.wileyonlinelibrary.com/journal/nem

Distributed Virtual Network Embedding for Software-Defined Networks Using Multiagent Systems

“…6. Our current method differs from our previous work [35], as the current work partitions the substrate network, while our previous work [30] did not partition the substrate network. Besides, in [35] agents cooperate to map a VNR at the same time, however, in this paper agents can embed multiple VNR at the same time.…”

“…In other words, in the uncoordinated approach, at first, all virtual nodes are mapped one by one and then all links are mapped one by one according to some criteria. However, in the Coordinated approach, the coordination between node mapping and link mapping is considered [30]. Also, the VNE algorithm can be done in a centralized or distributed approach.…”

Optimal network‐aware virtual data center embedding

“…Constraints ( 16) and (17) ensure that if a PM (or PL), that hosts at least one VM (or VL), should be active, respectively. Likewise, the first inequality of constraints (18) indicates that if a PS, that hosts at least one VS, must be active.…”

“…Ma 16 studies the cost‐aware multi‐domain VDCE problem and incorporate the energy consumption of VSs embedding in their model. Gilesh et al 17 consider the virtual data center network embedding (VDCNE) problem that considers only network resources. They propose a mixed integer programming model to solve VDCNE without considering other resources (e.g., CPU, memory, and storage).…”

“…Ma 16 studies the cost-aware multi-domain VDCE problem and incorporate the energy consumption of VSs embedding in their model. Gilesh et al 17…”