The deployment of emerging cloud computing technologies onto telecommunication (telco) infrastructures, coupled with highly stringent regulations relating to energy efficiency, power consumption, and CO2 emissions, put telco datacenters and their associated operations under a drastic transformation process. This paves the way for new optimization opportunities, such as dynamic datacenter sizing (DDS) with respect to power consumption constraints and volatile demands across multiple resource types. DDS boils down to determining the optimal subset of active servers in a datacenter, and it can be modeled as a planning problem where decisions regarding resource availability are based on projections of resource demands. This raises the need for accurate and efficient forecasting solutions, which shall be tailored to the problem at hand.To this end, our work focuses on the investigation, the development, and the evaluation of forecasting methods which predict the demand of 5G (and beyond) workloads across multiple resources. Concretely, using the daily pattern of load data pertaining to an operational data-plane function and the prevailing way of horizontal scaling in virtualized datacenters, we exemplify the evolution of resource demands of 5G applications residing at the network edge. This lets us simulate a dataset of demands across various resource types, upon which we study informative features, and then train and evaluate a wide suite of forecasting methods. Our experiments show that the intrinsic characteristics of the data render simple statistical algorithms capable of accurately capturing the underlying patterns, and even outperforming more complex algorithms. Last, we evaluate and discuss the trade-off between single- and multi-output forecasting models.

Resource optimization arguably comprises a crucial aspect for Network Function Virtualization (NFV) infrastructures. In this respect, the problem of virtualized network function (VNF) placement commonly entails the selection of the most appropriate server within a single or among multiple Points-of-Presence (PoPs). Nevertheless, CPU cache hierarchy and memory locality in NUMA multi-core servers along with the diversity in NFV resource profiles introduce significant challenges in terms of intra-server resource allocation; a problem that is often overlooked. As such, we stress on the need for fine-grained resource allocation in NFV infrastructures, and, to this end, we study various aspects of CPU allocation for VNF chains. We deem this intra-server resource allocation problem as complementary to the large body of literature that seeks to optimize VNF placement onto virtualized infrastructures. More particularly, we shed light on the intra-server VNF placement problem, treating CPU cores as the main resource allocation unit. To this end, we assess the performance of multiple CPU allocation combinations under varying server utilization levels and processing workloads with diverse requirements in terms of CPU and memory. Our experimentation approach lets us progressively gain useful insights, which ultimately form ground rules that can be leveraged for optimized server resource allocation.

Service Function Chain (SFC) embedding optimization is crucial for the resource efficiency of Network Function Virtualization infrastructures (NFVI). Nevertheless, high utilization and/or fragmentation levels of a NFVI can significantly restrict the feasible solution space of any SFC embedding method, leading to inefficient SFC placements, or even inhibit SFC embedding. To rectify this problem, we stress on the need for SFC graph transformation (SFC-GT), i.e., explore the potential of SFC graph expansion prior to its embedding. SFC-GT aims at decomposing virtualized network functions (VNFs) into multiple instances with lower resource demands, facilitating their placement onto the NFVI. In this respect, we discuss the trade-off between embedding flexibility and complexity, in the context of SFC-GT. We formulate SFC-GT as a multi-objective optimization problem and design a mixed-integer linear program (MILP) to tackle it. Our simulation results demonstrate notable resource efficiency gains when SFC-GT is utilized prior to SFC embedding.

Network Function Virtualization (NFV) poses the need for efficient embeddings of network services, usually defined in the form of service graphs, associated with resource and bandwidth demands. As the scope of NFV has been expanded in order to meet the requirements of virtualized cellular networks and emerging 5G services, the diversity of resource demands across dimensions, such as CPU, memory, and storage, increased. This requirement exacerbates the already challenging problem of network service embedding (NSE), rendering most existing NSE methods inefficient, as they commonly account for a single resource dimension (i.e., typically, the CPU). In this context, we investigate methods for NSE optimization across multiple resource dimensions. To this end, we study a range of multi-dimensional mapping efficiency metrics and assess their suitability for heuristic and exact NSE methods. Utilizing the most suitable and efficient metrics, we propose two heuristics and a mixed integer linear program (MILP) for optimized multi-dimensional NSE. In addition, we devise a virtual network function (VNF) bundling scheme that generates (resource-wise) balanced VNF bundles in order to augment VNF placement. Our evaluation results indicate notable resource efficiency gains of the proposed heuristics compared to a single-dimensional counterpart, as well as a minor degree of sub-optimality in relation to our proposed MILP. We further demonstrate how the bundling scheme affects the embedding efficiency, when coupled with our most efficient heuristic. Our study also uncovers interesting insights and potential implications from the utilization of multi-dimensional metrics within NSE methods.

The increasing resource demand and diversity of network services is taken under serious consideration by the various stakeholders, driving the architecture design of 5G (and beyond) networks. Network slicing, as a prominent aspect of next-generation network architectures, aims at satisfying the diverse service requirements in terms of throughput, latency, reliability, and/or security. However, the prevailing way of slice provisioning, i.e., in the form of isolated bundles of computing, storage, and network resources, makes cross-slice communication inefficient, especially at the network edge. This inevitably hinders opportunities for Business-to-Business (B2B) synergies at the event of service co-location. In this paper, we study this novel aspect of network slicing, i.e., cross-slice communication (CSC). We particularly promote a form of optimized CSC, at which two co-located slices can establish peering in a secure and controlled manner, by confining peering traffic within the boundaries of the datacenter, while still preserving the important aspect of resource isolation. Such optimized CSC can foster synergies between service providers without additional latency or traffic in the backhaul/transport network. In this context, we investigate various ways to establish optimized CSC at edge computing infrastructures, based on functionalities offered by state-of-the-art management and orchestration (MANO) frameworks, such as OpenSourceMANO.

The evolution of Internet of Things (IoT) has sparked significant research interest in edge computing. Within this scope and given the ever-increasing number of IoT and mobile devices, computation offloading is emerging as a cutting-edge and significant research area with enormous potential and practical applications.In this respect, we present the architecture design and experimental evaluation of an orchestration framework for smart computation offloading from IoT or mobile devices to edge cloud servers. The proposed orchestration platform, namely COSMOS, includes control-plane components for workload prediction, load balancing, and admission control. COSMOS is particularly tailored to the needs of an object identification service that receives images from a multitude of Points of Interest (PoIs), performs object identification using a trained model (based on Tensorflow), calculates the prediction accuracy, and finally returns to the end-users the identification outcome and accuracy along with useful information about the identified object. COSMOS has been deployed and evaluated in a large-scale experimental facility that employs OpenStack and OpenSourceMANO (OSM) for Network Function Virtualization (NFV) orchestration. Our experimental results indicate the feasibility of computation offloading for this object identification service and further uncover useful insights in terms of performance and scalability.

The wide adoption of cloud computing, along with the advent of Network Function Virtualization (NFV) and its auspicious applications, have generated a new class of combinatorial optimization problems, with network service embedding (NSE) being one of the most prominent. NSE methods aim at improved resource efficiency and increased revenues for cloud resource providers. However, these methods commonly handle resources types with a single dimension (e.g., virtual nodes with only computing demands), thus limiting the scope of the generated solutions.In an attempt to address NSE under a pragmatic scope, we investigate the potential gains of a heuristic algorithm, which takes into account both the CPU and the memory dimension of virtual nodes. To the best of our knowledge, the novelty of our work lies on the fact that the proposed heuristic exploits insights from research on multi-dimensional virtual machine allocation, i.e., the computation and accounting of a suitability metric across multiple resource dimensions. Our simulation results demonstrate that the proposed NSE method outperforms both a mixed integer linear program (MILP) and a similar heuristic, which do not account for multiple resource dimensions.