Search

In this work, we present and analyze methods and mechanisms for interconnecting a network slice control and management system of the mobile network, with an IEEE Time-Sensitive Network (TSN) control plane. IEEE TSN is gaining momentum as a key technology that is able to provide network service guarantees for Ethernet-based communications. Although Ultra-Reliable Low-Latency Communications (URLLC) have been thoroughly investigated in 5G, incorporating TSN technologies in the Transport Network is expected to unleash the potential of end-to-end deterministic communications, especially in industrial environments and time-critical applications like factory automation.We elaborate on the concepts of a TSN-aware Xhaul network, present a novel architecture, and describe a set of amendments required in order to enable network slicing. With the devised approach, a slice-aware TSN-enabled transport network can be controlled and managed in an end-to-end orchestrated way. Implementation experience and evaluation results are reported using TSN-enabled prototype devices, OpenAirInterface (OAI), and JOX slice orchestrator.

We propose a Software-Defined Network (SDN)based approach for ultra-fast recovery of Time Sensitive Networks (TSN) in the case of failure events. We exploit the Source Routing paradigm for explicit path control and the creation of a stateless TSN data plane. We further propose a TSN failure recovery routing heuristic used to minimise link congestion, while we also introduce the concept of TSN subgraphs to quickly reschedule the flows traversing the problematic area. We evaluate our approach using SDN-based Source Routing and Linux-based TSN scheduling integrated into Mininet.

Recurrent Neural Networks (RNNs) have been shown to perform very effectively in session-based recommendation settings, when compared to other commonly used methods that consider the last viewed item of the user and precomputed item-to-item similarities. However, there is little systematic study on how RNNs perform in comparison to methods that use embeddings for item representation for Collaborative Filtering. In this paper we use two industry datasets to compare RNNs with other sequential recommenders that use various embedding methods to represent items. The first dataset corresponds to a typical e-commerce session-based scenario demanding effective next-item recommendation. The second dataset represents a last-basket prediction setting. Results show that although the RNN greatly outperforms embedding methods in the next-item scenario, the dynamic nature of the RNNs has not the same positive impact in the last-basket prediction task. We also present and test a framework that enables the hybrid utilization of text content and item sequences using embeddings. Finally, we report on experiments with reranking methods that demonstrate the effectiveness of simple and practical methods, using item categories, to improve the results.

Multidimensional factor and cluster analysis and embedding-based machine learning were evaluated toward a knowledge-based recommendation system for supermarket e-marketing. The goal was to produce personalized notifications on special offers, optimized per individual customer’s predicted response. To this purpose, we firstly applied Multiple Correspondence Analysis and Hierarchical Clustering to extract insights on the ordering behaviors and to identify customer classes associated with predictable preference patterns. Secondly, a neural network model based on embeddings was developed to predict the customers’ ordering actions on a personalized level at large scale. Application of the factor and cluster analysis on the Instacart dataset resulted in the identification of typical and niche patterns with prediction value. The neural network model was successfully trained to predict with satisfactory accuracy individual customers’ future orders, to be used as a basis for composing personalized recommendations.

We present a graph-based approach for the data management tasks and the efficient operation of a system for session-based next-item recommendations. The proposed method can collect data continuously and incrementally from an ecommerce web site, thus seemingly prepare the necessary data infrastructure for the recommendation algorithm to operate without any excessive training phase. Our work aims at developing a recommender method that represents a balance between data processing and management efficiency requirements and the effectiveness of the recommendations produced. We use the Neo4j graph database to implement a prototype of such a system. Furthermore, we use an industry dataset corresponding to a typical e-commerce session-based scenario, and we report on experiments using our graph-based approach and other state-of-the-art machine learning and deep learning methods., Comment: Preprint version of the paper, the original paper is published on ACM DL. 6 pages, 1 figure, 1 table

In 5G radio access networks, meeting the performance requirements of the fronthaul network is quite challenging. Recent standardization and research activities are focusing on exploiting the IEEE Time Sensitive Networking (TSN) technology for fronthaul networks. In this work we evaluate the performance of Ethernet TSN networks based on IEEE 802.1Qbv and IEEE 802.1Qbu for carrying real fronthaul traffic and benchmark it against Ethernet with Strict priority and Round Robin scheduling. We demonstrate that both 802.1Qbv and 802.1Qbu can be well used to protect high-priority traffic flows even in overload conditions.

Network architectures compliant with the Software Defined Networking (SDN) design paradigm, are expected to provide extreme flexibility for service orientation and allow for efficient use of network resources of cloud systems. Nevertheless, radical reconsidering and removal of boundaries set out when studying multi-domain communications are required, in order to unleash the hidden potential of SDN and provide a “holistic” network view. Although many domain-specific efforts have been proposed in the literature and indeed they gain a lot of industrial attention, real multi-domain SDN implementations over converged wireless-optical networks are just starting to disclose. In this work we present an open end-to-end multi-domain SDN system, while focusing on the necessary abstractions and virtualization techniques to integrate virtual wireless and optical resources. With the proposed system, called CONTENT, we shed light in the field of wireless-optical network virtualization from an end-to-end perspective. We present the architecture and the integrated testbed that realize the envisioned system. Evaluation results are provided and implementation experience is reported using the integrated solution. The way SDN methodologies and techniques can be used to support NFV concepts in end-to-end fashion are also presented.

In this paper we address the support of wearable mHealth applications in LTE and future 5G networks following a holistic approach that spans across the elements of a mobile network. The communication requirements change from one application to another so we propose a measurement methodology to facilitate the selection of the user equipment to fulfil these requirements. We also discuss a new network architecture to support traffic prioritization, RAN programmability, low latency and group communications to over-the-top applications. Our proposal is validated using several realistic experimentation platforms and the results show that mHealth systems can benefit from our approach.

Part 2: Poster Papers; International audience; In this work we investigate the coupling of Flex-E and X-Ethernet technologies as a means to realize interface slicing and fast switching respectively. Flex-E is investigated as a key technology that is able to split a physical interface into isolated sub-channels, decouple MAC rate from PHY rate and achieve hard bandwidth isolation. X-Ethernet technology introduces Ethernet PCS switching, eliminates table lookup and buffer queuing and will be investigated as a fast switching mechanism that is exploiting the PCS layer relay. We present evaluation results from implementation activities that showcase the ability of an integrated Flex-E/X-Ethernet solution to enable network slicing features in transport networks. Furthermore we present a technical approach on the way the control plane of Flex-E/X-Ethernet network can be integrated with an orchestration and management solution.

The proliferation of mobile devices with ubiquitous Internet access has made content access patterns highly volatile and spatio- temporally varying. At the same time, virtualization techniques have enabled the emergence of virtual Content Delivery Network (vCDN) providers, that bundle together a virtual infrastructure by utilizing storage resources anywhere along the path between the access network and the corresponding data center, where the requested content actually resides. In this context, efficient content placement in the various storage layers depends on accurately estimating content access patterns from the different access networks at any given time. At the same time, different vCDN providers compete against each other to provide low content retrieval latency to their users. However, there are some interesting synergies that might emerge among otherwise competing vCDN providers: (i) host content of another vCDN provider and, (ii) provide own content to the users/customers of other vCDN providers. In this paper, we formulate the overall problem of content placement for multiple vCDN providers that employ collaboration as an overall social-welfare maximization problem. The solution to this problem gives the optimal placement, achievable only in the case of full information. Alleviating the need for full information and considering vCDN providers separately as profit seekers, we also devise a distributed algorithm for content placement by exchanging limited information among them. Extensive simulation experiments show that business collaboration among competing vCDN providers is beneficial, as compared to isolated offerings, and allows them to adapt faster to content pattern changes.

Current solutions for network virtualization in the transport network relies on Virtual Local Area Network (VLAN) and Virtual Private Network (VPN) technologies, which facilitate logically isolated connectivity, providing data traffic seperation. However, co-existing VLAN/VPNs on top of a common infrastructure may compete for the same physical resources and introduce performance degradation, especially in case of congestion. This paper introduces Flex-Ethernet (Flex-E) and investigates its use as candidate interface solution for facilitating “hard” slicing in transport networks. This study sheds light on how Flex-E together with newly developed techniques like Segment Routing (SR), can enable a dynamic path allocation with strict performance guarantees. The standardization gaps and open challenges are also elaborated.

Mobile computation offloading has been identified as a key-enabling technology to overcome the inherent processing power and storage constraints of mobile end devices. To satisfy the low-latency requirements of content-rich mobile applications, existing mobile cloud computing solutions allow mobile devices to access the required resources by accessing a nearby resource-rich cloudlet, suffering increased capital and operational expenditures. To address this issue, in this paper, we propose an infrastructure and architectural approach based on the orchestrated planning and operation of optical data center networks and wireless access networks. To this end, a novel formulation based on a multi-objective nonlinear programming model is presented that considers energy-efficient virtual infrastructure planning over the converged wireless, optical network interconnecting DCs with mobile devices, taking a holistic view of the infrastructure. Our modelling results identify trends and trade-offs relating to end-to-end service delay, mobility, resource requirements and energy consumption levels of the infrastructure across the various technology domains.

This work focuses on the design, analysis and evaluation of Dynamic Weighted Round Robin (DWRR) algorithms that can guarantee CPU service shares in clusters of servers. Our motivation comes from the need to provision multiple server CPUs in cloud-based data center environments. Using stochastic control theory we show that a class of DWRR policies provide the service differentiation objectives, without requiring any knowledge about the arrival and the service process statistics. The member policies provide the data center administrator with trade-off options, so that the communication and computation overhead of the policy can be adjusted. We further evaluate the proposed policies via simulations, using both synthetic and real traces obtained from a medium scale mobile computing application.

Network slicing is expected to be the main pillar around which virtualization technologies together with SDN control and NFV, will provide on-demand network and cloud infrastructures and facilitate rapid service deployment. In this paper we present JOX, an event-driven orchestrator for the virtu- alized network, operating on top of the Juju management system, that inherently supports network slicing. JOX is a python-based generic network slicing orchestrator, with a plugins architecture that is able to support different segments of a modern mobile edge network. We present a concrete prototype implementation of JOX for LTE, with experimental results considering footprint analysis, performance metrics, and implementation experience for slicing and orchestrating of an operational LTE network.

The upcoming 5G ecosystem is envisioned to build business-driven network slices to accommodate the different needs of divergent service types, applications, and services in support of vertical industries. In this article, we describe the network slicing concept by unveiling a novel network slicing architecture for integrated 5G communications. Further, we demonstrate its realization for the case of evolved LTE using state-of-the-art technologies. Finally, we elaborate on the LTE-specific requirements toward 5G, and point out existing challenges and open issues.

This chapter provides an overview of different scenarios that can be identified in emergency situations, such as natural disasters or accidents, thus providing an overview of new services that might be beneficial for these scenarios as well as the major challenges that have to be addressed by public safety LTE networks to maintain communications in these situations. Additionally, a case study about a new video hardware platform to be used by first responders is introduced alongside new components to be included in the operators’ networks to provide these new services in the area of public safety.

Towards 5G mobile networks, the low-latency and high-bandwidth services are highly anticipated; however, legacy 3G and 4G networks now suffers from the mobile data surge. In this sense, pushing network services to the network edge has the potential to improve the traffic latency, user experience, and offload Internet traffic. Although the LTE/LTE-A network can highly benefit from the Mobile Edge Computing (MEC) principle, a detailed MEC architecture is not currently in place. In this work, we propose a modular architecture for the Mobile Edge Host that is ETSI compliant and describe the functional mapping of the architecture to LTE systems. Proof-of-concept demonstrations based on the OpenAirInterface (OAI), a software implementation of LTE/LTE-A systems, present significant benefits of adopting the MEC concept in data caching use case.

Mobile-Edge Computing (MEC) is about offering application developers and service providers cloud-computing capabilities and an IT service environment at the edge of the mobile network. However, although cloud computing can be used to meet traditional challenges, like scalability concerns and provide for fast resource provisioning times, a multifaceted analysis is required when it comes in multi-operator environments with time-critical applications and services. In this work, we claim that the service importance must be at the epicenter when it comes to the scheduling and placement decision of whether to deploy the service at the edge network or not. Virtual machine (VM) scheduling decisions should avoid SLA violations for popular or time-critical services, and be fair between the service providers. A Lyapunov optimization framework is derived to solve this stochastic optimization problem that aims to maximize the revenue of the physical infrastructure owner in a multi-network operator-sharing environment with time-critical SLAs. A series of simulation experiments validate the high effectiveness of the proposed approach over benchmarking ones.

Together with recent advancements in Radio Access Network (RAN) technologies, Wi-Fi is expected to be at the center of research on the subject of ubiquitous wireless connectivity, towards building the new 5G ecosystem. Nevertheless, the necessary testbed infrastructure to support large scale experimentally driven research seems to be missing. In this work we present the design and implementation of a novel Virtual Wi-Fi Testbed. We present how a traditional Wireless Testbed can support hundreds of virtual Wi-Fi nodes that are open to experimenters. We discuss the design and implementation of the virtualization tools. We demonstrate the accuracy and the overhead analysis of the approach in the face of actual testbed conditions. Implementation experience is also reported on the benefits of using the proposed virtualization approach for a simple association algorithm.

In this work, we present novel Architectural Design Patterns towards open, cloud-based 5G communications. We provide a brief classification of technologies that cannot be ignored in the design process of 5G systems and illustrate how a new technological added value can be created, when current methodologies, design paradigms, as well as design patterns and their extensions are properly exploited in efficient Radio Access Network (RAN) architectures. We believe that in many cases, the required technology is already there; nevertheless the correct approach has to be worked out and placed within an appropriate context, especially in the case of the integration of complex RAN systems. The enhancements in RF optimization, the progress in cloud computing, Software Defined Networks (SDN) and Network Function Virtualization (NFV), new design concepts such as Network Slicing have to become part of the RAN design methodology. Diverse architectural concepts should break existing stereotypes to pave the way towards the true 5G system integration.

In this paper, we provide a revolutionary vision of 5G networks, in which SDN technologies are used for the programmability of the wireless network, and where a NFV-ready network store is provided to Mobile Network Operators (MNO), Enterprises, and Over-The-Top (OTT) third parties. The proposed network serves as a digital distribution platform of programmable Virtualized Network Functions (VNFs) that enables 5G application use-cases. Currently existing application stores, such as Apple's App Store for iOS applications, Google's Play Store for Android, or Ubuntu's Software Center, deliver applications to user specific software platforms. Our vision is to provide a digital marketplace, gathering 5G enabling Network Applications and Network Functions, written to run on top of commodity cloud infrastructures, connected to remote radio heads (RRH). The 5G Network Store will be the same to the network provider as the application store is currently to a software platform.

In this work, we present how programmable data-plane technology (software routers) offers an easy-to-apply mechanism to create virtual wireless networks and support buffering and scheduling decisions. Furthermore, we present a feedback-based buffering mechanism that is able to provide throughput ratio guarantees per virtual network, without requiring any modifications in the 802.11 driver and without relying on statistical knowledge of the workload per virtual network or knowledge regarding the channel conditions. We implement the proposed mechanism in a software router in a 802.11 Access Point and we evaluate its performance in a wireless testbed environment. The methodology and the mechanics developed are generic and with some modifications can be applied to differentiating services for other types of guarantees like delay.

One of the distinctive features of the new virtualized ecosystem, is the multi-stakeholder participation in the way cloud services are designed, deployed and offered. In this work, we address the emerging content replication problem a Content Delivery Network (CDN) provider needs to consider, when deploying its network over multi-domain, heterogeneous environments, where virtual network operators utilize both the CDN services and the virtualized infrastructures. In our model, the benefit that the CDN provider enjoys may be different per network operator for the same request, while our model takes into account the replication cost to every domain, as well as the user mobility, besides physical storage limitations. Since the optimal placement of the objects at the caches of the various domains resembles the multiple knapsack problem, which is NP-complete, we provide two approximate solutions to the emerging content placement problem. We evaluate the proposed policies through extensive simulations and we compare them against a myopic method, where a domain is unaware of the caching strategy of the other domains that is connected to.

Undoubtedly, SDN has the potential to fundamentally change the way end-to-end networks are provisioned and designed. By introducing programmability into networking, the handling of network changes and network provisioning via open software interfaces has become a reality; together with the hardware/software virtualization advances in server systems and software platforms, cloud services can now be rapidly deployed and efficiently managed. The CONTENT project is an EU funded effort for network and infrastructure virtualization over heterogeneous, wireless and metro optical networks, that can be used to provide end-to-end cloud services. In this work we present the wireless network virtualization solution, in the light of the CONTENT technical approach, where a convergent LTE/Wi-Fi network is virtualized and interconnected with an optical TSON metro network. We present our approach in designing and implementing converged virtual 802.11 and LTE wireless networks and the corresponding efficient wireless data-plane mechanisms required, in order to satisfy strict QoS requirements.

The NITOS wireless testbed, is one of the main building blocks in the wireless testbed experimentation, offered by the FIRE community, and a core Fed4Fire partner. Its main focus is on wireless accessed technologies and on all layers of the protocol stack. In the rapidly changing technological environ-ment, a unique opportunity is provided to enhance the NITOS testbed with cloud computing experimentation capabilities and increase the heterogeneity and diversity of possible services that can be offered to experimenters/ developers. In this paper we describe the NITOS approach on adopting cloud technologies and SDN capabilities and how we upgrade the meaning of “delivered service” in testbed experimentation using SOA extensions.

This paper proposes a next generation ubiquitous converged infrastructure to support Cloud and mobile Cloud computing services. The proposed infrastructure facilitates interconnection of fixed and mobile end users with computational resources through a heterogeneous network integrating optical metro and wireless access networks. To satisfy the low-latency requirements of content-rich mobile applications, a novel stochastic virtual infrastructure planning model that takes a holistic approach considering jointly the network and computational resources is presented. Our modelling results identify trends and trade-offs related to end-to-end service delay, resource requirements and energy consumption levels of the infrastructure across the various technology domains under traffic uncertainty.

Cloud service provisioning on top of virtual infrastructures is of major importance in modern ICT, since it is directly correlated to the way business models are designed and revenue is generated from the cloud service providers. In this work we examine an end-to-end content replication problem over cloud-based multi-technology infrastructures. We extend the classical model where every network node is a potential replica carrier and the link weights represent hops/delay and we examine replication schemes for content that a) is requested by customers belonging in different virtual networks and b) depending on the requester there is different impact on the system operational cost. We examine both centralized and distributed content replication management policies and we evaluate their performance through extended simulations, by means of total cost, the number of object replacements and the number of iterations required. © 2014 IEEE.

Although cloud computing and the Software Defined Network (SDN) framework are fundamentally changing the way we think about network services, multi-domain and multi-technology problems are not sufficiently investigated. These multi-domain, end-to-end problems concern communication paths that span from the wireless access and the wireless backhaul networks to the IT resources through optical networks. In this paper we present the CONTENT project approach to network and infrastructure virtualization over heterogeneous, wireless and metro optical networks, that can be used to provide end-to-end cloud services. The project goal is to drive innovation across multi-technology infrastructures and allow ICT to be delivered and consumed as a service by Virtual Network Operators. The communication mechanics between wireless and optical domains and the physical layer abstractions of a CONTENT Virtual Network are presented and the relation of the proposed approach with the SDN framework is investigated.

In this paper, we study the problem of resource allocation in the setting of multitier data centers. Our main motivation and objective is to provide applications hosted in the data center with different service levels. In such centers, there are several mechanisms the designer can use to achieve such objectives. We restrict our attention to CPU time at the service tier as the resource; the objective we consider is service differentiation, expressed as allocating prespecified percentages of this resource to applications. Then, mechanisms at the designer's disposal to provide desired service differentiation include the triplet of load balancing through the switch fabric, enqueueing at a server and scheduling at a server. We focus on the enqueueing component of control mechanisms. We provide, through analysis and simulations “rules of thumb” for situations where simple enqueueing policies can provide service differentiation.

Enterprise computing facilities, such as data centers or server farms typically employ service-oriented architectures (SOA) to support multiple, XML-based Web Services. They are typically architected in multiple computing tiers, in which one tier is used for, say, offloading the CPU-intensive XML processing onto a cluster of (potentially virtual) middle-ware appliances. Service differentiation in enterprise networks addresses the issues of managing the enterprise network resources in order to achieve desired performance objectives. In this paper, we define a dynamic algorithm that manages allocation of CPU time in the appliance tier. We evaluate the service differentiation capabilities of this algorithm via simulations.

In the current 4G era, the dual connectivity technique utilizes radio resources scheduled by two distinct base stations for a single user equipment to enhance the data throughput. Multi- connectivity, as a natural evolution of dual connectivity, is one of the key 5G techniques to improve both the user performance and overall resource utilization, allowing dynamic user traffic steering across multiple connections of one or more radio access technologies (RATs). However, one of the main challenge in multi-connectivity is to efficiently allocate resources across multiple connections under heterogeneous quality of service (QoS) requirements. In this paper, we examine a resource allocation problem under multi- connectivity in an evolved LTE network and propose a utility proportional fairness (UPF) resource allocation that supports QoS in terms of requested rates. We evaluate the proposed policy with the proportional fairness (PF) resource allocation through extensive simulations and characterize performance gain from both the user and network perspectives under different conditions.

A converged optical-wireless testbed is formed integrating TSON sub-wavelength networking in UK and the NITOS wireless solution in Greece. End-to-end service provisioning and flow differentiation are demonstrated across two networks communicating over the GEANT.

Communications for emergency services have traditionally been provided by dedicated radio technologies, like the TETRA system. However the need for more efficient mobile communications is gaining popularity in these scenarios due to extreme requirements by means of availability, price and spectrum usage. This trend will be further accentuated in the coming years with the appearance of 5G systems that promise better figures in almost all the Key Performance Indicators (KPIs). LTE can provide many of the functionalities required by emergency services and is currently being evolved towards the 5G era by incorporating heterogeneous networking, over the top applications, network function virtualisation and more. The Q4Health project focuses on the preparatives for market introduction of eHealth video applications for first responders based on LTE technology. This preparation includes the investigation of advanced LTE programmability features and the design and execution of several novel experiments in order to optimise the communications for this platform.

In this paper we focus on the problem of multidomain orchestration for Network Function Virtualization (NFV), over multi-technology environments. In order to facilitate service deployment in end-to-end setups, new orchestration designs are required that exploit and advance existing methodologies. We examine in detail the challenges on multi-domain NFV orchestration for the general case and we provide the current landscape and existing technologies. We also describe a reference architecture for the problem of multi-domain NFV orchestration, that also supports the concept of Network Slicing. Finally, we present a realization of the architecture proposed for the LTE network and we describe a use case with LTE-specific considerations.