ABSTRACT Vizing's generalized theorem states that any multigraph with maximum degree and maximum multiplicity has an edge coloring with at most colors. The runtime of Vizing's algorithm, which can find such a coloring, scales quadratically with . This is an important drawback for multigraphs arising from applications like photonic networks, where edge degrees and multiplicities can be very high. In this work, we propose a new edge coloring algorithm for multigraphs, which scales linearly with . To the best of our knowledge, no existing edge coloring algorithm, using so few colors, scales less than quadratically with . We describe one algorithm that finds a near‐optimal coloring and another algorithm that finds an optimal coloring with and worst‐case time complexity, respectively. We verify the working mechanisms as well as the performance of these algorithms, both for purely random test inputs and for synthetic test inputs whose structure is based on real applications.

Deterministic networks (DetNets) demand precise delay modeling to meet quality-of-service (QoS) requirements for time-sensitive applications. Queueing delay plays the most critical role among all contributors to end-to-end latency. Existing Time-Sensitive Networking (TSN) standards, such as Time-Aware Shaping (TAS) and Cyclic Queueing and Forwarding (CQF), rely on high-precision time synchronization to effectively manage queueing delays. Larger-scale DetNets suffer from less accurate time synchronization, resulting in alternative data plane approaches. The approach followed in this paper uses strict priority queueing (SPQ), network calculus (NC), and pre-configured queueing delay budgets to bound queueing delays and improve routing flexibility and simplicity. While it has been shown that queueing delay budget configurations affect routing efficiency, there are no studies on deriving optimal budgets. This paper addresses this gap by introducing a novel optimization model that jointly determines optimal flow routes and queueing delay budgets while leveraging NC to ensure delay compliance. The model is evaluated in an industrial setting, demonstrating its utility as a benchmark for routing solutions, and as a tool to identify scenarios in which SPQ can be beneficial and to derive queueing delay budget configurations for specific use cases. Moreover, the analysis highlights the differences between priorities and the importance of considering traffic demands, network topology, and capacity when fine-tuning the DetNet.

Distributed computer vision deployment frameworks aim to reduce end-to-end latency and backbone traffic by distributing computation across edge and cloud tiers. In typical architectures, nodes can process a portion of incoming image streams locally or in an edge server and offload the remainder to the cloud. However, runtime decisions about task scheduling and offloading often impose a significant computational burden on nodes, leading to unnecessary runtime overhead and underutilization of hardware resources for deployed applications. To address these challenges, we propose an offline mixed-integer optimisation framework that partitions workloads into local edge processing and deterministic offloading to the cloud tier. Each edge node handles multiple image streams by deploying parallel processing pipelines. Our framework enforces quality-of-service (QoS) constraints while maximising the proportion of computation performed at the edge. By precomputing an optimal configuration, the edge node can meet latency and resource requirements without incurring the overhead of on-the-fly optimisation. We evaluate our approach against several standard image stream processing approaches and demonstrate that our method satisfies QoS targets with lower processing delay and higher resource utilisation. We demonstrate that as the number of sensors increases, our framework can adjust the deployed configuration to enhance on-edge processing.

The necessity of processing real-time data at the network edge is growing. Low-power AI accelerators, especially edge GPUs, help meet this demand by mitigating cloud-related latency and bandwidth issues. However, GPUs remain underutilised, even in heavy workloads, due to a limited understanding of resource sharing in edge computing. This work analyses key GPU metrics: utilisation, memory, streaming multiprocessors (SMs), and tensorcores on NVIDIA Jetson devices under concurrent vision-inference workloads. Our findings show that while GPU utilisation can reach 100 % with optimisations, SMs and tensor cores often run at only 15-30 % capacity.

In the era of 5G and the upcoming 6G, current software stacks are ineffective when it comes to intelligent and fast decision-making, which necessitates the need for sustainable yet performant and scalable solutions. A significant amount of research has been done in this area, but an end-to-end solution was lacking. In this scope, we identify that we need a novel software stack which can handle the growing need for a vast amount of data generation with the help of the cloud-edge continuum, swarm programmability, along with secure deployments of applications. In this context, we propose OASEES, an architectural framework for a decentralised AI/ML computing stack that unifies diverse computing resources, peer-to-peer coordination protocols, and secure middleware. Our design outlines modular different compute infrastructures(CPUs, GPUs, TPUs and custom ASICs) orchestrated by a lightweight containerbased runtime and governed by a blockchain-enabled tamperproof data storage, decentralised coordination (decentralised autonomous organisation, voting, etc) to facilitate transparent discovery, allocation, and incentivization. We also propose a detailed performance and scalability metrics framework covering latency, throughput, resource utilisation, and cost-per-inference, intended as the foundation of subsequent evaluations.

Unmanned Aerial Vehicles (UAVs) are increasingly deployed in inspection and monitoring missions, yet onboard computation and communication impose significant energy burdens that limit flight time and operational scope. In this work, we introduce a novel, blockchain-enabled framework-grounded in the Distributed Autonomous Organization (DAO) paradigm-for orchestrating distributed analytics across a swarm of UAVs. Leveraging the OASEES project's smart-contract architecture, each drone embeds a Metrics Module for real-time power monitoring, a Behavioral Module for adaptive control, and a Blockchain Agent that autonomously proposes, votes on, and executes collective decisions. Three concurrent threads-Proposal Trigger, Voting, and Action Execution-enable fully decentralized governance of swarm behavior: from detecting critical energy thresholds and formulating swarm-wide conservation maneuvers, to executing approved strategies across all members. We validate our framework in a UAV-based infrastructure inspection scenario, employing a YOLOv5 object-detection pipeline to classify four corrosion classes on a telecommunications mast under three video-capture modalities (short-distance, long-distance, and horizontally concatenated streams). Across all configurations, our system achieves near-perfect precision, recall, and mean Average Precision (mAP50-95$\approx 0.995$), demonstrating both the efficacy of distributed workload inference and the feasibility of treating a single drone as a multi-feed processor. These results underscore the potential of DAO-driven UAV swarms for energy-aware, resilient aerial analytics, and pave the way for fully decentralized 5G/6G-enabled airborne networks.

High-Performance Computing (HPC) is indispensable in the current technological era. To preserve the development of innovative ideas and technologies, supercomputers must continue to grow in size. Consequently, the importance of the interconnection network that connects the computing resources, increases. The communication demands of today’s workloads can cause bottlenecks in the supercomputer’s network, resulting in the utilization of only 3% of their available computing power. This inefficiency presents an opportunity to enhance system efficiency through network-oriented resource management.In this paper, we outline the limitations of state-of-the-art resource management strategies and identify the challenges associated with solving the underutilization in large-scale HPC systems. We advocate that flexibility and direct control over resources are the fundamental principles for overcoming these challenges. To this end, we present our research strategy focussing on designing closed-loop resource allocation strategies. Opposed to existing strategies, our approach adaptively reacts to the dynamic behavior of the system through elastic allocation, ensuring optimal resource allocation.

Next-generation Internet will require strict end-to-end delay guarantees to support upcoming latency-sensitive applications. The IEEE 802.1 Time-Sensitive Networking (TSN) standard has become the de-facto solution for Ethernet-based L2 networks to support applications with strict latency, jitter and packet loss requirements. The IETF DetNet Working Group tries to expand on TSN to support real-time applications over larger-scale L3 networks. This paper proposes control and routing strategies that provide latency guarantees in L3 networks without requiring time synchronization among nodes. The proposed strategies include a link-state routing protocol and several exploration-based protocols that exploit queue-level information and network calculus to provide latency guarantees. Additionally, the use of queueing delay budgets enables independence among flows, while enabling fine-grained routing. This allows to make better routing decisions and to support applications with diverse latency requirements. Moreover, traffic shaping is only required at the network ingress. The strategies are evaluated extensively and compared in a simulation environment in multiple large-scale scenarios, considering acceptance rate, network utilization, path dissemination time, control overhead, and memory consumption, as well as how these metrics evolve w.r.t. different network scales. Experimental results demonstrate that representative delay-constrained traffic demands can be accommodated adequately by queue-level link-state routing protocols only in smaller-scale networks. In larger-scale network scenarios, breadth-first exploration-based protocols are required to provide stable performance w.r.t. acceptance rate and path dissemination times at the cost of only linearly increasing control overhead and memory footprint.

Time-Sensitive Networking (TSN) is a set of standards currently being defined by the IEEE 802.1 Time-Sensitive Networking Task Group [1] for Real-Time behavior in the network. Software-Defined Networking (SDN) provides a good solution for implementing TSN networks due to its characteristics such as run-time flexibility, benefits in management, cost efficiency, and performance. For achieving Real-Time behavior in TSN networks, the high-priority traffic should be scheduled precisely to fulfill its timing requirements. For this purpose, in SDN-based implementation, the control plane must have a good knowledge of the network topology and the delay in the network to be able to schedule the traffic. In this paper, we propose a topology discovery mechanism for the Central Network Controller (CNC) in TSNs based on the Link Layer Discovery Protocol (LLDP) able to discover accurate link latency characteristics as required for time-aware scheduling without relying on external time synchronization protocols such as PTP. We evaluate its feasibility and assess its performance in terms of required bandwidth and achieved accuracy.

Electromagnetic nano-communication has increasing attention in recent years. Several developments have been achieved in the fabrication, communication and management of various nano-network devices serving potential applications ranging from software-defined metamaterials, wireless robotic materials and body-centric communication. Such applications need uplink and downlink communication between the deployed nano-network and the external macro-world or the Internet through nano-interfaces. As a result, heterogeneous nano-network devices and their interoperability in different Internet of nano-things applications become new challenges for nano-network communication. In this regard, dynamic, flexible and distributed micro/nano-gateways can accommodate such sustainable issues and make the nano-network fully operational, regardless of the adopted application domain or the protocols used in communication. Network functions virtualization and software-defined networking technologies altogether can overcome these challenges. This article proposes SDN-based architecture and software module for the micro/nano-gateway. The proposed software module converts data formats and protocols between nano-network and traditional network domains allowing the nano-devices to be linked to the Internet. A prototype of the module is built, and the performance of the proposed algorithm is evaluated based on two communication scenarios; single tenant and multitenant. The result shows the effect of the total number of connected nano-devices and the number of packets sent by each device on the total average round-trip processing delay and the overall throughput of the micro/nano-gateway.

Recent developments in nanotechnology herald nanometer-sized devices expected to bring light to a number of groundbreaking applications. Communication with and among nanodevices will be needed for unlocking the full potential of such applications. As the traditional communication approaches cannot be directly applied in nanocommunication, several alternative paradigms have emerged. Among them, electromagnetic nanocommunication in the terahertz (THz) frequency band is particularly promising, mainly due to the breakthrough of novel materials such as graphene. For this reason, numerous research efforts are nowadays targeting THz band nanocommunication and consequently nanonetworking. As it is expected that these trends will continue in the future, we see it beneficial to summarize the current status in these research domains. In this survey, we therefore aim to provide an overview of the current THz nanocommunication and nanonetworking research. Specifically, we discuss the applications envisioned to be supported by nanonetworks operating in the THz band, together with the requirements such applications pose on the underlying nanonetworks. Subsequently, we provide an overview of the current contributions on the different layers of the protocol stack, as well as the available channel models and experimentation tools. Finally, we identify a number of open research challenges and outline several future research directions.

Many media services undergo a varying workload, showing periodic usage patterns or unexpected traffic surges. As cloud and NFV services are increasingly softwarized, they enable a fully dynamic deployment and scaling behaviour. At the same time, there is an increasing need for fast and efficient mechanisms to allocate sufficient resources with the same elasticity, only when they are needed. This requires adequate performance models of the involved services, as well as awareness of those models in the involved orchestration machinery. In this paper we present how a scalable content delivery service can be deployed in a resource- and time-efficient manner, using adaptive machine learning models for performance profiling. We include orchestration mechanisms which are able to act upon the profiled knowledge in a dynamic manner. Using an offline profiled performance model of the service, we are able to optimize the online service orchestration, requiring fewer scaling iterations.

In this demonstration, we present dynamic allocation of accelerator resources to SSH tunnels in an NFV environment. In order to accelerate a VNF, its compute-intensive operations are offloaded to hardware cores running on an FPGA. The CPU utilization information of VNFs is continuously processed by a service management component to dynamically decide the suitable target to run VNF's crypto-operations. We also demonstrate switching between the non-accelerated and hardware-accelerated SSH-tunnels triggered by a change in the nature of the data traffic flowing through the tunnel and indicate throughput gains obtainable in dynamically switching contexts.

The virtualization of compute and network resources enables an unseen flexibility for deploying network services. A wide spectrum of emerging technologies allows an ever-growing range of orchestration possibilities in cloud-based environments. But in this context it remains challenging to rhyme dynamic cloud configurations with deterministic performance. The service operator must somehow map the performance specification in the Service Level Agreement (SLA) to an adequate resource allocation in the virtualized infrastructure. We propose the use of a VNF profile to alleviate this process. This is illustrated by profiling the performance of four example network functions (a virtual router, switch, firewall and cache server) under varying workloads and resource configurations. We then compare several methods to derive a model from the profiled datasets. We select the most accurate method to further train a model which predicts the services' performance, in function of incoming workload and allocated resources. Our presented method can offer the service operator a recommended resource allocation for the targeted service, in function of the targeted performance and maximum workload specified in the SLA. This helps to deploy the softwarized service with an optimal amount of resources to meet the SLA requirements, thereby avoiding unnecessary scaling steps.

In pursuit of a flexible, resource efficient and high- performant 5G infrastructure, many operators, vendors and research consortia are currently developing, testing and inte­grating their NFV platform with associated management and orchestration (MANO) functionality. The SONATA NFV platform follows a micro-service design, which involves a tight coupling between an SDK, monitoring and MANO functionality, targeting a secure and stable software foundation. This experience paper gives a thorough overview on the encountered challenges, insights and resulting learnings when implementing and integrating the SONATA Service Platform using a continuous integration and delivery DevOps methodology. This is the result of a strong cooperation between prominent equipment vendors, network operators, software companies and universities, providing a set of constructive recommendations in hope of catalysing the development and deployment of NFV platforms.

Network slicing is the emerging paradigm in which operators use their resources to provide multiple logical networks and associated resources, with varying configurations and at the same time. More and more vertical industries need their machines and devices connected in networks with specific requirements. In order to provide networks fitted to these usecases, and not require that they adapt to the one-size-fits-all network as is currently the case with the mobile Internet, the telecom community vowed to include network slicing functionality within its next generation of mobile networking, 5G, as an end-to-end network solution. As the concept is new and still not fully grasped, we develop and refine the concept of a network slice both from a business and a technological point of view. We investigate how network slicing in the context of a vehicular network could be implemented and how it advances the state of the art. This involves a detailed study of the involved technologies across a range of infrastructures and network segments, as well as the resulting gaps in the existing technology landscape. Based on the lessons learned in this concrete usecase, network slicing is considered in a broader 5G landscape. We capture the main challenges and potential directions in order to make network slicing a true enabler of 5G-driven vertical industries.

Developing a virtualized network service does not only involve the implementation and configuration of the network functions it is composed of but also its integration and test with management solutions that will control the service in its production environment. These integration tasks require testbeds that offer the needed network function virtualization infrastructure (NFVI), like OpenStack, introducing a lot of management and maintenance overheads. Such testbed setups become even more complicated when the multi point-of-presence (PoP) case, with multiple infrastructure installations, is considered. In this demo, we showcase an emulation platform that executes containerized network services in user-defined multi-PoP topologies. The platform does not only allow network service developers to locally test their services but also to connect realworld management and orchestration solutions to the emulated PoPs. During our interactive demonstration we focus on the integration between the emulated infrastructure and state-of-theart orchestration solutions like SONATA or OSM.

The growing demand for flexibility and cost reduction in the telecommunication landscape directs the focus of service development heavily to programmability and softwarization. In the domain of Network Function Virtualization (NFV), one of the goals is to replace dedicated hardware devices (such as switches, routers, firewalls) with software-based network functionalities, showing comparable performance when deployed on common servers. In this paper, we discuss how current VNF implementation and deployment strategies impact the efficient monitoring of their resources. In a multi-tenant, NFV-based ecosystem, different Service Providers deploy VNFs on a shared infrastructure, where the Infrastructure Provider exposes only VNF specific metrics and little information about the physical hosts where the VNFs are eventually orchestrated. Especially in the situation where datacenters are overcommitted, detecting the risk of e.g. CPU starvation is not straight-forward, when no information from the physical host is available. A new monitoring technique is introduced, based on the skewness of the measured probability distribution of the VNF resource consumption. Our measurements show that this metric is a good indicator for the (un)availability of the required CPU resources in the datacenter.

Although geometric routing is proposed as a memory-efficient alternative to traditional lookup-based routing and forwarding algorithms, it still lacks: (i) adequate mechanisms to trade stretch against load balancing, and (ii) robustness to cope with network topology change. The main contribution of this paper involves the proposal of a family of routing schemes, called Forest Routing. These are based on the principles of geometric routing, adding flexibility in its load balancing characteristics. This is achieved by using an aggregation of greedy embeddings along with a configurable distance function. Incorporating link load information in the forwarding layer enables load balancing behavior while still attaining low path stretch. In addition, the proposed schemes are validated regarding their resilience towards network failures.

Network Service Chaining (NSC) is a service concept which promises increased flexibility and cost-efficiency for future carrier networks. The two recent developments, Network Function Virtualization (NFV) and Software-Defined Networking (SDN), are opportunities for service providers to simplify the service chaining and provisioning process and reduce the cost (in CAPEX and OPEX) while introducing new services as well. One of the challenging tasks regarding NFV-based services is to efficiently map them to the components of a physical network based on the services specifications/constraints. In this paper, we propose an efficient cost-effective algorithm to map NSCs composed of Network Functions (NF) to the network infrastructure while taking possible decompositions of NFs into account. NF decomposition refers to converting an abstract NF to more refined NFs interconnected in form of a graph with the same external interfaces as the higher-level NF. The proposed algorithm tries to minimize the cost of the mapping based on the NSCs requirements and infrastructure capabilities by making a reasonable selection of the NFs decompositions. Our experimental evaluations show that the proposed scheme increases the acceptance ratio significantly while decreasing the mapping cost in the long run, compared to schemes in which NF decompositions are selected randomly.

End-to-end service delivery often includes transparently inserted Network Functions (NFs) in the path. Flexible service chaining will require dynamic instantiation of both NFs and traffic forwarding overlays. Virtualization techniques in compute and networking, like cloud and Software Defined Networking (SDN), promise such flexibility for service providers. However, patching together existing cloud and network control mechanisms necessarily puts one over the above, e.g., OpenDaylight under an OpenStack controller. We designed and implemented a joint cloud and network resource virtualization and programming API. In this demonstration, we show that our abstraction is capable for flexible service chaining control over any technology domains.

Geometric routing is an alternative to traditional routing algorithms in which traffic is no longer forwarded using lookup tables, but using coordinates in an embedding of the underlying network. A major downside of current geometric routing algorithms is their inability to handle network failures in a graceful manner. Moreover, they cannot deal with dynamic graph topologies. This article presents a geometric routing scheme that uses an embedding based on a spanning forest. Allowing nodes to select the optimal spanning tree leads to both shorter paths and natural traffic redirection in case of network failures. By constructing the forest in such a way that its disconnected components have low redundancy, their coverage is maximized. Results show that this system is able to operate gracefully in severe failure scenarios, without any form of path protection or restoration. By means of an embedding regeneration procedure, the routing scheme is able to continuously adapt to an altering network topology. This geometric routing algorithm effectively combines two key objectives, namely low path stretch and high robustness. © 2015 Wiley Periodicals, Inc. NETWORKS, Vol. 66(4), 320–334 2015

Information-centric networking has been proposed to achieve efficient and reliable distribution of content. We propose a model to assign content locators to content names. Information routing decision is made based on geometric routing using the assigned locators. We consider a geometric routing scheme known as geodesic geometric routing. We demonstrate on the iLab.t virtual wall the successful operation of the proposed scheme and the gain of using content locators on capacity utilization by means of caching.

Mininet is a great prototyping tool which combines existing SDN-related software components (e.g., Open vSwitch, OpenFlow controllers, network namespaces, cgroups) into a framework, which can automatically set up and configure customized OpenFlow testbeds scaling up to hundreds of nodes. Standing on the shoulders of Mininet, we implement a similar prototyping system called ESCAPE, which can be used to develop and test various components of the service chaining architecture. Our framework incorporates Click for implementing Virtual Network Functions (VNF), NETCONF for managing Click-based VNFs and POX for taking care of traffic steering. We also add our extensible Orchestrator module, which can accommodate mapping algorithms from abstract service descriptions to deployed and running service chains.

In this paper, we use machine learning technique at the routers to study the link state protocol data to predict the existence of shared risk link groups (SRLG) in the network. In particular, we use the correlation between different link state updates (LSUs) issued by different network nodes (routers) upon failure. The concerned network router then runs a novel Bayesian network based statistical learning process to learn about the possible existence of SRLGs. The decision of this online learning is transferred to the routing information base (RIB) so that it can accordingly modify the routing table for the entire SRLG upon failure detection of one of the candidate node of that particular SRLG and hence reduce the protection switching time.