Modern cybersecurity systems rely increasingly on machine learning (ML) for threat detection, yet they often fall short in delivering context-specific mitigation strategies. To bridge this gap, we propose an explanation-aware Retrieval-Augmented Generation (RAG) framework that tightly integrates supervised ML-based attack detection with Large Language Model (LLM)-driven mitigation guidance. We propose a Weighted Reciprocal Rank Fusion (WRRF)—a novel ranking method that enhances multi-query retrieval by incorporating retriever-side confidence scores. This ensures that semantically relevant and high-confidence documents from cybersecurity knowledge bases (ENISA, NIST, CISA) are prioritized during response generation. Our system begins by classifying suspicious network traffic using a Random Forest classifier trained on the UNSW-NB15 dataset. It then constructs explanation-rich prompts grounded in key anomalous features to query semantically indexed domain-specific documents. Using a multi-query strategy, the framework retrieves diverse candidate documents, which are then aggregated using WRRF to improve contextual alignment and ranking fidelity. Experimental evaluations across multiple response-generation baselines—including OpenAI, standard RAG, and RRF—demonstrate that WRRF achieves superior performance in mitigation accuracy, semantic relevance to traffic indicators, document source diversity, and response precision.

Intrusion detection in federated learning (FL) is challenged by the presence of adversarial clients who poison model updates to degrade global performance. We address this problem by proposing RAID, a reputation-based aggregation method that robustly combines client contributions in a distributed intrusion detection setting. Unlike standard Federated Averaging (FedAvg), which naively averages all updates (and can be arbitrarily manipulated by even a single malicious client), RAID assigns each client a dynamic trust weight based on the historical consistency of their updates. This approach down-weights anomalous or adversarial gradients while still preserving contributions from honest clients with diverse data. We evaluate RAID on a network intrusion detection dataset under varying poisoning attack rates (up to 40% of clients compromised), comparing against FedAvg and a Median Absolute Deviation (MAD) based robust aggregator. Key results show that RAID consistently outperforms both baselines across all attack levels, maintaining higher accuracy and F1 scores even under aggressive poisoning. For example, at 40% adversarial clients, RAID achieves an F1-0.743, significantly outperforming FedAvg and MAD (F1-0.62). These results demonstrate that RAID substantially improves robustness against poisoning attacks with minimal overhead, making it well-suited for secure federated intrusion detection in adversarial environments.

Incident management in telecommunications networks generates large volumes of incident management tickets (IMTs), each containing heterogeneous and often unstructured text describing service outages, performance degradations, or security issues. Accurately categorizing these IMTs into multiple impact and cause labels is essential for rapid diagnosis and resolution. However, existing rule-based and standard language-model-based approaches struggle with noisy data, overlapping categories, and limited contextual understanding. To address these challenges, we propose two complementary solutions for automated multi-label classification of IMTs. To mitigate the effects of noisy data and overlapping categories, the first solution employs an encoder-based language model (i.e., Bidirectional Encoder Representations from Transformers (BERT)) with a relevance-guided feature selection strategy that focuses on semantically meaningful attributes. To improve contextual understanding and label consistency, the second solution leverages a decoder-based large language model (i.e., Phi-3.5) enhanced with retrieval-augmented generation (RAG) and a novel probabilistic re-ranking mechanism to refine label predictions. Experimental results show that our encoder-only model achieves an F1 score of 79.20%, while our RAG-enhanced decoder model achieves 94.98%, outperforming traditional machine learning models and BERT baselines by 23.59% and 29% on average, respectively. These findings demonstrate that combining fine-tuned language models with intelligent retrieval and re-ranking significantly improves classification accuracy in incident management systems.

As 5G and beyond mobile networks evolve, their increasing complexity necessitates advanced, automated, and datadriven fault diagnosis methods. While traditional data-driven methods falter with modern network complexities, Transformer models have proven highly effective for fault diagnosis through their efficient processing of sequential and time-series data. However, these Transformer-based methods demand substantial labeled data, which is costly to obtain. To address the lack of labeled data, we propose a novel active learning (AL) approach designed for Transformer-based fault diagnosis, tailored to the time-series nature of network data. AL reduces the need for extensive labeled datasets by iteratively selecting the most informative samples for labeling. Our AL method exploits the interpretability of Transformers, using their attention weights to create dependency graphs that represent processing patterns of data points. By formulating a one-class novelty detection problem on these graphs, we identify whether an unlabeled sample is processed differently from labeled ones in the previous training cycle and designate novel samples for expert annotation. Extensive experiments on real-world datasets show that our AL method achieves higher F1-scores than state-of-the-art AL algorithms with 50% fewer labeled samples and surpasses existing methods by up to 150% in identifying samples related to unseen fault types.

As cyber threats continue to evolve, there is a need for autonomous cyber defense (ACD) strategies capable of fast and context-aware responses. Reinforcement learning (RL) has shown promise for automating cyber defense by exploring and learning effective countermeasures, yet it often struggles with sparse reward signals and insufficient context to handle diverse attack scenarios. Furthermore, the convergence time taken by an RL agent is often high, which makes it difficult to train the RL agent in online settings. To address these challenges, we propose a large language model (LLM)-enhanced RL method that builds and queries a knowledge graph (KG) derived from agent-environment interactions. We leverage the pre-trained knowledge of an LLM on different cybersecurity frameworks and use the LLM to analyze a part of the KG to generate appropriate actions for the RL agent. We infuse the knowledge extracted from the LLM into the RL agent’s training loop in two ways. First, the state vector of the RL agent is augmented with the most effective action and its corresponding reward, as determined from the KG. Second, the suggested action from the LLM is used as a reference policy. In addition, we introduce a regularization term in the loss function to make the RL policy close to the reference policy. To validate our approach, we develop a custom RL environment guided by the MITRE ATT&CK framework, enabling the agent to generate tailored mitigation strategies for detected cyber attacks. Experimental results show that our proposed approach significantly outperforms the baseline RL by over $75 \%$ in terms of taking better mitigation actions.

The integration of Network Functions Virtualization (NFV) systems into mobile edge and core networks has heightened the need for effective anomaly detection and localization methods. The complexity of NFV demands robust mechanisms for network resilience, security, and performance. Machine Learning approaches have demonstrated promising solutions in crafting adaptive and efficient mechanisms for detecting and localizing potential anomalies within NFV systems. Particularly, Unsupervised Learning (UL) methods have garnered significant attention for their potential to detect anomalies without the need for labeled data. However, UL methods are susceptible to even minor levels of anomalous samples in the training data, termed contamination, which can severely compromise their performance. This paper proposes a novel approach using the Noisy-Student technique for anomaly detection. It addresses data contamination by combining a density-estimation teacher model for pseudo-labeling with a weakly-supervised student model based on a Masked Autoencoder trained on the pseudo-labeled data. For anomaly localization, we introduce a heuristic tailored for our anomaly detection model and two Explainable Artificial Intelligence (XAI)-based approaches applicable to any detection model. Extensive experiments on three NFV datasets demonstrate superior performance, with up to a 20% improvement in anomaly detection and up to a 22% improvement in localization, in terms of F1-score.

Data-driven algorithms play a pivotal role in the automated orchestration and management of network slices in 5G and beyond networks, however, their efficacy hinges on the timely and accurate monitoring of the network and its components. To support 5G slicing, monitoring must be comprehensive and encompass network slices end-to-end (E2E). Yet, several challenges arise with E2E network slice monitoring. Firstly, existing solutions are piecemeal and cannot correlate network-wide data from multiple sources (e.g., different network segments). Secondly, different slices can have different requirements regarding Key Performance Indicators (KPIs) and monitoring granularity, which necessitates dynamic adjustments in both KPI monitoring and data collection rates in real-time to minimize network resource overhead. To address these challenges, in this paper, we present Monarch, a scalable monitoring architecture for 5G. Monarch is designed for cloud-native 5G deployments and focuses on network slice monitoring and per-slice KPI computation. We validate the proposed architecture by implementing Monarch on a 5G network slice testbed, with up to 50 network slices. We exemplify Monarch’s role in 5G network monitoring by showcasing two scenarios: monitoring KPIs at both slice and network function levels. Our evaluations demonstrate Monarch’s scalability, with the architecture adeptly handling varying numbers of slices while maintaining consistent ingestion times between 2.25 to 2.75 ms. Furthermore, we showcase the effectiveness of Monarch’s adaptive monitoring mechanism, exemplified by a simple heuristic, on a real-world 5G dataset. The adaptive monitoring mechanism significantly reduces the overhead of network slice monitoring by up to 76% while ensuring acceptable accuracy.

In the complex landscape of modern networks, the necessity of Intrusion Detection System (IDS) has become paramount. An IDS is a crucial cybersecurity tool that plays a pivotal role in safeguarding networks against a wide array of threats and attacks. The application of deep learning models for intrusion detection is becoming popular among research communities due to its success in many other domains. However, deep learning models require a significant amount of labeled data to achieve effective training. Obtaining labeled data for intrusion detection can be challenging and costly. To address it, Deep Transfer Learning (DTL) can be employed. This research introduces an innovative traffic classification method tailored for 5G networks. The approach leverages deep transfer learning by utilizing pre-trained models and fine-tuning them. We evaluate several deep-learning models in a transfer learning setting. The Inception model being identified as the top-performing model shows an improvement of approximately 10% in terms of F1-score between IDS-based DTL and the same scheme without DTL.

In the rapidly evolving landscape of 5G and its successor technologies, the Next Generation Radio Access Network (NG-RAN) stands out as a transformative pillar. Functional splitting, a core concept in NG-RAN, splits the traditional base station into distinct functional entities, notably the Distributed Unit (DU), Centralized Unit (CU) and Radio Unit (RU). With flexible functional splitting, Infrastructure Providers (InPs) can dynamically allocate RAN resources to cater to each network slice's distinct throughput and latency demand. However, the problem of optimally selecting functional splits, placement of RAN functions in DU/CU with constrained computational capacities and determining routing paths present an NP-hard challenge. The coexistence of multiple slices on shared infrastructure may necessitate slice isolation for security, performance, and operational reasons, adding another layer of complexity. To address this multifaceted problem, we formulate an Integer Linear Programming (ILP) model that seeks to maximize the InP profit considering computation, virtual machine instantiation and routing costs. Using Gurobi optimizer, we show that optimal slice admission solutions directly impact InP profit and that enhanced computational capacities can increase the number of slices admitted.

Changes are an integral part of the day-to-day operation of large telecommunications networks as they allow to keep pace with technological advancements, meet growing network demands, ensure scalability, enhance security, improve service quality, and meet customer expectations. Changing configurations, installing devices, and migrating traffic are some examples of these changes. These changes are documented by opening tickets through a ticket management system. Automation in the ticket management system is now becoming highly desirable to manage the large number of submitted tickets. An automated ticket management system supports the management of a ticket by automating several parts of a ticket's lifecycle. In this context, ticket classification problem consists in assigning an appropriate label to a ticket to be utilized in the later stages of the ticket management cycle. In this paper, we use a collection of network-change tickets from a real network operator to solve a ticket classification problem. We observe that the network-change ticket dataset is highly skewed in the number of tickets for different possible classes. We address this challenge of classification in a highly imbalanced dataset by proposing two token-prioritization strategies along with other components. We compare three variations of our proposed approach with three methods from the literature and show that the variations of the proposed approach outperform existing methods by up to 7% in terms of F1 score.

The Fifth generation (5G) mobile network is expected to provide high bandwidth, low latency, and rapid user connectivity. 5G Mobile operators are seeking an effective solution that would enable them to support heterogeneous use cases with different Quality of Service (QoS) requirements by utilizing the existing physical infrastructure. 5G supports Network Slicing (NS), an end-to-end (E2E) logical network that is mutually isolated, has independent control, and can be managed independently. By slicing the network, mobile operators can effectively manage several network instances over a single infrastructure to provide a variety of applications, use cases, and business services while satisfying heterogeneous QoS requirements. With the advancement of Machine Learning (ML), future communication networks will need to use data-driven decision-making to achieve desired network performance. In this paper, we demonstrated a prediction mechanism using Machine and Deep Learning Algorithms in traditional and incremental ways to select the suitable network slice for various user requirements and device types. Using a publicly available dataset and Incremental Learning model called Stochastic Gradient Descent (SGD), we successfully classified incoming user requests to appropriate network slices with an accuracy of 99.33%.

Automated decision making algorithms are expected to play a key role in management and orchestration of network slices in 5G and beyond networks. State-of-the-art algorithms for automated orchestration and management tend to rely on data-driven methods which require a timely and accurate view of the network. Accurately monitoring an end-to-end (E2E) network slice requires a scalable monitoring architecture that facilitates collection and correlation of data from various network segments comprising the slice. The state-of-the-art on 5G monitoring mostly focuses on scalability, falling short in providing explicit support for network slicing and computing network slice key performance indicators (KPIs). To fill this gap, in this paper, we present MonArch, a scalable monitoring architecture for 5G, which focuses on network slice monitoring, slice KPI computation, and an application programming interface (API) for specifying slice monitoring requests. We validate the proposed architecture by implementing MonArch on a 5G testbed, and demonstrate its capability to compute a network slice KPI (e.g., slice throughput). Our evaluations show that MonArch does not significantly increase data ingestion time when scaling the number of slices and that a 5-second monitoring interval offers a good balance between monitoring overhead and accuracy.

An intrusion detection system (IDS) is an essential component of any modern network. The purpose of an IDS is to detect intrusion and generate appropriate alarms so that the intrusion can be mitigated. Implementing an IDS in a Software Defined Network (SDN) is easier since an SDN controller has a centralized view of the whole network. Researchers have made many efforts to use machine learning (ML) for developing network-based IDS in SDN. The network-based IDS analyzes different characteristics of incoming network traffic to detect intrusion. Early detection of intrusion is crucial for an IDS because if the intrusion is not detected quickly enough, it can cause severe damage, such as data breaches and service shutdowns. This paper focuses on detecting intrusion in SDN as early as possible using real-time flow-based features. Our aim is to detect intrusion with less amount of packets per flow, which not only facilitates early intrusion detection but also is useful when an intrusion flow has less number of packets. We show that although ML models perform well in offline training on a dataset, their performance decreases ~25% when fewer packets are used to generate features for the ML model. In all our experiments, a simple Random Forest (RF) algorithm outperforms a complex deep learning model on a publicly available dataset for intrusion detection in SDN.

5G transport networks will support dynamic services with diverse requirements through network slicing. Elastic Optical Networks (EONs) facilitate transport network slicing by flexible spectrum allocation and tuning of transmission configurations. A major challenge in supporting dynamic services is the lack of priori knowledge of future slice requests. As a consequence, slice embedding can become sub-optimal over time, leading to spectrum fragmentation and skewed utilization. This in turn can block future slice requests, impacting operator revenue. To address this issue, operators can periodically re-optimize slice embedding for reducing fragmentation. In this paper, we address this problem of re-optimizing network slice embedding on EONs for minimizing fragmentation. The problem is solved in its splittable version, which significantly increases problem complexity, but also offers more opportunities for a larger set of re-configuration actions. We employ simulated annealing for systematically exploring the large solution space. We also propose a greedy algorithm to address the practical constraint of limiting the number of re-configuration steps. Moreover, we present a novel method based on Deep Reinforcement Learning (DRL) for determining when performing re-configuration is most effective. Our extensive simulations demonstrate that the greedy algorithm yields a solution very close to that obtained using simulated annealing while requiring orders of magnitude lesser re-configuration actions. Finally, we show that by applying the greedy algorithm periodically on the network according to the DRL-based time selection algorithm, a significant improvement in the total number of accepted slice requests can be achieved with only performing a limited number of re-configuration operations.

Network traffic classification has a huge application in software-defined networking (SDN) where we talk about more control over the network traffic. With the increase of encrypted protocols in the network, the problem of traffic classification has become extremely challenging. Many researchers have proposed different techniques to do traffic classification. This demo paper presents an application of our proposed method for traffic classification in an SDN environment. The proposed method leverages one of the self-supervised learning approaches, an emerging field of deep learning, to classify network traffic. This paper shows that the proposed method can outperform the corresponding supervised approach by $\sim 2$% in terms of accuracy using data collected from an SDN testbed. Furthermore, an SDN application is developed to show that the trained model is able to classify real-time traffic.

Network traffic classification is used in many applications including network provisioning, malware detection, resource management, and so on. In modern networks, use of encrypted protocols is a norm rather than an exception. Existing network traffic classification techniques fall short in working with encrypted traffic. Although deep learning based techniques have been shown to perform well in the case of encrypted traffic classification, they require an abundance of labeled data to achieve high accuracy. However, labeled data is rarely available in sufficient volumes in real network settings as they require domain experts to annotate data with labels. Therefore, in this paper, we propose a self-supervised approach that can achieve high accuracy on encrypted network traffic classification with a few labeled data. The proposed method is evaluated on three publicly available datasets. The empirical result shows that our method not only achieves high accuracy on encrypted traffic but also has the ability to apply the acquired knowledge on a different dataset. In our experiments, our method outperforms the state-of-the-art baseline methods by ~3% in terms of accuracy even with a much lower volume of labeled data.

Due to the scarcity of labeled faulty data, Unsupervised Learning (UL) methods have gained great traction for anomaly detection and localization in Network Functions Virtualization (NFV) systems. In a UL approach, training is performed on only normal data for learning normal data patterns, and deviation from the norm is considered as an anomaly. However, it has been shown that even small percentages of anomalous samples in the training data (referred to as contamination) can significantly degrade the performance of UL methods. To address this issue, we propose an anomaly-detection approach based on the Noisy-Student technique, which was originally introduced for leveraging unlabeled datasets in computer-vision classification problems. Our approach not only provides robustness against training-data contamination, but also can leverage this contamination to improve anomaly-detection accuracy. Moreover, after an anomaly is detected, localization of the anomalous virtualized network functions in an unsupervised manner is a challenging task in the absence of labeled data. For anomaly localization in NFV systems, we propose to exploit existing local AI-explainability methods to achieve a high localization performance and propose our own novel AI-explainability method, specifically designed for the anomaly-localization problem in NFV, to improve the performance further. We perform a comprehensive experimental analysis on two datasets collected on different NFV testbeds and show that our proposed solutions outperform the existing methods by up to 22% in anomaly detection and up to 19% in anomaly localization in terms of F1-score.

Network slicing has been envisaged as a key enabler to satisfy diverse requirements of 5G networks, by creating multiple isolated end-to-end virtual networks dedicated to different services. An accurate view of these end-to-end 5G network slices is essential for both artificial intelligence (AI) driven slice orchestration, and data-driven automated service assurance. However, the existing open-source implementations of the 5G core do not natively support slice Key Performance Indicator (KPI) monitoring. In this demonstration, we show how to deploy a functional 5G testbed using a combination of open-source frameworks and tools, with a guide for configuring multiple network slices published on GitHub [1]. We also show the feasibility of monitoring and visualizing network slice KPIs using a representative cloud-gaming use-case.

5G networks are expected to support numerous novel services and applications with versatile quality of service (QoS) requirements such as high data rates and low end-to-end (E2E) latency. It is widely agreed that E2E latency can be reduced by moving the computational capability closer to the network edge. The limited amount of computational resources of the edge nodes, however, poses the challenge of efficiently utilizing these resources while, at the same time, satisfying QoS requirements. In this work, we employ mixed-integer linear programming (MILP) techniques to formulate and solve a joint user association, service function chain (SFC) placement, where SFCs are composed of virtualized service functions (VSFs), and resource allocation problem in 5G networks composed of decentralized units (DUs), centralized units (CUs), and a core network (5GC). Specifically, we compare four approaches to solving the problem. The first two approaches minimize, respectively, the E2E latency experienced by users and the service provisioning cost. The other two instead aim at minimizing VSF migrations along with their impact on users’ quality of experience with the last one minimizing also the number of inter-CU handovers. We then propose a heuristic to address the scalability issue of the MILP-based solutions. Simulations results demonstrate the effectiveness of the proposed heuristic algorithm.

5G transport networks will support dynamic services with diverse requirements through network slicing. Elastic Optical Networks (EONs) facilitate transport network slicing by flexible spectrum allocation and tuning of transmission configurations such as modulation format and forward error correction. A major challenge in supporting dynamic services is the lack of a priori knowledge of future slice requests. In consequence, slice embedding can become sub-optimal over time, leading to spectrum fragmentation and skewed utilization. This in turn can block future slice requests, impacting operator revenue. Therefore, operators need to periodically re-optimize slice embedding for reducing fragmentation. In this paper, we address this problem of re-optimizing network slice embedding on EONs for minimizing fragmentation. The problem is solved in its splittable version, which significantly increases problem complexity, but offers more opportunities for a larger set of re-configuration actions. We employ simulated annealing for systematically exploring the large solution space. We also propose a greedy algorithm to address the practical constraint to limit the number of re-configuration steps taken to reach a defragmentated state. Our extensive simulations demonstrate that the greedy algorithm yields a solution very close to that obtained using simulated annealing while requiring orders of magnitude lesser number of re-configuration actions.

Elastic Optical Networks (EONs) enable finer-grained resource allocation and tuning of transmission configurations for right-sized resource allocation. These features make EONs excellent choice for 5G transport networks supporting highly dynamic traffic with diverse Quality-of-Service (QoS) requirements. 5G network slices are expected to host applications with a dynamic nature (e.g., augmented/virtual reality broadcasting), which will result in slice resource requirement changing over time. The initial resource allocation to network slices has to be adapted to accommodate such changes without causing significant disruption to existing traffic and using minimal additional resources. In this paper, we address the problem of scaling bandwidth demand of network slices on an EON-enabled 5G transport network. In contrast to the state-of-the-art, we do not assume any specific technologies for minimizing disruption when accommodating the scaling request. Rather, we propose an Integer Linear Program (ILP) and a heuristic algorithm for accommodating scaling requests by choosing from a comprehensive set of reconfiguration actions. We carefully design a novel cost model for capturing traffic disruptions and additional resource usage by these different actions. Our extensive simulations using realistic network topologies shed light on the trade-off between additional resource usage and disruption while accommodating slice scaling requests by employing a comprehensive set of reconfiguration actions. Simulation results also show that our heuristic algorithm can find solutions that remain within 10% of ILP-based solutions, while executing several orders of magnitude faster than ILP.

Modern network services are constantly increasing their requirements in terms of bandwidth, latency and cost efficiency. To satisfy these requirements, the concept of network slicing has been introduced in the context of next-generation 5G networks. However, to successfully provision resources to slices, a complex optimization problem must be addressed to allocate resources over a cloud network, i.e., a distributed computing infrastructure interconnected through high-capacity network links. In this study, we propose two new latency and energy-aware optimization models for provisioning 5G slices in cloud networks comprising both distributed computing and network resources. The proposed approaches differ from other existing solutions since we conduct our studies with respect to the end-to-end latency. Relevant models of latency and energy consumption are proposed based on a comprehensive review of the state-of-the-art. To effectively solve those optimization problems, a configurable heuristic is also proposed and investigated over different network topologies. Performance of the proposed heuristic is compared against near-optimal solutions. Moreover, we assess the importance of matching between resource provisioning algorithms and architectural assumptions related to 5G network slices and a proper problem modeling.

This paper addresses Connectivity-aware Virtual Network Embedding (CoViNE) problem, which consists in embedding a virtual network (VN) on a substrate network while ensuring VN connectivity (without any bandwidth guarantee) against multiple substrate link failures. CoViNE provides a weaker form of survivability incurring less resource overhead than traditional VN survivability models. To optimally solve CoViNE, we present an Integer Linear Program (ILP), namely CoViNE-opt. CoViNE-opt enumerates an exponential number of edge-cuts in a VN severely limiting its scalability. Therefore, we decompose CoViNE into three sub-problems: i) augmenting a VN with virtual links to provide necessary connectivity, ii) identifying the virtual links that should be embedded disjointly, and iii) computing a VN embedding while satisfying the disjointness constraints. We introduce conflicting set abstraction that allows to address sub-problems (i) and (ii) without enumerating all the edge-cuts of a VN. We propose two novel solutions to CoViNE leveraging conflicting set, namely CoViNE-ILP and CoViNE-fast. CoViNE-ILP uses a heuristic algorithm to address sub-problems (i) and (ii), while an ILP is used for sub-problem (iii). In contrast, CoViNE-fast uses heuristics for solving all three sub-problems. Through simulation, we evaluate the optimality and scalability of our solutions and demonstrate a failure restoration use-case enabled by CoViNE.

5G network slicing allows partitioning of network resources to meet stringent end-to-end service requirements across multiple network segments, from access to transport. These requirements are shaping technical evolution in each of these segments. In particular, the transport segment is currently evolving in the direction of elastic optical networks (EONs), a new generation of optical networks supporting a flexible optical-spectrum grid and novel elastic transponder capabilities. In this paper, we focus on the reliability of 5G transport-network slices in EON. Specifically, we consider the problem of slicing 5G transport networks,i.e., establishing virtual networks on 5G transport, while providing dedicated protection. As dedicated protection requires a large amount of backup resources, our proposed solution incorporates two techniques to reduce backup resources: (i) bandwidth squeezing,i.e., providing a reduced protection bandwidth than the original request; and (ii) survivable multi-path provisioning. We leverage the capability of EONs to fine tune spectrum allocation and adapt modulation format and forward error correction for allocating spectrum resources. Our numerical evaluation over realistic network topologies quantifies the spectrum savings achieved by employing EON over traditional fixed-grid optical networks, and provides new insights on the impact of bandwidth squeezing and multi-path provisioning on spectrum utilization. One key takeaway from our evaluation is that multi-path provisioning can guarantee up to 40% of the bandwidth requested by a VN during failures by provisioning only 10% additional spectrum resources. This also caused VN blocking ratio for BSR up to 40% to remain very close to that of the no-backup case.

5G networks are characterized by massive device connectivity, supporting a wide range of novel applications with their diverse Quality of Service (QoS) requirements. This poses a challenge since 5G as one-fits-all technology has to simultaneously address all these requirements. Network slicing has been proposed to cope with this challenge, calling for efficient slicing and slice placement strategies in order to ensure that the slice requirements (e.g., latency, data rate) are met, while the network resources are utilized in the most optimal manner. In this paper, we compare different end-to-end (E2E) slice placement strategies by formulating and solving a Mixed Integer Linear Programming (MILP) slice placement problem and study their trade-offs. E2E slice requests are modelled as Service Functions Chains (SFC), in which each core network and radio access network component is represented as a Virtual Network Function (VNF). Based on the analysis of the results, we then propose a slice placement heuristic algorithm whose objective is to minimize the number of VNF migrations in the network and their impact onto the slices while, at the same time, optimizing the network utilization and making sure that the QoS requirements of the considered slice requests are satisfied. The results of the simulations demonstrate the efficiency of the proposed algorithm.

In 5G networks, slicing allows partitioning of network resources to meet stringent end-to-end service requirements across multiple network segments, from access to transport. These requirements are shaping technical evolution in each of these segments. In particular, the transport segment is currently evolving in the direction of the so-called elastic optical networks (EONs), a new generation of optical networks supporting a flexible optical-spectrum grid and novel elastic transponder capabilities. In this paper, we focus on the reliability of 5G transport-network slices in EON. Specifically, we consider the problem of slicing 5G transport networks, i.e., establishing virtual networks on 5G transport, while providing dedicated protection. As dedicated protection requires a large amount of backup resources, our proposed solution incorporates two techniques to reduce backup resources: (i) bandwidth squeezing, i.e., providing a reduced protection bandwidth with respect to the original request; and (ii) survivable multi-path provisioning. We leverage the capability of EONs to fine tune spectrum allocation and adapt modulation format and Forward Error Correction (FEC) for allocating rightsize spectrum resources to network slices. Our numerical evaluation over realistic case-study network topologies quantifies the spectrum savings achieved by employing EON over traditional fixed-grid optical networks, and provides new insights on the impact of bandwidth squeezing and multi-path provisioning on spectrum utilization.

Elastic Optical Network (EON) virtualization has recently emerged as an enabling technology for 5G network slicing. A fundamental problem in EON slicing (known as Virtual Network Embedding (VNE)) is how to efficiently map a virtual network (VN) on a substrate EON characterized by elastic transponders and flexible grid. Since a number of 5G services will have strict latency requirements, the VNE problem in EONs must be solved while guaranteeing latency targets. In existing literature, latency has always been modeled as a constraint applied on the virtual links of the VN. In contrast, we argue in favor of an alternate modeling that constrains the latency of virtual paths. Constraining latency over virtual paths (vs. over virtual links) poses additional modeling and algorithmic challenges to the VNE problem, but allows us to capture end-to-end service requirements. In this paper, we first model latency in an EON by identifying the different factors that contribute to it. We formulate the VNE problem with latency guarantees as an Integer Linear Program (ILP) and propose a heuristic solution that can scale to large problem instances. We evaluated our proposed solutions using real network topologies and realistic transmission configurations under different scenarios and observed that, for a given VN request, latency constraints can be guaranteed by accepting a modest increase in network resource utilization. Latency constraints instead showed a higher impact on VN blocking ratio in dynamic scenarios.

Network operators must continuously scale the capacity of their optical backbone networks to keep apace with the proliferation of bandwidth-intensive applications. Today’s optical networks are designed to carry large traffic aggregates with coarse-grained resource allocation, and are not adequate for maximizing utilization of the expensive optical substrate. Elastic Optical Network (EON) is an emerging technology that facilitates flexible allocation of fiber spectrum by leveraging finer-grained channel spacing, tunable modulation formats and Forward Error Correction (FEC) overheads, and baud-rate assignment, to right size spectrum allocation to customer needs. Virtual Network Embedding (VNE) over EON has been a recent topic of interest due to its importance for 5G network slicing. However, the problem has not yet been addressed while simultaneously considering the full flexibility offered by an EON. In this paper, we present an optimization model that solves the VNE problem over EON when lightpath configurations can be chosen among a large (and practical) set of combinations of paths, modulation formats, FEC overheads and baud rates. The VNE over EON problem is solved in its splittable version, which significantly increases problem complexity, but is much more likely to return a feasible solution. Given the intractability of the optimal solution, we propose a heuristic to solve larger problem instances. Key results from extensive simulations are: (i) a fully-flexible VNE can save up to 60% spectrum resources compared to that where no flexibility is exploited, and (ii) solutions of our heuristic fall in more than 90% of the cases, within 5% of the optimal solution, while executing several orders of magnitude faster.

The 5th generation mobile network (5G) is expected to support numerous services with versatile quality of service (QoS) requirements such as high data rates and low end-to-end (E2E) latency. It is widely agreed that E2E latency can be significantly reduced by moving content/computing capability closer to the network edge. However, since the edge nodes (i.e., base stations) have limited computing capacity, mobile network operators shall make a decision on how to provision the computing resources to the services in order to make sure that the E2E latency requirement of the services are satisfied while the network resources (e.g., computing, radio, and transport network resources) are used in an efficient manner. In this work, we employ integer linear programming (ILP) techniques to formulate and solve a joint user association, service function chain (SFC) placement, and resource allocation problem where SFCs, composed of virtualized service functions (VSFs), represent user requested services that have certain E2E latency and data rate requirements. Specifically, we compare three variants of an ILP-based algorithm that aim to minimize E2E latency of requested services, service provisioning cost, and VSF migration frequency, respectively. We then propose a heuristic in order to address the scalability issue of the ILP-based solutions. Simulations results demonstrate the effectiveness of the proposed heuristic algorithm.

A key challenge in network virtualization is to efficiently map a virtual network (VN) on a substrate network (SN), while accounting for possible substrate failures. This is known as the survivable VN embedding (SVNE) problem. The state-of-the-art literature has studied the SVNE problem from infrastructure providers' (InPs') perspective, i.e., provisioning backup resources in the SN. A rather unexplored solution spectrum is to augment the VN with sufficient spare backup capacity to survive substrate failures and embed the resulting VN accordingly. Such augmentation enables InPs to offload failure recovery decisions to the VN operator, thus providing more flexible VN management. In this paper, we study the problem of jointly optimizing spare capacity allocation in a VN and embedding the VN to guarantee full bandwidth in the presence of multiple substrate link failures. We formulate the optimal solution to this problem as a quadratic integer program that we transform into an integer linear program. We also propose a heuristic algorithm to solve larger instances of the problem. Based on analytical study and simulation, our key findings are: 1) provisioning shared backup resources in the VN can yield ~33% more resource efficient embedding compared to doing the same at the SN level and 2) our heuristic allocates ~21% extra resources compared to the optimal, while executing several orders of magnitude faster.

Network virtualization (NV), considered as a key enabler for overcoming the ossification of the Internet allows multiple heterogeneous virtual networks to co-exist over the same substrate network. Resource allocation problems in NV have been extensively studied for single layer substrates such as IP or Optical networks. However, little effort has been put to address the same problem for multi-layer IP-over-optical networks. The increasing popularity of multi-layer networks for deploying backbones combined with their unique characteristics ( e.g., topological flexibility of the IP layer) calls for the need to carefully investigate the resource provisioning problems arising from their virtualization. In this paper, we address the problem of multi-layer virtual network embedding (MULE; similar to multi-layer networks, this hybrid species brings the best of two species together.) on IP-over-optical networks. We propose two solutions to MULE: 1) an integer linear program formulation for the optimal solution (OPT-MULE) and 2) a heuristic to address the computational complexity of the optimal solution (FAST-MULE). We demonstrate through extensive simulations that on average our heuristic performs within $\boldsymbol \approx 1.47\boldsymbol \times $ of optimal solution while executing several orders of magnitude faster. Simulation results also show that FAST-MULE incurs ≈66% less cost on average than the state-of-the-art heuristic while accepting ≈60% more virtual network requests on average.

Network Virtualization (NV), considered as a key enabler for overcoming the ossification of the Internet allows multiple heterogeneous virtual networks to co-exist over the same substrate network. Resource allocation problems in NV have been extensively studied for single layer substrates such as IP or Optical networks. However, little effort has been put to address the same problem for multi-layer IP-over-Optical networks. The increasing popularity of multi-layer networks for deploying backbones combined with their unique characteristics (e.g., topological flexibility of the IP layer) calls for the need to carefully investigate the resource provisioning problems arising from their virtualization. In this paper, we address the problem of MUlti-Layer virtual network Embedding (MULE) on IP-overOptical networks. We propose two solutions to MULE: an Integer Linear Program (ILP) formulation for the optimal solution and a heuristic to address the computational complexity of the optimal solution. We demonstrate through extensive simulations that on average our heuristic performs within ≈1.47 × of optimal solution and incurs ≈66% less cost than the state-of-the-art heuristic.

Perceived as a key enabling technology for the future Internet, Network Virtualization (NV) allows an Infrastructure Provider (InP) to better utilize their Substrate Network (SN) by provisioning multiple Virtual Networks (VNs) from different Service Providers (SPs). A key challenge in NV is to efficiently map the VN requests from SPs on an SN, known as the Virtual Network Embedding (VNE) problem. VNE algorithms are typically online in nature. A VN embedding can become suboptimal over time due to the arrival and departure of other VNs as well as due to changes in SN such as failures. One way to mitigate the impact of such dynamism is to periodically reallocate resources for the existing VNs. VNE reallocation can increase an InP's revenue by decreasing bandwidth consumption and by increasing the possibility of accepting future VNs. In this paper, we study Reallocation of Virtual Network Embedding (ReViNE) problem to minimize the number of over utilized substrate links and total bandwidth cost on the SN. We propose an Integer Linear Programming formulation for the optimal solution (ReViNE-OPT) and a simulated annealing based heuristic (ReViNE-FAST) to solve larger problem instances. Simulation results show that on average our proposed heuristic performs within ∼19% of the optimal solution. Moreover, ReViNE-FAST generates more than 2.5× better solutions compared to the state-of-the-art simulated annealing based heuristic for VNE reallocation.

A service-function chain, or simply a chain, is an ordered sequence of service functions, e.g., firewalls and load balancers, composing a service. A chain deployment involvesselectingand instantiating a number of virtual network functions (VNFs), i.e., softwarized service functions,placingVNF instances, androutingtraffic through them. In the current optimization-models of a chain deployment, the instances of the same function are assumed to be identical, while typical service providers offer VNFs with heterogeneous throughput and resource configurations. The VNF instances of the same function are installed in a single physical machine, which limits a chain to the throughput of a few instances that can be installed in one physical machine. Furthermore, theselection,placement, androutingproblems are solved in isolation. We present distributed service function chaining that coordinates these operations, places VNF-instances of the same functiondistributedly, and selects appropriate instances from typical VNF offerings. Such a deployment uses network resources more efficiently and decouples a chain’s throughput from that of physical machines. We formulate this deployment as a mixed integer programming (MIP) model, prove its NP-Hardness, and develop a local search heuristic called Kariz. Extensive experiments demonstrate that Kariz achieves a competitive acceptance-ratio of 76%–100% with an extra cost of less than 24% compared with the MIP model.

Network virtualization has evolved as a key enabling technology for offering the next generation network services. Recently, it is being rolled out in data center networks as a means to provide bandwidth guarantees to cloud applications. With increasing deployments of virtual networks (VNs) in commercial-grade networks with commodity hardware, VNs need to tackle failures in the underlying substrate network. In this paper, we study the problem of recovering a batch of VNs affected by a substrate node failure. The combinatorial possibilities of alternate embeddings of the failed virtual nodes and links of the VNs make the task of finding the most efficient recovery both non-trivial and intractable. Furthermore, any recovery approach ideally should not cause any service disruption for the unaffected parts of the VNs. We take into account these issues to design a generalized recovery approach that can achieve customized objectives such as fair treatment on the failed VNs, partial treatment based on priority, and so on. We provide integer linear programming (ILP) formulations for two variants of our recovery scheme, namely, fair recovery model and priority-based recovery model. We also propose a fast and scalable heuristic algorithm to tackle the computational complexity of the ILP solution. Evaluation results demonstrate that our heuristic performs close to the optimal solution and outperforms the state-of-the-art algorithm.

Network visualization (NV) has evolved as a key enabling technology for offering the next generation network services. Recently, it is being rolled out in data center networks as a means to provide bandwidth guarantees to cloud applications. With increasing deployments of virtual networks (VNs) in commercial-grade networks with commodity hardware, VNs need to tackle failures in the underlying substrate network. In this paper, we study the problem of recovering a batch of VNs affected by a substrate node failure. The combinatorial possibilities of alternate embeddings of the failed virtual nodes and links of the VNs makes the task of finding the most efficient recovery both non-trivial and intractable. Furthermore, any recovery approach ideally should not cause any service disruption for the unaffected parts of the VNs. We take into account these issues to design a recovery approach for maximizing recovery and minimizing the cost of recovery and network disruption. We provide an Integer Linear Programming (ILP) formulation of our recovery scheme. We also propose a fast and scalable heuristic algorithm to tackle the computational complexity of the ILP solution. Evaluation results demonstrate that our heuristic performs close to the optimal solution and outperforms the state-of-the-art algorithm.

Network virtualization is enabling infrastructure providers (InPs) to offer new services to higher level service providers (SPs). InPs are usually bound by Service Level Agreements (SLAs) to ensure various levels of resource availability for different SPs' virtual networks (VNs). They provision redundant backup resources while embedding an SP's VN request to conform to the SLAs during physical failures in the infrastructure. An extreme of this backup resource provisioning is to reserve a dedicated backup of each element in an SP's VN request. Such dedicated protection scheme can enable an InP to ensure fast VN recovery, thus, providing high uptime guarantee to the SPs. In this paper, we study the 1 + 1-Protected Virtual Network Embedding (1 + 1-ProViNE) problem. We propose Dedicated Protection for Virtual Network Embedding (DRONE), a suite of solutions to the 1 + 1-ProViNE. DRONE includes an Integer Linear Programming (ILP) formulation for optimal solution (OPT-DRONE) and a heuristic (FAST-DRONE) to tackle the computational complexity in computing the optimal solution. Trace driven simulations show that FAST-DRONE allocates only 14.3% extra backup resources on average compared to the optimal solution, while executing 200-12000x faster.

Network virtualization is enabling infrastructure providers (InPs) to offer new services to service providers (SPs). InPs are usually bound by service level agreements to ensure various levels of resource availability for different SPs' virtual networks (VNs). They provision redundant backup resources while embedding an SP's VN request to conform to the SLAs during physical failures in the infrastructure. An extreme backup resource provisioning is to reserve a mutually exclusive backup of each element in an SP's VN request. Such dedicated protection scheme can enable an InP to ensure fast VN recovery, thus, providing high uptime guarantee to the SPs. In this paper, we study the 1 + 1-Protected Virtual Network Embedding (1 + 1-ProViNE) problem. We propose Dedicated Protection for Virtual Network Embedding (DRONE), a suite of solutions to the 1 + 1-ProViNE problem. DRONE includes an integer linear programming formulation for optimal solution (OPT-DRONE) and a heuristic (FAST-DRONE) to tackle the computational complexity of the optimal solution. Trace driven simulations show that FAST-DRONE allocates only 14.3% extra backup resources on average compared to the optimal solution, while executing 200-1200 times faster. Simulation results also show that FAST-DRONE can accept four times more VN requests on average compared to the state-of-the-art solution for providing dedicated protection to VNs.

Internet applications are deployed on the same network infrastructure, yet they have diverse performance and functional requirements. The Internet was not originally designed to support the diversity of current applications. Network virtualization enables heterogeneous applications and network architectures to coexist without interference on the same infrastructure. Embedding a virtual network (VN) into a physical network is a fundamental problem in network virtualization. A VN embedding that aims to survive physical (e.g., link) failures is known as the survivable VN embedding (SVNE). A key challenge in the SVNE problem is to ensure VN survivability with minimal resource redundancy. To address this challenge, we propose survivability in multi-path link embedding (SiMPLE). By exploiting path diversity in the physical network, SiMPLE provides guaranteed VN survivability against single link failure while incurring minimal resource redundancy. In case of multiple arbitrary link failures, SiMPLE provides maximal survivability to the VNs. We formulate this problem as an integer linear program and implement it using GNU linear programming kit. We propose a greedy proactive approach to solve larger instances of the problem in case of single link failures. In presence of more than one link failures, we propose a greedy reactive algorithm as an extension to the previous one, which opportunistically recovers the lost bandwidth in the VNs. Simulation results show that SiMPLE outperforms full backup and shared backup schemes for SVNE, and produces near-optimal results.

Internet applications are deployed on the same network infrastructure, yet they have diverse performance and functional requirements. The Internet was not originally designed to support the diversity of current applications. Network Virtualization can enable heterogeneous applications and network architectures to coexist without interference on the same infrastructure. Embedding a Virtual Network (VN) into a physical network is a fundamental problem in Network Virtualization. A VN Embedding that aims to survive physical (e.g., link) failures is known as the Survivable Virtual Network Embedding (SVNE). A key challenge in the SVNE problem is to ensure VN survivability with minimal resource redundancy. To address this challenge, we propose SiMPLE. By exploiting path diversity in the physical network, SiMPLE provides guaranteed VN survivability against single link failure. In addition, SiMPLE produces highly surviv-able VN embeddings in presence of multiple link failures while incurring very low resource redundancy. We provide an ILP formulation for this problem and implement it using GLPK. We also propose a greedy algorithm to solve larger instances of the problem. Simulation results show that our solution outperforms full backup and shared backup schemes for SVNE, and produces near-optimal results.

Ensuring content availability in a persistent manner is essential for providing any consistent service over peer-to-peer (P2P) systems. This paper introduces an efficient protocol, called DATA, to design highly available P2P systems irrespective of peer uptime and churn. Our approach utilizes the diurnal pattern of globally dispersed peers to develop a grouping strategy where each group aims to ensure 24 × 7 data availability within the group. Simulation results reveal that our protocol converges fast and ensures high availability for each group with minimal overhead.

Several protocols for ad hoc network try to reduce redundancy as an effective measure against broadcast problems. Though these protocols ensure good performance in a favorable environment, they perform poorly when node cooperation cannot be guaranteed due to intentional misbehavior or environmental hostility. As a result, the expected behavior of nodes to forward packets, which is the basic assumption of all broadcast approaches, cannot be achieved always. In this paper, we analyze the performance deterioration of these algorithms in hostile environment. As a remedy, we focus on the reverse direction and interestingly find that adding redundancy in a controlled manner can greatly compensate the performance loss due to node misbehavior. Here we propose a novel approach that tune the amount of redundancy so that reachability and routing load both remain at a satisfactory level. Comparing their relative performance we end up with the conclusion that though redundancy is undesired, controlled redundancy is effective in special situations like uncooperative environments.