Deterministic Network Calculus (DNC) provides a rigorous algebra for worst-case performance analysis of networks. It allows researchers to compute bounds on worst-case characteristics of systems through algebraic expressions that may appear simple on paper but often require heavy computations, making software support essential. Libraries such as Nancy offer a rich API for DNC computations in C#, but they require programming expertise. In contrast, RTaW’s min-plus playground (MPPG) provides a simpler, calculator-like syntax for DNC computations; however, it is proprietary and web-hosted, limiting offline use and reproducibility. We present nancy-playground, an open-source, locally runnable console that implements the same MPPG syntax while executing computations through the Nancy library. This tool enables reproducible scripting, interactive exploration, and a seamless transition to full programs by converting MPPG scripts into C# code. In this paper, we describe the design and implementation of nancy-playground, as well as its main features for researchers and practitioners in the DNC community.

Digital Twins (DTs) have been proposed as digital replicas of physical entities (e.g., manufacturing equipment), which one can observe in real-time and interact with. Digital Twins of Networks (DTNs) are increasingly being discussed in the literature, as an enabler for efficient data-driven network management and performance-driven network optimization (e.g., to support dynamic reconfiguration, or anticipate the effects of faults). A DTN includes service mapping models, i.e. models that can be fed with acquired data to produce insight on the network itself - e.g., to run what-if scenarios, based on multiple underlying technologies, from Machine Learning to analytical models, e.g. Markov Chains. In this paper we examine the case of DTNs of mobile networks, DTMNs, tailored to 5G and beyond, where issues of dynamic reconfiguration and fault anticipation are critical. We argue that simulation services should be offered by the DTMN in order to allow performance-driven network optimization, and that discrete-event network simulators are ideal instruments to be employed for this purpose. We discuss the challenges that need be addressed to make this happen, e.g., centralized vs. distributed implementation, gathering input from the physical network, security issues and hosting, and we review the possibilities offered by network simulation in terms of what-if analysis, defining the concepts of lockstep and branching analysis. We present a framework to endow a DTMN with simulation services and we exemplify it using Simu5G, a popular 5G/B5G simulation library for OMNeT++, as a reference case study.

The quest for trustworthiness in Artificial Intelligence (AI) is increasingly urgent, especially in the field of next-generation wireless networks. Future Beyond 5G (B5G)/6G networks will connect a huge amount of devices and will offer innovative services empowered with AI and Machine Learning tools. Nevertheless, private user data, which are essential for training such services, are not an asset that can be unrestrictedly shared over the network, mainly because of privacy concerns. To overcome this issue, Federated Learning (FL) has recently been proposed as a paradigm to enable collaborative model training among multiple parties, without any disclosure of private raw data. However, the initiative to natively integrate FL services into mobile networks is still far from being accomplished. In this paper we propose a novel FL-as-a-Service framework that provides the B5G/6G network with flexible mechanisms to allow end users to exploit FL services, and we describe its applicability to a Quality of Experience (QoE) forecasting service based on a vehicular networking use case. Specifically, we show how FL of eXplainable AI (XAI) models can be leveraged for the QoE forecasting task, and induces a benefit in terms of both accuracy, compared to local learning, and trustworthiness, thanks to the adoption of inherently interpretable models. Such considerations are supported by an extensive experimental analysis on a publicly available simulated dataset. Finally, we assessed how the learning process is affected by the system deployment and the performance of the underlying communication and computation infrastructure, through system-level simulations, which show the benefits of deploying the proposed framework in edge-based environments.

Electronic payment methods have become increasingly popular for business transactions, both online and in-person, across the globe. Anomalies like online fraud and default payments, which can result in substantial financial losses, have become more common as the usage of credit cards in online purchases has increased. To address this issue, researchers have explored various machine learning models and their ensemble techniques for detecting anomalies in credit card transaction data. However, detecting anomalies in this data can be challenging due to overlapping class samples and an imbalanced class distribution. Therefore, the detection rate of anomalies from minority class samples is relatively low, and general learning algorithms can be biased towards the majority class samples. In this paper, we propose a model called Credit Card Anomaly Detection (CCAD) that leverages the base learners paradigm and meta-learning ensemble techniques to improve the detection rate of credit card anomalies. We utilize four outlier detection algorithms as base learners and XGBoost algorithm as meta learner in the proposed stacked ensemble approach to detect anomaly in credit card transactions. We apply stratified sampling technique and k-fold cross-validation process to address the issues of data imbalance and overfitting. In addition, the discordance rate is calculated to enhance the accuracy of ensemble learning performances. The proposed model is trained and tested using two datasets: CCF (Credit Card Fraud) and CCDP (Credit Card Default Payment). Experimental results demonstrate that our approach outperforms existing approaches, particularly in detecting anomalies from the minority class instances of these datasets.

Networks with hop-by-hop flow control occur in several contexts, from data centers to systems architectures (e.g., wormhole-routing networks on chip). A worst-case end-to-end delay in such networks can be computed using Network Calculus (NC), an algebraic theory where traffic and service guarantees are represented as curves in a Cartesian plane. NC uses transformation operations, e.g., the min-plus convolution, to model how the traffic profile changes with the traversal of network nodes. NC allows one to model flow-controlled systems, hence one can compute the end-to-endservice curvedescribing the minimum service guaranteed to a flow traversing a tandem of flow-controlled nodes. However, while the algebraic expression of such an end-to-end service curve is quite compact, its computation is often intractable from an algorithmic standpoint: data structures tend to grow quickly to unfeasibly large sizes, making operations intractable, even with as few as three hops. In this paper, we propose computational and algebraic techniques to mitigate the above problem. We show that existing techniques (such as reduction tocompact domains) cannot be used in this case, and propose an arsenal of solutions, which include methods to mitigate the data representation space explosion as well as computationally efficient algorithms for the min-plus convolution operation. We show that our solutions allow a significant speedup, enable analysis of previously unfeasible case studies, and - since they do not rely on any approximation - still provide exact results.

Digital Twins (DTs) have been proposed as digital replicas of physical entities (e.g., manufacturing plants), which one can observe and interact with, e.g., to perform what-if analysis. In this paper, we argue that mobile networks need DTs as well, and network simulators appear to be promising candidates to fulfill that role. We discuss the challenges that need be addressed to make this happen, e.g., centralized vs. distributed implementation, gathering input from the physical network, security issues and hosting, and the possibilities offered by network simulation in terms of what-if analysis, defining the concepts of lockstep and branching analysis. We exemplify the above concepts using Simu5G, a popular 5G/B5G simulation library for OMNeT++, as a reference case study.

Due to the trends of centralizing the EIE architecture and new computing-intensive applications, high-performance hardware platforms are currently finding their way into automotive systems. However, the Systems-on-Chip (SoCs) currently available on the market have significant weaknesses when it comes to providing predictable performance for time-critical applications. The main reason for this is that these platforms are optimized for average-case performance. This shortcoming represents one major risk in the development of current and future automotive systems. In this paper we describe how highperformance and predictability could (and should) be reconciled in future HW /SW platforms. We believe that this goal can only be reached via a close collaboration among system suppliers, IP providers, semiconductor companies, and OS/hypervisor vendors. Furthermore, academic input will be needed to solve remaining challenges and to further improve initial solutions.

This paper describes the SAPIENT system, a real-time monitoring and control infrastructure for Air Traffic Management. Within the latter, aircrafts constantly measure the state and quality of their datalinks, and report these measurements to a ground entity, tagging them with a time/space reference. The ground entity, then, builds a map of the monitored portion of the sky, and can feed back information to the aircrafts themselves regarding conditions that they would not be able to measure otherwise. This allows optimal vertical handover decisions to be made, increasing service continuity and improving communication performance. We show that the SAPIENT system can be implemented using existing technologies, without the need for expensive hardware. Moreover, we show via simulation that a small, negligible increase in the communication overhead due to SAPIENT reporting brings about considerable benefits.

Computing Systems are evolving towards more complex, hetero-geneous systems where multiple computing cores and accelera-tors on the same system concur to improve computing resources utilization, resources re-use and the efficiency of data sharing across workloads. Such complex systems require equally complex tools and models to design and engineer them so that their use-case requirements can be satisfied. Adaptive Traffic Profiles (ATP) introduce a fast prototyping technology, which allows one to model the dynamic memory behavior of computer system de-vices when executing their workloads. ATP defines a standard file format and comes with an open source transaction generator engine written in C++. Both ATP files and the engine are porta-ble and pluggable to different host platforms, to allow workloads to be assessed with various models at different levels of abstraction. We present here the ATP technology developed at Arm and published in [5]. We present a case-study involving the usage of ATP, namely the analysis of the worst-case latency at a DRAM controller, which is assessed via two separate toolchains, both using traffic modelling encoded in ATP.

Multi-access Edge Computing (MEC) allows users to run applications on demand near their mobile access points. MEC applications will exploit 5G infrastructure, and they will have to be designed by taking into account the characteristics of 5G mobile networks. This work describes how to use a system-level simulator of 5G networks - namely Simu5G, which evolves the popular 4G network simulator SimuLTE - as a real-time 5G net-work emulator. This allows designers of networked applications - and MEC ones in particular - to use it as a testbed during the deployment. We describe the system setup of Simu5G as an emulator, and its emulation capabilities and scale. Moreover, we present a case study of a MEC testbed using Intel's Open Network Edge Services Software (OpenNESS) toolkit, based on a recent demon-stration in 5GAA (5G Automotive Association).

This paper presents Simu5G, a new OMNeT++-based system-level simulator of 5G networks. Simu5G is built starting from the SimuLTE simulation library, which models 4G (i.e., LTE/LTE-A) networks, and is compatible with the latter, thus allowing the simulation of 4G-5G coexistence and transition scenarios. We discuss the modelling of the protocol layers, network entities and functions, and validate our abstraction of the physical layer using 3GPP-based scenarios. Moreover, we report profiling results related to Simu5G execu-tion, and we describe how it can be employed to evaluate Radio Access Network configurations, as well as end-to-end scenarios involving communication and computation, e.g., with Multi-access Edge Computing ap-plications.

Cellular network nodes should be dynamically switched on/off based on the load requirements of the network, to save power and minimize inter-cell interference. This should be done keeping into account global interference effects, which requires a centralized approach. In this paper, we present an architecture, realized within the Flex5GWare EU project, that manages a large-scale cellular network, switching on and off nodes based on load requirements and context data. We describe the architectural framework and the optimization model that is used to decide the activity state of the nodes. We present simulation results showing that the framework adapts to the minimum power level based on the cell loads.

One-to-one device-to-device (D2D) communications are expected to play a major role in future releases of LTE-A, as well as in future 5G networks . Despite the abundance of works on resource allocation for D2D communications, few works, if any, discuss how D2D should be realized within the LTE-A protocol stack. While it is generally understood that D2D endpoints should be able to communicate both on the direct path or sidelink (SL) and on the relayed path (RP) through the eNB, little has been said on how this can be achieved in practice. In this paper we present a comprehensive proposal for a data-plane architecture for D2D communication: we define how communications should occur on the SL and the RP, and propose a solution for the challenges associated with mode switching between the SL and the RP. In particular, we argue that two different communication modes on the RP are required to allow D2D connections to be kept alive across cell borders in a multicell environment. Our proposal is scalable, since it does not require any signaling, and is guaranteed to not introduce losses. We evaluate our proposal through detailed system-level simulations, also focusing on its interplay with transport-layer protocols.

In an LTE-Advanced network, network-controlled Device-to-Device (D2D) communications can be combined in a multihop fashion to distribute broadcasts over user-defined (and possibly large) areas, with small latencies and occupying few resources. Such a service may be exploited for several purposes, (e.g. Internet of Things, Vehicular communications). Engineering a multihop D2D-based broadcast service requires working at both the application level on the User Equipment (UE) and at the resource-allocation level within the eNodeBs. This paper describes the necessary modifications at both the UE and the eNodeB, what the main issues are, and how to solve them efficiently. We evaluate the performance of the above service using system-level simulations, and demonstrate its advantages over standard broadcasting techniques.

Direct (or device-to-device, D2D) communications are being investigated in the framework of LTE- Advanced. They allow one-to-one communications between two endpoints, under the control of the eNodeB, which allocates resources for the d2d flow, but does not act as a relay for its traffic. The direct link can also be used for file transfer or proximity-based browsing, i.e. applications running on TCP. In this paper, we evaluate the performance of TCP-based traffic transported through the direct link, in several scenarios. We show and explain non-intuitive results, which arise from the interplay of TCP and LTE-A protocol mechanisms, and compare the existing TCP versions in a dynamic environment, where mode switches between the direct and the infrastructure link may induce periodic losses.

Direct (or D2D) communications allow two UEs to communicate without passing through the eNodeB. However, the two UEs may still need to relay their communication through the eNB from time to time, hence should be able to switch from the direct to the re-layed mode seamlessly, without this affecting the QoS. In this paper we show that in conventional systems a mode switching may cause relevant losses, and propose two architectures to miti-gate or solve this problem. Our proposals do not require extra signaling or additional functionalities to be added to the network, hence are scalable and inexpensive. We assess their effectiveness through detailed system-level simulations.

In this paper, we evaluate the schedulability of traffic with arbitrary end-to-end deadline constraints in Wireless Mesh Networks (WMNs). We formulate the problem as a mixed integer linear optimization problem, and show that, depending on the flow aggregation policy used in the network, the problem can be either convex or non-convex. We optimally solve the problem in both cases, and prove that the schedulability does depend on the aggregation policy. This allows us to derive rules of thumb to identify which policy improves the schedulability with a given traffic. Furthermore, we propose a heuristic solution strategy that allows good suboptimal solutions to the scheduling problem to be computed in relatively small times, comparable to those required for online admission control in relatively large WMNs.

Computing network paths under worst-case delay constraints has been the subject of abundant literature in the past two decades. Assuming weighted fair queueing scheduling at the nodes, this translates to computing paths and reserving rates at each link. The problem is 𝒩𝒫-hard in general, even for a single path; hence polynomial-time heuristics have been proposed in the past that either assume equal rates at each node, or compute the path heuristically and then allocate the rates optimally on the given path. In this paper we show that the above heuristics, albeit finding optimal solutions quite often, can lead to failing of paths at very low loads, and that this could be avoided by solving the problem, i.e. path computation and rate allocation, jointly at optimality. This is possible by modeling the problem as a mixed-integer second-order cone program and solving it optimally in split-second times for relatively large networks on commodity hardware; this approach can also be easily turned into a heuristic one, trading a negligible increase in blocking probability for one order of magnitude of computation time. Extensive simulations show that these methods are feasible in today's Internet service provider networks and they significantly outperform the existing schemes in terms of blocking probability.

In this paper, we compute the actual worst-case end-to-end delay for a flow in a feed-forward network of first-in-first-out (FIFO)-multiplexing service curve nodes, where flows are shaped by piecewise-affine concave arrival curves, and service curves are piecewise affine and convex. We show that the worst-case delay problem can be formulated as a mixed integer linear programming problem, whose size grows exponentially with the number of nodes involved. Furthermore, we present approximate solution schemes to find upper and lower delay bounds on the worst-case delay. Both only require to solve just one linear programming problem and yield bounds that are generally more accurate than those found in the previous work, which are computed under more restrictive assumptions.

In LTE and its evolutions, energy efficiency is a critical aspect, also in view of the dramatic traffic growth foreseen for the next years. Cell Discontinuous Transmission (DTX) techniques can be important tools to achieve the needed efficiency in the networks, and one possibility is to implement the DTX by switching off the eNB at some subframes (MBSFN subframes) and not in others (where reference signals are also transmitted). Switching schedules in LTE are made for larger periods (e.g., 40/80ms or even more). We present an algorithm that i) estimates how many resources will be needed in a period, and ii) decides how many resource blocks to activate in each subframe so as to maximize the power efficiency. We show that the power saving is significant, close to the theoretical minimum at low loads. This comes with no reduction in cell throughput and without impacting the QoS (a tolerable extra delay is added only at low loads).

We introduce OptiMOS, a method for selecting deadlines for the scheduling of downlink voice flows in a centralized-scheduling point-to-multipoint network. These include cellular networks such as LTE, UMTS HSDPA, or WiMAX, and WLANs with centralized control. OptiMOS selects the deadline of each packet so as to maximize the expected Mean Opinion Score (MOS) for the talkspurt it belongs to. It requires that absolute time is both included in RTP timestamps at the source and checked for scheduling decisions, in a cross-layer approach. OptiMOS can be employed in conjunction with any deadline-based scheduling algorithm, and does not require a specific playout buffering scheme at the receiver. Simulations of an LTE system show a performance improvement in terms of MOS for the VoIP flows, good tolerance to relatively large synchronization errors and qualitatively similar results with different playout buffers.

In this paper we investigate link scheduling for Wireless Mesh Networks (WMNs) carrying real-time (i.e., delay-constrained) traffic. We show that the problem of computing a conflict-free link schedule with end-to-end delay constraints can be formulated as a mixed-integer non linear problem that can be optimally solved in reasonable time (i.e., minutes) for relatively large WMNs (up to 20-30 nodes). We use the above result to explore the schedulability region of a WMN with a given routing and input traffic, assessing whether and when aggregating flows which traverse the same path makes a given input flow set schedulable. Furthermore, we devise a heuristic solution strategy, which computes good suboptimal solutions within up to few seconds, thus being amenable for online admission control.

Interdomain Traffic Engineering (TE) across domains employing Path Computation Elements should allow a source domain to select a good AS path, i.e. one likely to allow the actual setup of an interdomain tunnel. This is impossible if the AS path is computed online during path setup (which happens, e.

Link scheduling is used in wireless mesh networks (WMNs) to guarantee interference-free transmission on the shared wireless medium in a time division multiple access approach. Several papers in the literature address the problem of link scheduling guaranteeing a minimum throughput to the flows traversing the WMN. However, none of the existing works address the problem of computing a schedule that guarantees that prespecified end-to-end delay constraints are met. In this paper, we make a first step forward in this direction by defining a link scheduling algorithm that works in sinktree WMNs, i.e. those whose traffic is routed towards a common sink (i.e., the Internet gateway). Our iterative algorithm exploits a delay-based admission control procedure, devised through network calculus, which tests the feasibility of a schedule from the point of view of delay guarantees. Preliminary analyses reported in this paper show that the algorithm finds feasible solutions in few iterations.

In this paper we propose a scheduling algorithm for supporting quality of service (QoS) in an IEEE 802.11e network using the HCF controlled channel access (HCCA) function. The algorithm consists of an offline procedure that generates a service schedule over a base period, and an online procedure that applies the latter to actually schedule transmission opportunities to HCCA flows

Multiservice networks host heterogeneous applications, requiring different qualities of service (QoS), the coexistence of which can be efficiently accounted by employing scheduling algorithms which are capable of providing different QoS simultaneously. In a previous work, we defined a reference dual-class (DC) paradigm, according to which rate-guaranteed flows are restrained from using more than their minimum guaranteed rate in the presence of backlogged best-effort flows and the latter share all the remaining capacity according to predetermined weights. The timed token service discipline (TTSD), which applies at the output link of a switch the same rules used to control medium access by the timed token protocol, was also introduced and analyzed therein. It was proven that TTSD shares most of the capacity which is not strictly needed by the rate-guaranteed flows among the best-effort ones, thus achieving one of the goals of the DC paradigm. However, in TTSD, best-effort flows can only share the available capacity equally. We take into account the issue of differentiating the capacity sharing among the best-effort flows: We define a generalized TTSD (GTTSD) in which the latter actually share capacity according to predefined weights in a weighted fair queuing service discipline. Formal analysis and simulation results show that GTTSD closely approximates the DC paradigm.

Deficit Round-Robin (DRR) is a scheduling algorithm devised for providing fair queueing in the presence of variable length packets. The main attractive feature of DRR is its simplicity of implementation: in fact, it can exhibit O(1) complexity, provided that specific allocation constraints are met. However, according to the original DRR implementation, meeting such constraints often implies tolerating high latency and poor fairness. In this paper, we first derive new and exact bounds for DRR latency and fairness. On the basis of these results, we then propose a novel implementation technique, called Active List Queue Method (Aliquem), which allows a remarkable gain in latency and fairness to be achieved, while still preserving O(1) complexity. We show that DRR achieves better performance metrics than those of other round-robin algorithms such as Pre-Order Deficit Round-Robin and Smoothed Round-Robin. We also show by simulation that the proposed implementation allows the average delay and the jitter to be reduced.

Interconnection networks of parallel systems are used for servicing traffic generated by different applications, often belonging to different users. When multiple users contend for channel bandwidth, fairness in bandwidth sharing becomes a key requirement. In fact, enforcing a fair sharing of channel bandwidth improves flow isolation, thus preventing misbehaving flows from affecting the performance of other flows. We present a novel packet scheduling algorithm, called eligibility-based round robin (EBRR), devised to provide fair queueing in interconnection networks. In fact, EBRR meets the constraints imposed by wormhole switching, which is the most popular switching technique in interconnection networks of parallel systems. It can also be applied to packet switching wide area networks (WANs), such as IP and ATM. We show that EBRR has O(1) complexity and better delay and fairness properties than existing algorithms of comparable complexity. Here, we also investigate the means for assessing the fairness of a scheduler: we show that using the relative fairness bound as a fairness measure may lead to erroneous results. We then propose an alternative measure, called the generalized relative fairness bound, that allows fairness to be assessed more precisely.

Multiservice networks face the challenge of managing several traffic classes simultaneously. Service disciplines devised for multiservice networks therefore need to be flexible, i.e., able to provide different types of service, in order to accommodate different traffic classes efficiently. In this paper, we present and analyze an innovative Timed Token Service Discipline (TTSD, which has been given Patent Application No. TO2000A001000) which is able to schedule both rate-guaranteed and best-effort traffic simultaneously at a feasible computational complexity. The service paradigm approximated by TTSD is unveiled by means of a fluid-flow analysis and proven to be different from the well-known Generalized Processor Sharing. We then describe the TTSD properties, deriving minimum rate guarantees and delay bounds for leaky-bucket constrained traffic, both in a single node and in a multinode environment. Furthermore, algorithms for selecting the TTSD parameters in order to achieve a prespecified rate and delay guarantees are proposed.