The rising popularity of Vehicle-to-Network (V2N) applications is driven by the Ultra-Reliable Low-Latency Communications (URLLC) service offered by 5G. Distributed resources can help manage heavy traffic from these applications, but complicate traffic routing under URLLCfs strict delay requirements. In this paper, we introduce the V2N Computation Offloading and CPU Activation (V2N-COCA) problem, aiming at the monetary/energetic cost minimization via computation offloading and edge/cloud CPU activation decisions, under stringent latency constraints. Some challenges are the proven nonmonotonicity of the objective function and the no-existence of closed-formulas for the sojourn time of tasks. We present a provably tight approximation for the latter, and we design BiQui, a provably asymptotically optimal and computationally efficient algorithm for the V2N-COCA problem. We then study dynamic scenarios, introducing the Swap-Prevention problem, to account for changes in the traffic load and minimize the switching on/off of CPUs without incurring into overcosts.We prove the problemfs structural properties and exploit them to design Min-Swap, a provably correct and computationally effective algorithm for the Swap-Prevention Problem. We assess both BiQui and Min-Swap over real-world vehicular traffic traces, performing a sensitivity analysis and a stress-test. Results show that (i) BiQui is nearoptimal and significantly outperforms existing solutions; and (ii) Min-Swap reduces by a ≥90% the CPU swapping incurring into just ≤0.14% extra cost.

The rising popularity of Vehicle-to-Network (V2N) applications is driven by the Ultra-Reliable Low-Latency Communications (URLLC) service offered by 5G. The availability of distributed resources could be leveraged to handle the enormous traffic arising from these applications, but introduces complexity in deciding where to steer traffic under the stringent delay requirements of URLLC. In this paper, we introduce the V2N Computation Offloading and CPU Activation (V2N-COCA) problem, which aims at finding the computation offloading and the edge/cloud CPU activation decisions that minimize the operational costs, both monetary and energetic, under stringent latency constraints. Some challenges are the proven non-monotonicity of the objective function w.r.t. offloading decisions, and the no-existence of closed-formulas for the sojourn time of tasks. We present a provably tight approximation for the latter, and we design BiQui, a provably asymptotically optimal and with linear computational complexity w.r.t. computing resources algorithm for the V2N-COCA problem. We assess BiQui over real-world vehicular traffic traces, performing a sensitivity analysis and a stress-test. Results show that BiQui significantly outperforms state-of-the-art solutions, achieving optimal performance (found through exhaustive searches) in most of the scenarios.

Next-generation mobile networks will rely on their autonomous operation. Virtual Network Functions empowered by Artificial Intelligence (AI) and Machine Learning (ML) can adapt to varying environments that encompass both network conditions and the cloud platform executing them. In this view, it becomes paramount tounderstand whyAI/ML algorithms made a decision, to be able to reason upon those decisions and, eventually, take further decisions related toe.g., network orchestration. In this paper, we present ATHENA, an ML-based radio resource scheduler for virtualized Radio Access Network (RAN) system. Our real-software implementation shows that the proposed ML-based approach can outperform the baseline solution. We discuss how additional re-orchestration actions can be taken by analyzing our scheduling decisions and learning from the past.

This work contributes towards optimizing edge analytics in Mobile Edge Computing (MEC) systems. We consider requests for computing tasks that are generated from users and can be satisfied either locally at their devices, or they can be offloaded to an edge server in their proximity for remote execution. We study a multi-user MEC system with limited energy autonomy for the mobile devices and with limitations on the computing capability of both mobile devices and at an edge server, where users can offload part of their computation load. We define a utility over “resource residuals”, that capture the difference between the resources assigned through our decisions, and those needed in practice, and we aim at the minimization of regret, i.e., of the difference between the utility obtained by an optimal offline benchmark that knows the system evolution in hindsight, and our online decision policy. We design an algorithm that jointly learns policies for offloading computations and scheduling them for execution at the shared MEC server. We prove that our algorithm is asymptotically optimal, i.e., it has no regret over the optimal static offline benchmark, and that its performance is independent of the number of devices in the system. From our numerical evaluation we conclude that our algorithm adapts to unpredictable demand changes, it learns to identify resource-limited devices, and it learns to share the server’s resources.

We learn optimal user association policies for traffic from different locations to Access Points(APs), in the presence of unknown dynamic traffic demand. We aim at minimizing a broad family of α-fair cost functions that express various objectives in load assignment in the wireless downlink, such as total load or total delay minimization. Finding an optimal user association policy in dynamic environments is challenging because traffic demand fluctuations over time are non-stationary and difficult to characterize statistically, which obstructs the computation of cost-efficient associations. Assuming arbitrary traffic patterns over time, we formulate the problem of online learning of optimal user association policies using the Online Convex Optimization (OCO) framework. We introduce a periodic benchmark for OCO problems that generalizes state-of-the-art benchmarks. We exploit inherent properties of the online user association problem and propose PerOnE, a simple online learning scheme that dynamically adapts the association policy to arbitrary traffic demand variations. We compare PerOnE against our periodic benchmark and prove that it enjoys the no-regret property, with additional sublinear dependence of the network size. To the best of our knowledge, this is the first work that introduces a periodic benchmark for OCO problems and a no-regret algorithm for the online user association problem. Our theoretical findings are validated through results on a real-trace dataset.

Caching decisions typically seek to cache content that satisfies the maximum possible demand aggregated over all users. Recommendation systems, on the contrary, focus on individual users and recommend to them appealing content in order to elicit further content consumption. In our paper, we explore how these, phenomenally conflicting, objectives can be jointly addressed. First, we formulate an optimization problem for the joint caching and recommendation decisions, aiming to maximize the cache hit ratio under minimal controllable distortion of the inherent user content preferences by the issued recommendations. Then, we prove that the problem is NP-complete and that its objective function lacks those monotonicity and submodularity properties that would guarantee its approximability. Hence, we proceed to introduce a simpler heuristic algorithm that essentially serves as a form of lightweight control over recommendations so that they are both appealing to end-users and friendly to network resources. Finally, we draw on both analysis and simulations with real and synthetic datasets to evaluate the performance of the algorithm. We point out its fundamental properties, provide bounds for the achieved cache hit ratio, and study its sensitivity to its own as well as system-level parameters.

Caching decisions by default seek to maximize some notion of social welfare: the content to be cached is determined so that the maximum possible aggregate demand over all users served by the cache is satisfied. Recommendation systems, on the contrary, are oriented towards user individual preferences: the recommended content should be most appealing to the user so as to elicit further content consumption. In our paper we explore how these, phenomenically conflicting, objectives can be jointly addressed. To this end, we depart radically from current practice with recommender systems, and we approach them as network traffic engineering tools that can actively shape content demand towards optimizing user- and network-centric performance objectives. We formulate the resulting joint theoretical optimization problem of deciding on the cached content and the recommendations to each user so that the cache hit ratio is maximized subject to a maximum tolerable distortion that the recommendation should undergo. We conclude on its complexity, and we propose a practical algorithm for its solution. The algorithm is essentially a form of lightweight control over the user recommendations so that the recommended content is both appealing to the end user and more friendly to the caching system and the network resources.