An increasing number of applications rely on complex inference tasks that are based on machine learning (ML). Currently, there are two options to run such tasks: either they are served directly by the end device (e.g., smartphones, IoT equipment, smart vehicles), or offloaded to a remote cloud. Both options may be unsatisfactory for many applications: local models may have inadequate accuracy, while the cloud may fail to meet delay constraints. In this paper, we present the novel idea of inference delivery networks (IDNs), networks of computing nodes that coordinate to satisfy ML inference requests achieving the best trade-off between latency and accuracy. IDNs bridge the dichotomy between device and cloud execution by integrating inference delivery at the various tiers of the infrastructure continuum (access, edge, regional data center, cloud). We propose a distributed dynamic policy for ML model allocation in an IDN by which each node dynamically updates its local set of inference models based on requests observed during the recent past plus limited information exchange with its neighboring nodes. Our policy offers strong performance guarantees in an adversarial setting and shows improvements over greedy heuristics with similar complexity in realistic scenarios.

Modern information and communication systems have become increasingly challenging to manage. The ubiquitous system logs contain plentiful information and are thus widely exploited as an alternative source for system management. As log files usually encompass large amounts of raw data, manually analyzing them is laborious and error-prone. Consequently, many research endeavors have been devoted to automatic log analysis. However, these works typically expect structured input and struggle with the heterogeneous nature of raw system logs. Log parsing closes this gap by converting the unstructured system logs to structured records. Many parsers were proposed during the last decades to accommodate various log analysis applications. However, due to the ample solution space and lack of systematic evaluation, it is not easy for practitioners to find ready-made solutions that fit their needs. This paper aims to provide a comprehensive survey on log parsing. We begin with an exhaustive taxonomy of existing log parsers. Then we empirically analyze the critical performance and operational features for 17 open-source solutions both quantitatively and qualitatively, and whenever applicable discuss the merits of alternative approaches. We also elaborate on future challenges and discuss the relevant research directions. We envision this survey as a helpful resource for system administrators and domain experts to choose the most desirable open-source solution or implement new ones based on application-specific requirements.

A non-traditional type of financial asset, cryptocurrencies based on public blockchains are still little understood in their real-world behavior. Exogenous events such as cyber-attacks and their media coverage can strongly affect their supply and demand, adoption and usage, efficiency, and infrastructural development, thus influencing their price stability and market valuation. Given the great technical complexity of blockchains, we believe that a pure economic analysis of risks associated with cryptocurrencies is simply not sufficient to convey the relationships between attacks and the disruptive effects that these can bring on the operation of cryptocurrencies. On these grounds, we survey the most common types of attacks for Proof-of-Work (PoW) cryptocurrencies and evaluate their impact on the returns of a number of real-world cryptocurrencies for which market data are available. Due to data availability, our event study analysis focuses on instances of 51% attacks, hard forks, and wallet attacks. The main goal of our analysis is to understand the relationship between technical events (cyber-attacks and coordinated user/miner behavior) and the economic impacts surrounding them. We aim to develop a deeper understanding of these systems that are objects of great research interest in separate disciplines, supporting policymakers in their regulatory decisions concerning crypto-assets and associated cyber-related financial risks.

VLC based indoor positioning system is a natural choice thanks to LED technology and future trend. We can reuse existing LED lighting installations to design an appropriate scheme to provide indoor localization. By receiving basic information such as the light intensity and knowing the position of the transmitters for example through Visible Light Communication (VLC), one can estimate its position with the potential of high accuracy. In this paper, we use a multi-color sensor and build an indoor localization system based on the light fingerprint, which is suitable for smart cities and homes, health-care centers, hospitals or similar. There are, however, several practical problems that make it a technology that is not yet sufficiently developed to be used in our everyday life. In this paper, a new perspective is given by looking into indoor localization using multi-color fingerprinting and different machine learning methods. Experiment results have shown its effectiveness and potentials with mean localization errors of around ten centimeters.

An automated offline design process for optimized cross-layer schedulers can produce augmented scheduling algorithms tailored to a target deployment scenario. We discuss the application of ODDS, a reinforcement learning technique we introduced, for augmenting LTE MAC scheduler algorithms of practical significance. ODDS observes the correlation between the value of a utility function and the cross-layer state parameters seen by an instrumented baseline scheduler, determining the best actions via an offline Monte Carlo exploration of the problem space. The result of the ODDS process is a compact definition of a scheduling policy that has been optimized for a target scenario and utility function. In this paper we instrument a production scheduler definition to evaluate the potential of augmented schedulers in practical use, and experiment with awareness to application traffic properties by using a multi-class utility function, yielding scheduling policies that behave differently depending on the features of individual traffic classes.

The upcoming 5G architecture is expected to heavily rely on network functions implemented by software deployed on commodity hardware architectures. Multiple standardization efforts are underway to specify interfaces between virtualized and real infrastructure, and procedures for interoperability among functions. However, the practical feasibility of function implementation in such abstract and disembodied conditions is scarcely covered in the latest literature. In this paper, we argue for a Network Function Virtualization (NFV) framework that provides 5G network functions built around a modular software router model, rather than following the traditional VM-container approaches. We illustrate its advantages in enabling support for efficient processing on heterogeneous hardware and in ensuring consistency of flow/session semantics across distributed 5G data planes. Finally, we report on the state of Programmable Data Plane, our architecture to implement 5G network functions as modular pipelines orchestrated across multiple devices.

The study of human activity patterns traditionally relies on the continuous tracking of user location. We approach the problem of activity pattern discovery from a new perspective which is rapidly gaining attention. Instead of actively sampling increasing volumes of sensor data, we explore the participatory sensing potential of multiple mobile social networks, on which users often disclose information about their location and the venues they visit. In this paper, we present automated techniques for filtering, aggregating, and processing combined social networking traces with the goal of extracting descriptions of regularly-occurring user activities, which we refer to as “user routines”. We report our findings based on two localized data sets about a single pool of users: the former contains public geotagged Twitter messages, the latter Foursquare check-ins that provide us with meaningful venue information about the locations we observe. We analyze and combine the two datasets to highlight their properties and show how the emergent features can enhance our understanding of users' daily schedule. Finally, we evaluate and discuss the potential of routine descriptions for predicting future user activity and location.

In this position paper we introduce Virtual Telco, a comprehensive effort to simplify the management of deployed telecommunication services by using a Cloud computing approach. The main objective of Virtual Telco is to replace the costly dedicated hardware implementing several centralized control plane functions and other services with distributed solutions that may be allocated on-demand over a pool of dependable, dynamically contracted computing and networking resources that are easy to manage. Virtual Telco relies on state-of-the-art techniques in the domains of virtualization and distributed systems to meet the challenging criteria of reliability, scalability, and timeliness required by present and future telecommunication standards and services. As a representative example of a Virtual Telco application, we propose the case of the distributed mobility management entity (MME) for next-generation LTE cellular networks.

With the convergence between phone and data networks in LTE and 4G, cellular signaling traffic is increasingly carried over IP. Control plane functions, once performed by dedicated machinery, are evolving into large-scale network applications with strict requirements on delay, availability, and processing throughput as mandated by 3GPP standards. In this paper we present DMME, a distributed architecture that implements mobility management for next-generation cellular systems. DMME is a scalable and cost-effective drop-in replacement for the LTE mobility management entity (MME). We evaluate the DMME scheme via analysis and simulation under several deployment scenarios, using mobility patterns drawn from synthetic models and from traces collected in a production network. We also report preliminary performance figures by our DMME prototype implementation. Our results confirm that distributed architectures are a viable choice to reliably support high-throughput, latency-sensitive control plane functions.

Large-scale live media streaming is a challenge for traditional server-based approaches. To appropriately support big audiences, broadcasters must be able to allocate huge bandwidth and computational resources. The costs involved with such an infrastructure exclude all but the established content producers from exploiting the Internet as a distribution medium. Publishers of not-yet-popular content, unless they manage to properly predict their maximum audience size, will likely fail to dimension correctly their broadcast infrastructure. Peer-to-peer systems for live streaming allow the users to support content distribution by contributing their unused resources: this increases the scalability of the content distribution while reducing at the same time the economical burden on the streaming provider. This paper presents and evaluates PULSE, an unstructured mesh-based peer-to-peer system designed to support live streaming to large audiences under the arbitrary resource availability as is typically the case for the Internet. PULSE is a highly dynamic system: it constantly optimizes its mesh of data connections using a feedback-driven peer selection strategy that is based on pairwise incentives. We evaluate the behavior of PULSE under realistic scenarios via simulation and emulation, and present the advantages of our approach, namely a best-effort response to system-wide resource scarcity, high resilience to node churn, and good hop-count properties of the average data distribution paths.

With the widespread availability of inexpensive broadband Internet connections for home-users, a large number of bandwidth-intensive applications previously not feasible have now become practical. This is the case for multimedia live streaming, for which end-user's dial-up/ISDN modem connections once were the bottleneck. The bottleneck is now mostly found on the server side: the bandwidth required for serving many clients at once is large and thus very costly to the broadcasting entity. Peer-to-peer systems for on-demand and live streaming have proved to be an encouraging solution, since they can shift the burden of content distribution from the server to the users of the network. In this work we introduce PULSE, a P2P system for live streaming whose main goals are flexibility, scalability, and robustness. We present the fundamental concepts that stand behind the design of PULSE along with its intended global behavior, and describe in detail the main algorithms running on its nodes.