In the traditional two-rack distributed storage system (DSS) model, due to the assumption that the storage capacity of each node is the same, the minimum bandwidth regenerating (MBR) point becomes infeasible. In this paper, we design a new non-homogeneous two-rack model by proposing a generalization of the threshold function used to compute the tradeoff curve. We prove that by having the nodes in the rack with higher regenerating bandwidth stores more information, all the points on the tradeoff curve, including the MBR point, become feasible. Finally, we show how the non-homogeneous two-rack model outperforms the traditional model in the tradeoff curve between the storage per node and the repair bandwidth.
SHIELD is an EU-funded project, targeting at the design and development of a novel cybersecurity framework, which offers security-as-a-Service in an evolved telco environment. The SHIELD framework leverages NFV (Network Functions Virtualization) and SDN (Software-Defined Networking) for virtualization and dynamic placement of virtualised security appliances in the network (virtual Network Security Functions - vNSFs), Big Data analytics for real-time incident detection and mitigation, as well as attestation techniques for securing both the infrastructure and the services. This papers discusses key use cases and requirements for the SHIELD framework and presents a high-level architectural approach.
17 May 2021 in IEEE Transactions on Industrial Informatics
This article presents a simple solution to estimate the location of products in a retail store, using an autonomous ground robot with a radio frequency identification payload. The model used and explained in this article is designed to be as simple and versatile as possible, while achieving accurate location estimations when compared with other proposed models in the state of the art. In addition, the solution developed meets the business requirements of the retail industry, such as locating at stock keeping unit level, as opposed to item level, or expressing the location in terms of store fixtures (e.g., shelf and rack) as opposed to $(x,y)$ coordinates. The research results are obtained from experiments of the model in different environments, achieving accurate location estimations in a controlled laboratory environment. Moreover, for the first time, the model has been tested in a large retail store, where the results obtained met the requirements of the store owners.
This demo showcases some of the capabilities foreseen for the security infrastructure designed by the H2020 SHIELD project. SHIELD exploits NFV for adaptive monitoring of an IT infrastructure and for feeding the data to an analytics engine to detect attacks in real time. An intelligent reaction system is then activated to reconfigure the SDN/NFV infrastructure so that the attacks are thwarted. The SDN/NFV infrastructure itself is protected from attacks thanks to trusted computing techniques, that permit to quickly identify misbehaving nodes. The proposed demo will present detection and reaction to a DDoS attack (by on-the-fly deployment of new virtual network security functions and/or change of network paths), as well as detection of software attacks against virtual network functions (executed in Docker containers) and unauthorized modification of the SDN switching tables and NFV configurations.
23 November 2021 in IEEE Transactions on Industrial Electronics
Radio frequency identification (RFID) technology is being widely adopted by retailers due to its accuracy, versatility, and reduction of operational costs. Most commonly, RFID in retail is used for taking frequent and accurate inventories of items in the stores. Usually, RFID inventories use handheld RFID devices, which makes the task tedious, costly, and prone to human errors. More reliable, fully automatic alternatives exist, such as smart shelves, overhead RFID antennas, and RFID-equipped robots. Among them, robots seem to be the preferred choice by retailers with large stores. However, retailers need an objective way to compare the different options for inventory solutions and to calculate the return on investment of each of them before they make an investment decision. In this article, we present a metric for assessing, comparing, and predicting the performance of autonomous RFID-based robots in retail stores. The metric is based on a theoretical model of both the store and the robot, and predicts the performance of a given robot when inventorying a specific store. The metric also allows to compare the performance of different RFID robots in different stores. The metric has been developed using experimental data and has been validated in a real store.
Nowadays, it is possible to optimize the storage size of large files in distributed environments, maintaining the same availability level in the system. Although replication (backups) is the most used option, it is possible to make use of erasure coding in order to significantly compress the storage size [1]. However, using erasure coding, more information needs to be transmitted than when using replication, in order to replace a node which has failed. This is called the code repair problem. The amount of transmitted information can be an important issue when the file size is very large. If network coding is used in conjunction with erasure codes, the transmitted information can be reduced by compressing the sent data in the regeneration phase. This solution minimizes the code repair problem and consists of the use of regenerating codes introduced by Dimakis et al. [2]. Nevertheless, using network coding, computational resources for any node in the system are required, since linear operations must be carried out and systems of equations must be solved in the nodes. This requirement is too strong for many simple storage devices [3]. In this paper a new family of Minimum Storage Regenerating codes is proposed. These codes not only compress the storage size using erasure coding and achieve optimality for the amount of transmitted information when d = k + 1, but also demand few computational requirements. Therefore, they could be used for simple hard discs without computational resources.