arXiv subjects

Dario Pompili

Publications and source records attributed to Dario Pompili.

44 records · Page 3Linked to original sources

arXiv2017

Cooperative Hierarchical Caching in 5G Cloud Radio Access Networks (C-RANs)

Over the last few years, Cloud Radio Access Network (C-RAN) has arisen as a transformative architecture for 5G cellular networks that brings the flexibility and agility of cloud computing to wireless communications. At the same time, content caching in wireless networks has become an essential solution to lower the content-access latency and backhaul traffic loading, which translate into user Quality of Experience (QoE) improvement and network cost reduction. In this article, a novel Cooperative Hierarchical Caching (CHC) framework in C-RAN is introduced where contents are jointly cached at the BaseBand Unit (BBU) and at the Radio Remote Heads (RRHs). Unlike in traditional approaches, the cache at the BBU, cloud cache, presents a new layer in the cache hierarchy, bridging the latency/capacity gap between the traditional edge-based and core-based caching schemes. Trace-driven simulations reveal that CHC yields up to 80% improvement in cache hit ratio, 21% decrease in average content-access latency, and 20% reduction in backhaul traffic load compared to the edge-only caching scheme with the same total cache capacity. Before closing the article, several challenges and promising opportunities for deploying content caching in C-RAN are highlighted towards a content-centric mobile wireless network.

cs.IT↗

arXiv2017

Cloud-based Deep Learning of Big EEG Data for Epileptic Seizure Prediction

Developing a Brain-Computer Interface~(BCI) for seizure prediction can help epileptic patients have a better quality of life. However, there are many difficulties and challenges in developing such a system as a real-life support for patients. Because of the nonstationary nature of EEG signals, normal and seizure patterns vary across different patients. Thus, finding a group of manually extracted features for the prediction task is not practical. Moreover, when using implanted electrodes for brain recording massive amounts of data are produced. This big data calls for the need for safe storage and high computational resources for real-time processing. To address these challenges, a cloud-based BCI system for the analysis of this big EEG data is presented. First, a dimensionality-reduction technique is developed to increase classification accuracy as well as to decrease the communication bandwidth and computation time. Second, following a deep-learning approach, a stacked autoencoder is trained in two steps for unsupervised feature extraction and classification. Third, a cloud-computing solution is proposed for real-time analysis of big EEG data. The results on a benchmark clinical dataset illustrate the superiority of the proposed patient-specific BCI as an alternative method and its expected usefulness in real-life support of epilepsy patients.

cs.LG↗

arXiv2017

Energy-efficient Analog Sensing for Large-scale, High-density Persistent Wireless Monitoring

The research challenge of current Wireless Sensor Networks~(WSNs) is to design energy-efficient, low-cost, high-accuracy, self-healing, and scalable systems for applications such as environmental monitoring. Traditional WSNs consist of low density, power-hungry digital motes that are expensive and cannot remain functional for long periods on a single charge. In order to address these challenges, a \textit{dumb-sensing and smart-processing} architecture that splits sensing and computation capabilities among tiers is proposed. Tier-1 consists of dumb sensors that only sense and transmit, while the nodes in Tier-2 do all the smart processing on Tier-1 sensor data. A low-power and low-cost solution for Tier-1 sensors has been proposed using Analog Joint Source Channel Coding~(AJSCC). An analog circuit that realizes the rectangular type of AJSCC has been proposed and realized on a Printed Circuit Board for feasibility analysis. A prototype consisting of three Tier-1 sensors (sensing temperature and humidity) communicating to a Tier-2 Cluster Head has been demonstrated to verify the proposed approach. Results show that our framework is indeed feasible to support large scale high density and persistent WSN deployment.

cs.NI↗

arXiv2017

Robust Orchestration of Concurrent Application Workflows in Mobile Device Clouds

A hybrid mobile/fixed device cloud that harnesses sensing, computing, communication, and storage capabilities of mobile and fixed devices in the field as well as those of computing and storage servers in remote datacenters is envisioned. Mobile device clouds can be harnessed to enable innovative pervasive applications that rely on real-time, in-situ processing of sensor data collected in the field. To support concurrent mobile applications on the device cloud, a robust and secure distributed computing framework, called Maestro, is proposed. The key components of Maestro are (i) a task scheduling mechanism that employs controlled task replication in addition to task reallocation for robustness and (ii) Dedup for task deduplication among concurrent pervasive workflows. An architecture-based solution that relies on task categorization and authorized access to the categories of tasks is proposed for different levels of protection. Experimental evaluation through prototype testbed of Android- and Linux-based mobile devices as well as simulations is performed to demonstrate Maestro's capabilities.

cs.DC↗

arXiv2016

Collaborative Multi-bitrate Video Caching and Processing in Mobile-Edge Computing Networks

Recently, Mobile-Edge Computing (MEC) has arisen as an emerging paradigm that extends cloud-computing capabilities to the edge of the Radio Access Network (RAN) by deploying MEC servers right at the Base Stations (BSs). In this paper, we envision a collaborative joint caching and processing strategy for on-demand video streaming in MEC networks. Our design aims at enhancing the widely used Adaptive BitRate (ABR) streaming technology, where multiple bitrate versions of a video can be delivered so as to adapt to the heterogeneity of user capabilities and the varying of network connection bandwidth. The proposed strategy faces two main challenges: (i) not only the videos but their appropriate bitrate versions have to be effectively selected to store in the caches, and (ii) the transcoding relationships among different versions need to be taken into account to effectively utilize the processing capacity at the MEC servers. To this end, we formulate the collaborative joint caching and processing problem as an Integer Linear Program (ILP) that minimizes the backhaul network cost, subject to the cache storage and processing capacity constraints. Due to the NP-completeness of the problem and the impractical overheads of the existing offline approaches, we propose a novel online algorithm that makes cache placement and video scheduling decisions upon the arrival of each new request. Extensive simulations results demonstrate the significant performance improvement of the proposed strategy over traditional approaches in terms of cache hit ratio increase, backhaul traffic and initial access delay reduction.

cs.NI↗

arXiv2016

Automatic and Manual Segmentation of Hippocampus in Epileptic Patients MRI

The hippocampus is a seminal structure in the most common surgically-treated form of epilepsy. Accurate segmentation of the hippocampus aids in establishing asymmetry regarding size and signal characteristics in order to disclose the likely site of epileptogenicity. With sufficient refinement, it may ultimately aid in the avoidance of invasive monitoring with its expense and risk for the patient. To this end, a reliable and consistent method for segmentation of the hippocampus from magnetic resonance imaging (MRI) is needed. In this work, we present a systematic and statistical analysis approach for evaluation of automated segmentation methods in order to establish one that reliably approximates the results achieved by manual tracing of the hippocampus.

cs.CV↗

arXiv2016

Octopus: A Cooperative Hierarchical Caching Strategy for Cloud Radio Access Networks

Recently, implementing Radio Access Network (RAN) functionalities on cloud-based computing platform has become an emerging solution that leverages the many advantages of cloud infrastructure, such as shared computing resources and storage capacity, while lowering the operational cost. In this paper, we propose a novel caching framework aimed at fully exploiting the potential of such Cloud-based RAN (C-RAN) systems through cooperative hierarchical caching which minimizes the network costs of content delivery and improves users' Quality of Experience (QoE). In particular, we consider the cloud-cache in the cloud processing unit (CPU) as a new layer in the RAN cache hierarchy, bridging the capacity-performance gap between the traditional edge-based and core-based caching schemes. A delay cost model is introduced to characterize and formulate the cache placement optimization problem, which is shown to be NP-complete. As such, a low complexity, heuristic cache management strategy is proposed, constituting of a proactive cache distribution algorithm and a reactive cache replacement algorithm. Extensive numerical simulations are carried out using both real-world YouTube video requests and synthetic content requests. It is demonstrated that our proposed Octopus caching strategy significantly outperforms the traditional caching strategies in terms of cache hit ratio, average content access delay and backhaul traffic load.

cs.NI↗

arXiv2015

Dynamic Radio Cooperation for Downlink Cloud-RANs with Computing Resource Sharing

A novel dynamic radio-cooperation strategy is proposed for Cloud Radio Access Networks (C-RANs) consisting of multiple Remote Radio Heads (RRHs) connected to a central Virtual Base Station (VBS) pool. In particular, the key capabilities of C-RANs in computing-resource sharing and real-time communication among the VBSs are leveraged to design a joint dynamic radio clustering and cooperative beamforming scheme that maximizes the downlink weighted sum-rate system utility (WSRSU). Due to the combinatorial nature of the radio clustering process and the non-convexity of the cooperative beamforming design, the underlying optimization problem is NP-hard, and is extremely difficult to solve for a large network. Our approach aims for a suboptimal solution by transforming the original problem into a Mixed-Integer Second-Order Cone Program (MI-SOCP), which can be solved efficiently using a proposed iterative algorithm. Numerical simulation results show that our low-complexity algorithm provides close-to-optimal performance in terms of WSRSU while significantly outperforming conventional radio clustering and beamforming schemes. Additionally, the results also demonstrate the significant improvement in computing-resource utilization of C-RANs over traditional RANs with distributed computing resources.

cs.IT↗