A5000838509- Cloud Computing and Resource Management
- Software-Defined Networks and 5G
- Caching and Content Delivery
- IoT and Edge/Fog Computing
- Machine Learning and Data Classification
- Network Traffic and Congestion Control
- Advanced Data Storage Technologies
- Data Stream Mining Techniques
- Parallel Computing and Optimization Techniques
- Privacy-Preserving Technologies in Data
- Adversarial Robustness in Machine Learning
- Advanced Neural Network Applications
- Peer-to-Peer Network Technologies
- Generative Adversarial Networks and Image Synthesis
- Interconnection Networks and Systems
- Anomaly Detection Techniques and Applications
- Distributed and Parallel Computing Systems
- Stochastic Gradient Optimization Techniques
- Distributed systems and fault tolerance
- Advanced Optical Network Technologies
- Software System Performance and Reliability
- Image and Video Quality Assessment
- Blockchain Technology Applications and Security
- Mobile Crowdsensing and Crowdsourcing
- Imbalanced Data Classification Techniques
University of Turin
2022-2025
ABB (Switzerland)
2017-2022
ABB (India)
2021
IBM (United States)
2011-2019
IBM Research - Zurich
2011-2019
Polytechnic University of Turin
2005-2014
Switch
2012
Kiel University
1979-2009
Institut de Biologie et de Chimie des Protéines
2008
University of Lübeck
1983
In this work, we propose a VAE-LSTM hybrid model as an unsupervised approach for anomaly detection in time series. Our utilizes both VAE module forming robust local features over short windows and LSTM estimating the long term correlation series on top of inferred from module. As result, our algorithm is capable identifying anomalies that span multiple scales. We demonstrate effectiveness five real world problems find method outperforms three other commonly used methods.
The usage of synthetic data is gaining momentum in part due to the unavailability original privacy and legal considerations its utility as an augmentation authentic data. Generative adversarial networks (GANs), a paragon generative models, initially for images subsequently tabular data, has contributed many state-of-the-art synthesizers. As GANs improve, synthesized increasingly resemble real risking leak privacy. Differential (DP) provides theoretical guarantees on loss but degrades...
OpenFlow is an open standard that can be implemented in Ethernet switches, routers and wireless access points (AP). In the framework, packet forwarding (data plane) routing decisions (control run on different devices. switches are charge of forwarding, whereas a controller set up switch table per-flow basis, to enable flow isolation resource slicing. We focus data path analyze implementation Linux based PCs. compare switching, layer-2 switching layer-3 IP performance. Forwarding throughput...
Nowadays, at least two billion people are experiencing a complete lack of wireless cellular network coverage. These users live in rural areas and low-income regions, where operators not keen to invest, mainly due high capital expenditure operational costs, as well the scarcity electricity from grid. We tackle this challenge by proposing 5G explicitly designed serve areas. Our solution investigates possibility mounting remote radio heads on top unmanned aerial vehicles, large cells (LCs)...
In today's commercial data centers, the computation density grows continuously as number of hardware components and workloads in units virtual machines increase. The service availability guaranteed by centers heavily depends on reliability physical servers. this study, we conduct an analysis 10K hosted five over observation period one year. Our objective is to establish a sound understanding differences similarities between failures machines. We first capture their failure patterns, i.e.,...
While data sharing is crucial for knowledge development, privacy concerns and strict regulation (e.g., European General Data Protection Regulation (GDPR)) unfortunately limit its full effectiveness. Synthetic tabular emerges as an alternative to enable while fulfilling regulatory constraints. The state-of-the-art synthesizers draw methodologies from generative Adversarial Networks (GAN) address two main types in the industry, i.e., continuous categorical. In this paper, we develop CTAB-GAN,...
Recent trends in deep learning (DL) imposed hardware accelerators as the most viable solution for several classes of high-performance computing (HPC) applications such image classification, computer vision, and speech recognition. This survey summarizes classifies recent advances designing DL suitable to reach performance requirements HPC applications. In particular, it highlights advanced approaches support accelerations including not only GPU TPU-based but also design-specific FPGA-based...
With the advancement of virtualization technologies and benefit economies scale, industries are seeking scalable IT solutions, such as data centers hosted either in-house or by a third party. Data center availability, often via cloud setting, is ubiquitous. Nonetheless, little known about in-production performance centers, especially interaction workload demands resource availability. This study fills this gap conducting large scale survey servers within time period that spans two years. We...
We analyze workload traces from production data centers and focus on their VM usage patterns of CPU, memory, disk, network bandwidth. Burstiness is a clear characteristic many these time series: there exist peak loads within periodic but also that do not have periodicity. present PRACTISE, neural based framework can efficiently accurately predict future loads, timing. Extensive experimentation using IBM illustrates PRACTISE's superiority when compared to ARIMA baseline models, with average...
Current trends in telecommunication networks foresee the adoption of fifth generation (5G) wireless near future. However, a large number people are living without coverage and connectivity. To face this issue, we consider possibility deploying 5G rural low-income zones. After detailing current state-of-the-art, main challenges that need to be faced. Moreover, define pillars follow order deploy such zones, as well proposal future network architecture.
Hardware virtualization is the prevalent way to share data centers among different tenants. In this paper we present a large scale workload characterization study that aims better understanding of state-of-the-practice, i.e., how in private cloud are used by their customers, physical resources shared tenants using virtualization, and technologies actually employed. Our focuses on all corporate major infrastructure provider geographically dispersed across entire globe reports observed usage...
VoIP has widely been addressed as the technology that will change Telecommunication model opening path for convergence. Still today this revolution is far from being complete, since majority of telephone calls are originated by circuit-oriented networks. In paper first time to best our knowledge, we present a large dataset measurements collected FastWeb backbone, which one worldwide Telecom operator offer and high-speed data access end-user. Traffic characterization focus on several layers,...
Virtualization is the ubiquitous way to provide computation and storage services datacenter end-users. Guaranteeing sufficient data efficient access central all operations, yet little known of effects virtualization on workloads. In this study, we collect analyze field from production datacenters that operate within private cloud paradigm, during a period three years. The our study consist 8,000 physical boxes, hosting over 90,000 VMs, which in turn use 22 PB storage. Storage analyzed...
Peer to streaming (P2P-TV) applications have recently emerged as cheap and efficient solutions provide real time services over the Internet. For sake of simplicity, typical P2P-TV systems are designed optimized following a pure layered approach, thus ignoring effect design choices on underlying transport network. This simple however, may constitute threat for network providers, due congestion that traffic can potentially generate. In this article, we present discuss architecture an...
As energy costs become increasingly greater contributors to a cloud provider's overall costs, it is important for the recoup these from its tenants profitability via appropriate pricing design. The poor predictability of real-world tenants' demand and responses (DRs) make such design challenging problem. We formulate leader-follower game-based framework with goal maximizing cloud's profit. key distinguishing aspect our approach emphasis on modeling both as working low in their inputs....
Classification algorithms have been widely adopted to detect anomalies for various systems, e.g., IoT and cloud, under the common assumption that data source is clean, i.e., features labels are correctly set. However, collected from field can be unreliable due careless annotations or malicious transformation incorrect anomaly detection. In this paper, we present a two-layer learning framework robust detection (RAD) in presence of labels. The first layer quality model filters suspicious data,...
Computational sprinting speeds up query execution by increasing power usage for short bursts. Sprinting policy decides when and how long to sprint. Poor policies inflate response time significantly. We propose a model-driven approach that chooses between based on their expected time. However, alters executions at runtime, creating complex dependency queuing processing Our performance modeling employs offline profiling, machine learning, first-principles simulation. Collectively, these...
Generative Adversarial Networks (GANs) are typically trained to synthesize data, from images and more recently tabular under the assumption of directly accessible training data. Recently, federated learning (FL) is an emerging paradigm that features decentralized on client's local data with a privacy-preserving capability. And, while GANs FL systems has just been demonstrated, it unknown if for can be learned sources. Moreover, remains unclear which distributed architecture suits them best....
Deep neural networks (DNNs) are becoming the core components of many applications running on edge devices, especially for real time image-based analysis. Increasingly, multi-faced knowledge is extracted via executing multiple DNNs inference models, e.g., identifying objects, faces, and genders from images. The response times multi-DNN highly affect users' quality experience safety as well. Different exhibit diversified resource requirements execution patterns across layers networks, which...
We explore the efficacy of dynamic effective capacity modulation (i.e., using virtualization techniques to offer lower resource than that advertised by cloud provider) as an explicit control knob for a provider's profit maximization complementing more well-studied approach pricing. Our focus is on emerging ecosystems wherein we expect tenants modify their demands strategically in response such and prices. consider simple model provider offers single type virtual machine its devise...
To strike a balance between optimizing for energy versus performance in data centers is extremely tricky because the workloads are significantly different with varying constraints on performance. This issue exacerbated introduction of MapReduce over and above conventional web applications. In particular, batch interactive MapReduce, e.g., Spark system, availability locality drive while exhibiting degrees delay sensitivities. this paper we consider an minimization framework (which formulated...
Decentralised Machine Learning (DML) enables collaborative machine learning without centralised input data. Federated (FL) and Edge Inference are examples of DML. While tools for DML (especially FL) starting to flourish, many not flexible portable enough experiment with novel processors (e.g., RISC-V), non-fully connected network topologies, asynchronous collaboration schemes. We overcome these limitations via a domain-specific language allowing us map schemes an underlying middleware, i.e....