Systems and Networks
Distributed live streaming on mesh networks
Luca Baldesi
Publications | luca.baldesi [at] unitn.it (Email) | Website
More...
The demand of streaming services increases continuously. Typical centralized data distribution schemes poorly adapt to the scenarios of mesh networks. On the other hand, decentralized, cooperative distribution is capable of granting the flexibility and the high degree of scalability that the forecast scenario requires. It is crucial to maximize the efficiency and effectiveness of cooperative distribution of multimedia applications by developing new theoretically models and new methodologies according to the QoE requirements. Centrality metrics and cross-layer techniques can be further investigated in this context and can play a major role in future optimized architectures.
Strategic allocation in highly distributed and heterogeneous computing and information systems under concurrency and multitenancy
Francescomaria Faticanti
Publications | f.faticanti [at] unitn.it (Email)
More...
The emerging paradigms for cloud computing involve multi-level systems with different owners, multiple islands of the infrastructure and edge components with additional hardware. The decision on how to allocate resources in such a dynamic and complex system involves the decision of several actors having possibly conflicting objectives. Hence, the resulting operating point of the system is established by actors’ utilities, the economic figures into play and the mechanisms implementing pricing and resources matching. Our aim is to investigate the resource management and the task scheduling in such systems, and find new efficient methods to improve the performance of this kind of systems.
Improved routing through reinforcement learning and stochastic optimization
Antonio Massaro
Publications | antonio.massaro [at] unitn.it (Email)
More...
In software defined networks, routing is performed by a central controller, which installs flow fules for any new flow entering a switch in the network. If the installation requests exceed the communication capacity between switches and controller, the system can underperform. We study algorithmic solutions based on stochastic optimization and reinforcement learning, to allow switches to take local, autonomous decisions with respect to flow routing. The objective is to defer some routing decisions to the switches, without the need of interrogating the controller, hence decongesting the communication channel between it and the switches.
Prospects and Challenges for Future Transport Networks
Federico Pederzolli
Publications | f.pederzolli.1 [at] unitn.it (Email)
More...
The rate of growth of Internet traffic puts ever increasing strain on transport networks, which are realized with a mixture of packet (IP/MPLS) and circuit (Optical) technologies. This project aims to increase the effectiveness of transport networks by jointly allocating resources in both layers and exploiting recent advances in the optical domain.
Risk assessment for Cyber-Insurance
Ganbayar Uuganbayar
Publications | ganbayar.uuganbayar [at] unitn.it (Email) | Website
More...
As many surveys outline cyber insurance is an immature and faces many problems regardless its fast-growing market. The most troublesome are lack of available statistics, information asymmetries, risk correlation, security interdependence and risk assessment that is the crucial one. The risk assessment practice faces a number of difficulties, to name a few: lack of available data, rapidly changing threat landscape, quick evolution of technologies and agile systems, large intangible losses, etc. Taking into account the above-mentioned facts, the research in risk assessment for cyber insurance is challenging and promising at the same time.