PhD Alumni

Poli Alessandro

Present position: Temporary Research Assistant at Department of Electronics and Information - Politecnico di Milano, Italy

Thesis title:  Design of Large-Scale Distributed and Peer-to-Peer Infrastructures with Quality of Service
Advisor:  Paolo Giacomazzi
Research area:  Telecommunications Networks
Thesis abstract:  
The first part of my work is devoted to the performance analysis of peer-to-peer video streaming systems. Peer-to-peer systems are an effective way for content providers to distribute media content to a large set of users, with limited investments for computing and networking infrastructures. These systems rely on the provisioning of network and computing resources by users that are receiving video streaming services. These systems appear to be self-scalable, since an increase in the number of users is compensated by an additional availability of resources, at no costs for content providers. However, performances of these systems can be greatly affected by the uncontrollable behavior of the users, who can disconnect without notification, consequently impacting on the quality of the video stream received by other users. Results obtained with this work can be beneficial to content providers in order to evaluate the cost/benefits tradeoff that peer-to-peer systems can offer; moreover, results can be exploited by designers of peer-to-peer video streaming systems for increasing the efficiency and the overall performance.
The second part of my work is devoted to present two additional themes, both related to the design and the optimization of ICT (Information and Communication Technology) infrastructures. Two types of infrastructures have been considered: distributed infrastructures, made up by server farms running applications used by the users of an information system, and municipal wireless access networks, used to provide a wide range of services, including the internet access, to the citizens of large cities. The first scenario refers to the design of distributed infrastructures using the new Application Oriented Networking technologies, in order to jointly allocate computing and network resources; this approach is innovative, since current literature faces the two problems separately. The adopted optimization methodology minimizes costs, given a predefined performance level. The second research scenario is the optimization of municipal wireless networks, given the conformation of the ground, buildings’ positions, the list of physical supports able to host access points, and the positions of interconnection points towards the backbone network. The methodology is able to identify the minimum cost solution that satisfies end-to-end performance constraints.
These two themes share with the peer-to-peer video streaming analysis, the property of referring to systems whose optimization, by means of the algorithms commonly used in the literature, is not feasible. The problem stands in the lack of scalability of the tools and the algorithms used by the academic community – but also in the industry – when facing with large scale infrastructures, such as the ones considered in this thesis. The scalability issue is mainly given by implementation problems, and also by the algorithmic approach; the search for a strictly optimal solution can be put in place only if the size of the problem is very limited. For example, in the field of municipal wireless networks, current best algorithms, aiming at finding optimal solutions, are able to cope with a maximum of 10-30 nodes, while interesting real case scenarios must take into account hundreds of thousand candidate node sites. A gap of such extent between current algorithms and the real size of the problems, from our point of view, is substantial, and needs to be bridged.