Description
Date depot: 1 janvier 1900
Titre: Mesure et analyse des services de transport pour les applications cloud
Directeur de thèse:
Dario ROSSI (LTCI (EDMH))
Domaine scientifique: Sciences et technologies de l'information et de la communication
Thématique CNRS : Non defini
Resumé:
Internet, born as an U.S. military research experiment, is now a worldwide network used by more than three million1 people every day. In addition to interconnecting more and more users, it supports many and diverse applications related to business (e.g., e-commerce, Internet banking), professional activities (e.g., e-mail, videoconferencing, collaborative document editing), education (e.g., e-learning, crowd-sourced encyclopedia), entertainment (e.g., video streaming, on-line gaming), and personal interactions (e.g., instant messaging, social networking, photo sharing).
The set of services and platforms related to Internet is continuously expanding, increasing the number of activities that users can conduct online considerably, but also impacting the architecture of the network. Listing few examples: (i) The rise of Content Delivery Networks (1) redefined the paradigm of content distribution and availability; (ii) Cloud Infrastructures (2) are customary for storing and processing huge amount of data; (iii) Mobile Devices (3) are now a common means to access web pages and make use of Internet services; (iv) Cyber Threats (4) are becoming an urgent open issue and a source of troubles, considering the amount of sensitive information being exchanged over the network.
Due to the pervasiveness and the heterogeneity of the network, researchers and professionals are facing a constantly renewed interest in understanding the dynamics interacting in such ever-evolving ecosystem. In order to provide a comprehensive understanding, there is the strong need of tools and techniques able to inspect and to characterize various aspects of network traffic.
The proposed research activity has its foundations in active measurements campaigns and passive network traffic monitoring. In the former case, active measurements are made possible through the development of ad-hoc test-beds that are instrumented to perform programmable operations and to extract relevant metrics from the network traffic. Operations to be accomplished and meaningful metrics may change according to the scope of the experiment (i.e., different applications types will require different test-beds and measures) and to the final goal of the research.
Passive monitoring is instead performed with the advantage of a network probe that observes packets flowing through the monitored link. The probe is required to run a software, e.g. (5), able to reconstruct and abstract traffic flows through the annotation of statistics including (but not limited to) the IP address and the port number used by client and server, the amount of bytes sent and received, the flow duration, the application layer protocol, etc. The network probe needs to be installed in a place where the traffic of a significant amount of hosts (order of thousands) can be monitored. Suitable installation places are, for instance, the Point of Presence of a large-scale ISP, the border router of a University campus, or the link connecting a corporation to the Internet. This is required in order to collect data of a relevant population of users, and prevent unwanted biases caused by the finitude of the collected data.
This hybrid methodology allows the study of a huge variety of topics in the computer networks field. The proposed research work focuses on network traffic characterization and measurements applied to three main areas: (i) Cloud Storage Services for personal and data-intensive applications; (ii) Network performance assessment and evaluation of quality perceived by end-users related to the deployment of new protocols and technologies; and (iii) Network events correlation and traffic mining applied to security aspects and end-users protection. In the following a brief description about the intended research on each of the mentioned topics is provided.
Cloud Storage
In recent years, data storage became a fundamental service with companies, universities and also private end-users having the need of storing large amount. Cloud storage services are emerging as strong alternative to local storage, allowing professional customers to save the costs and the burdens of buying and maintaining expensive hardware (6), (7), and attracting private enthusiasts to backup content with great simplicity and synchronize multiple devices seamlessly (8). Despite the popularity of these services, very little is known about their features: Where are data stored? Which technologies are adopted to transfer them in a secure way? Which is the workload that the cloud infrastructure has to handle? Which traffic load that has to be delivered by the network? Are these services efficient during common usage? Many of these questions still do not find an answer.
Doctorant.e: Bocchi Enrico