Cooperation Strategies for Enhanced Connectivity at Home

WHILE AT HOME , USERS MAY EXPERIENCE A POOR I NTERNET SERVICE while being connected to their 802.11 Access Points (APs). The AP is just one component of the Internet Gateway (GW) that generally includes a backhaul connection (ADSL, fiber,etc..) and a router providing a LAN. The root cause of performance degradation may be poor/congested wireless channel between the user and the GW or congested/bandwidth limited backhaul connection. The latter is a serious issue for DSL users that are located far from the central office because the greater the distance the lesser the achievable physical datarate. Furthermore, the GW is one of the few devices in the home that is left always on, resulting in energy waste and electromagnetic pollution increase. This thesis proposes two strategies to enhance Internet connectivity at home by (i) creating a wireless resource sharing scheme through the federation and the coordination of neighboring GWs in order to achieve energy efficiency while avoiding congestion, (ii) exploiting different king of connectivities, i.e., the wired plus the cellular (3G/4G) connections, through the aggregation of the available bandwidth across multiple access technologies. In order to achieve the aforementioned strategies we study and develop: • A viable interference estimation technique for 802.11 BSSes that can be implemented on commodity hardware at the MAC layer, without requiring active measurements, changes in the 802.11 standard, cooperation from the wireless stations (WSs). We extend previous theoretical results on the saturation throughput in order to quantify the impact in term of throughput loss of any kind of interferer. We im- plement and extensively evaluate our estimation technique with a real testbed and with different kind of interferer, achieving always good accuracy. • Two available bandwidth estimation algorithms for 802.11 BSSes that rely only on passive measurements and that account for different kind of interferers on the ISM band. This algorithms can be implemented on commodity hardware, as they require only software modifications. The first algorithm applies to intra-GW while the second one applies to inter-GW available bandwidth estimation. Indeed, we use the first algorithm to compute the metric for assessing the Wi-Fi load of a GW and the second one to compute the metric to decide whether accept incoming WSs from neighboring GWs or not. Note that in the latter case it is assumed that one or more WSs with known traffic profile are requested to relocate from one GW to another one. We evaluate both algorithms with simulation as well as with a real test-bed for different traffic patterns, achieving high precision. • A fully distributed and decentralized inter-access point protocol for federated GWs that allows to dynamically manage the associations of the wireless stations (WSs) in the federated network in order to achieve energy efficiency and offloading con- gested GWs, i.e, we keep a minimum number of GWs ON while avoiding to create congestion and real-time throughput loss. We evaluate this protocol in a federated scenario, using both simulation and a real test-bed, achieving up to 65% of energy saving in the simulated setting. We compare the energy saving achieved by our protocol against a centralized optimal scheme, obtaining close to optimal results. • An application level solution that accelerates slow ADSL connections with the parallel use of cellular (3G/4G) connections. We study the feasibility and the potential performance of this scheme at scale using both extensive throughput measurement of the cellular network and trace driven analysis. We validate our solution by implementing a real test bed and evaluating it “in the wild, at several residential locations of a major European city. We test two applications: Video-on-Demand (VoD) and picture upload, obtaining remarkable throughput increase for both applications at all locations. Our implementation features a multipath scheduler which we compare to other scheduling policies as well as to transport level solution like MTCP, obtaining always better results.