A Content-aware Resource Management And Scheduling Platform For Home Networks

Internet plays a significant role in our daily lives. As one of the most popular user networks, home network helps to connect the physical work to the information world. With the prosperity of mobile computing and smart home, home network draws the attentions of academic and industry communities.Data transmission is the fundamental service supplied by home networks, and the quality is always troubling users. Generally, networks treat all packets equally and do best-effort transmissions, when network resource contention occurs, the users would feel serious quality degradation. We find out that traffic flows contribute to the Quality of Experience (QoE) differently, thus differentiating traffic flows helps to improve QoE. Moreover, previous works about scheduling at home gateways cannot help to solve the trouble well, as the access network predominates the performance bottleneck. Based on the observations, we propose a content-aware access network flow scheduling system, Conan, which consists of two parts:home gateways and logically centralized SDN controller that controls the access network. Based on Conan, we propose a schedul-ing policy:expedited forwarding for interactive flows, bandwidth guarantee for HTTP video streaming and best-effort forwarding for the others. Experimental results show low latency of interactive flow and high QoE of HTTP video streaming.Device management is another trouble for users. In home network, the users have to access devices using IP address, which is tough for users while DHCP mechanism makes IP address vary with time. To solve this trouble, we design and implement a device name system, which enable users to access devices through their permanent registered name that is much more meaningful for users. Besides that, in order to satisfy the requirements of storage, we design and implement a storage system based on multiple cloud storage services, which employs file splitting and encoding to protect privacies.The main contributions of this thesis includes:(1) we design and implement a content-aware access network scheduling system, Conan. (2) We propose a class-based flow scheduling policy and a self-adaption bandwidth tuning algorithm for HTTP video flows. (3) To solve the device management trouble, we design and implement a device name system. (4) we propose a storage system based on multiple cloud storage space.