Green and energy-efficient networking for cloud and content-delivery systems

The Internet has transformed to become a network of networks supporting traffic for various services and applications. Recently, cloud-computing and content-based services have become major contributors of this traffic, and are predicted to grow in future. Datacenters with large infrastructure and wide-area networks (WAN) have been the key enablers for the success of these services. Cloud-computing systems and content-delivery networks (CDN) use distributed dataceneters connected by a WAN to provide these specific services. Satisfying these increasing demands requires more physical equipment in datacenters and network facilities. Energy consumption of these infrastructures is also growing, increasing the electricity expense and greenhouse gas (GHG) emission concerns for infrastructure owners. Making datacenters and networks green and energy-efficient while providing scalable content and cloud-computing services, therefore, has become a major priority.;With the increase of content-based services, telecom network operators are investigating in building CDN systems over their networks. These CDNs can also take advantage of peer-to-peer (P2P) networking systems to increase scalability. In this dissertation, we propose energy consumption models for CDN, P2P, and hybrid CDN-P2P systems, and propose mathematical models for analyzing their energy efficiency. Heuristic algorithms for content replication and delivery have been devised to reduce energy consumption of CDNs. For hybrid CDN-P2P systems, our algorithms can exploit the inherent scalablity of P2P networks to decrease load on CDN servers while limiting system-wide energy consumption.;Increasing renewable energy or green energy consumption in datacenters is crucial to address the GHG emission concerns. However, irregular availability of renewable energy in unfriendly geographical locations create major challenges for its full utilization. Datacenter infrastructures are being built close to the renewable energy sites to help production and consumption of green energy. To maximize the benefits, efficient operation techniques in cloud are necessary for load balancing and service migration to match datacenter energy demand and green energy availability. To this end, we propose renewable-energy-aware virtual machine (VM) migration techniques after considering additional network energy consumption for migration to ensure effectiveness. We also propose VM migration bandwidth provisioning schemes to reduce resource and energy consumption in WAN.