|dc.description.abstract||The current Internet presents a high barrier to entry for new service providers, due to its inability to accommodate new protocols and technologies, and lack of competition among the network providers. Recently, network virtualization has gained considerable attention as a possible solution, as it enables multiple networks to concurrently run over a shared substrate. It allows for deploying diverse network protocols and technologies customized for specific networked services and applications. Moreover, any party can take on the role of a network provider by simply offering his virtual network infrastructure to customers, increasing competition in the market. However, the ﬁrst challenge in realizing a fair and competitive market in a virtual network environment is to have a service negotiation and contracting mechanism in place, that will allow (i) multiple infrastructure providers to participate in a fair and faithful competition, and (ii) a service provider to negotiate the price and quality of service with the providers.
In this thesis, we present V-Mart, an open market model and enabling framework for automated service negotiation and contracting in a virtual network environment. To the infrastructure providers, V-Mart fosters an open and fair competition realized by a two
stage auction. The V-Mart auction model ensures that bidders (infrastructure providers) bid truthfully, have the flexibility to apply diverse pricing policies, and still gain proﬁt from hosting customers’ virtual resources. To the service providers, V-Mart offers virtual network partitioning algorithms that allow them to divide their virtual networks among competing infrastructure providers while minimizing the total cost. V-Mart offers two types of algorithms to suit different market scenarios. The algorithms not only consider virtual resource hosting price but also the service provider’s preference for resource co-location and the high cost of inter-provider communication. Through extensive simulation experiments we show the efficiency and effectiveness of the algorithms under various market conditions.||en