Utility maximization for resolving throughput/reliability trade-offs in an unreliable network with multipath routing
Vladimir Marbukh · 2007
This paper proposes a framework for balancing competing user (i.e., application) level requirements by resolving the corresponding trade-offs in a distributed system with limited resources. Assuming that each user’s preferences can be characterized by some utility function, the goal of balancing competing requirements for each user as well as across different users is to maximize the aggregate utility. The framework assumes a presence of Intelligent Plane, which isolates users from details of the network properties and mechanisms of implementation of the user level requirements. The Intelligent Plane performs the following tasks: (a) maps the user level requirements into the network resource requirements, (b) maps the resource congestion prices into prices of the user level requirements, and (c) maps the user willingness to pay for the user level requirements into payments for the specific sets of resources. Once payments for the specific sets of resources are identified, the resources are allocated to the users by a “TCPfriendly” algorithm. The paper discusses this framework for a particular case of balancing user requirements for throughput and survivability in an unreliable network, where survivability is achieved through redundancy, e.g., using multipath routing.