Abstract-Many network operators offer some type of tiered service, in which users may select only from a small set of service levels (tiers). Such a service has the potential to simplify a wide range of core network functions, allowing the providers to scale their operations efficiently. In this work, we provide a theoretical framework for reasoning about and tackling algorithmically the general problem of service tier selection. Drawing upon results from discrete location theory, we formulate the problem as a p-median problem under a new directional distance measure, and we develop efficient algorithms for a number of important variants. Our main finding is that, by appropriately selecting the set of service levels, network providers may realize the benefits of tiered service with only a small sacrifice in network resources.

Abstract-Many network operators offer some type of tiered service, in which users may select only from a small set of service levels (tiers). In this work, we study the problem of designing a tiered-service network that allocates bandwidth in multiples of a basic bandwidth unit. Such a packet-switched network can enjoy many of the benefits, in terms of control and management, of a TDM network, but without the associated data plane rigidities.

Abstract-Many network operators offer some type of tiered service, in which users may select only from a small set of service levels (tiers). Such a service has the potential to simplify a wide range of core network functions, allowing the providers to scale their operations efficiently. In this work, we study a number of problem variants related to service tier selection. Our contributions include: 1) a faster algorithm for obtaining optimal service tiers; 2) a new formulation and optimal algorithm to optimize jointly the number and magnitude of each service tier; and 3) the concept of "TDM emulation" in which all service tiers are multiples of the same (software-configurable) bandwidth unit, and a suite of algorithms to select jointly the basic unit and service tiers. Our work provides a systematic framework for reasoning about and tackling algorithmically the general problem of service tier selection, and has applications to a number of networking contexts, including access networks (e.g., determining the tiers for ADSL, cable modem networks or PONs) and core networks (e.g., LSP sizing for MPLS networks).