Firms are increasingly dependent on data networks to complete day-to-day operations such as video conferencing, chat/online customer services, real time browsing/E-Commerce applications and other data applications. In general, a firm performs two types of tasks over a network, time-fixed and size-fixed. These tasks might have to be completed using networks with different quality of service levels (QoS) and pricing.; We investigate the customer's optimal behavior when there are quality related costs in addition to the cost of acquisition while using networks with bandwidth and QoS constraints. We consider the customer's cost minimization problem assuming that an environment exists in which network capacity can be purchased at competitive prices from different suppliers with different service quality. We model QoS as a function of loss probability, delay and jitter. We also assume that the suppliers can offer any quality of service and some capacity at competitive prices. Therefore, the customer has access to sufficient resources with different QoS levels, capacities, and prices to perform tasks with different QoS and capacity requirements. Existing literature on QoS mostly focuses on supplier side issues. We focus on the customer's point of view at a tactical level. This study, to the best our knowledge, is among the first to formulate and solve this problem.; First, we analyze the problem with all-you-can-send pricing where the supplier charges a fixed price for a specific bandwidth and duration. We show that the resulting problem is at least as hard as the two-dimensional bin-packing problem. We develop a heuristic (Heuristic A) that can take advantage of the trade-off between the cost of acquiring resources and the opportunity cost of degradation in realized quality. We analyze how a customer behaves as a decision maker under different pricing settings through experiments. The customer's strategy varies based on prices, capacity and quality of resources. Given the opportunity, customers mix and match resources to optimize their position.; The underlying problem is a computationally intractable problem. A relaxation of this problem can actually create workable solutions for real life situations. On that end we develop a lower bound (LB) formulation and solve it optimally. The solutions are reasonable when prices set randomly.; A second formulation, which is a more realistic relaxation of the original problem, is solved optimally by the Generalized Bender's Decomposition (GBD) method. We show that GBD solves the small problems in reasonable time and the solutions are very realistic. However, more work is needed to make it work in large problems.; Finally we analyze a different formulation using a pricing scheme called tax-band pricing in which the total price of a resource is a convex piecewise linear function of capacity. We show that this problem is easier to solve.
展开▼
