Algorithms for Services with Multiple Levels of Quality

Abstract

We consider the problem of serving a set of sensors/end-users at multi-level quality of service. Consider for example the problem, which arises in augmented reality applications, of single-hop radio broadcasting a common content from a provider. Each end-user requires a minimal quality level of the common content. The provider must satisfy all end-users at their required minimal quality level with a pointto-point encrypted transmission. Since the number of different quality level versions of the common content cannot be as large as the number of end-users, the provider has to partition the end-users into groups. The provider adopts for each group a quality level which is the maximum among the levels required by the end-users of that group. In doing so, the provider incurs into an encryption and transmission cost which is proportional to the product between the end-user group size and the adopted quality level. Such a problem consists in finding a partition of the end-users that satisfies their requests and minimizes the overall provider cost. These quality of service problems are optimally solved in polynomial time by dynamic programming techniques for an arbitrary number of end-users. A faster greedy heuristic is also proposed which quickly finds a solution very close to the optimum one.