Allocating computing resources to various applications running on distributed computing platforms in order to satisfy real-time quality-of-service (QoS) requirements is an important problem, as is evident from a recent NASA research challenge problem concerning an autonomous hot-spot convergence system for the next-generation constellation-based satellite sensor webs. This paper presents a new resource management mechanism based on service levels and utility, and motivated by the hot-spot example, to address this problem. Existing approaches to adaptive resource and QoS management typically perform either reallocation of resources or adaptation of application QoS levels in response to dynamic environment changes. While success has been achieved with both of these approaches, this paper demonstrates the potential synergy of implementing both types of techniques. A proof-of-concept implementation is shown and our initial experimental results are analyzed.