Skip to Main Content
The feasibility of using adaptive object migration to enable the execution of heavy applications in pervasive environments, is determined by the computational efficiency of adaptation algorithms and the efficacy of their decisions. These two factors, which are largely predicated by the resource constraints of devices, are heavily influenced by the granularity at which adaptation decisions are performed. This paper proposes a new type of adaptation granularity which combines the efficiency of coarse level approaches with the efficacy of fine-grained adaptation. A novel approach for achieving this level of granularity through the dynamic decomposition of runtime class graphs is presented and empirically evaluated on a corpus of real world applications. It is shown that the approach improves the efficacy of adaptation decisions by reducing network overheads by a minimum of 17% to as much 99%, while maintaining comparable decision making efficiency to class level adaptation.