A Distributed Approach to Local Adaptation Decision Making for Sequential Applications in Pervasive Environments

The use of adaptive object migration strategies, to enable the execution of computationally heavy applications in pervasive computing spaces requires improvements in the efficiency and scalability of existing local adaptation algorithms. The paper proposes a distributed approach to local adaptation which reduces the need to communicate collaboration metrics, and allows for the partial distribution of adaptation decision making. The algorithm's network and memory utilization is mathematically modeled and compared to an existing approach. It is shown that under small collaboration sizes, the existing algorithm could provide up to 30% less network overheads while under large collaboration sizes the proposed approach can provide over 900% less network consumption. It is also shown that the memory complexity of the algorithm is linear in contrast to the exponential complexity of the existing approach.