Abstract. Emerging grid computing infrastructures such as cloud com-puting can only become viable alternatives for the enterprise if they can provide stable service levels for business processes and SLA-based cost-ing. In this paper we describe and apply a three-step approach to map SLA and QoS requirements of business processes to such infrastructures. We start with formalization of service capabilities and business process requirements. We compare them and, if we detect a performance or reli-ability gap, we dynamically improve performance of individual services deployed in grid and cloud computing environments. Here we employ translucent replication of services. An experimental evaluation in Ama-zon EC2 verified our approach