RaaS (Resource-as-a-Service) clouds represent a new market and technology driven paradigm shift from renting prebuilt virtual machines towards renting resources to compose elastic machines. By offering cloud users fine-grained resources, including CPU, memory, disk and network based on their demands, RaaS clouds have the potential to address several inherent problems in IaaS clouds. RaaS architecture requires a new service layer to abstract and compose heterogeneous and dynamic fine-grained cloud resources at high frequency. This paper proposes a REST service framework consisting of Resource-Oriented Network (RON) and monadic service composition. RON abstracts heterogeneous cloud resources with uniform REST resources and represents their dynamic relations with hypertext. Monadic service composition provides a concise functional programming language with monad to concurrently control large scale RON through dynamically generated workflows. A prototype system has been implemented based on Docker containers and Linux control groups. Our experimental results show that the approach is feasible and the performance is consistent with the client and server workloads.