The lack of a single system restricts the use of parallel processing on clusters. We propose an approach for building an efficient single system image cluster Operating System (OS). The proposed system is based on container, a concept which provides the abstraction of shared memory segment at the cluster scale. Small pieces of software, called linkers, are used to integrate containers in an existing single node OS. Hence, the existing OS programming API is kept unmodified but the underlying OS services can take benefit of all cluster resources. Our approach has been validated by a prototype based on Linux. Existing applications running on symmetric multiproc- essors (SMP) on top of Linux can be executed on top of our cluster OS without modification.