A type of multiprocessor or architecture in which there is a hierarchy of the units of replication. At the lowest level, "processors" are replicated to form a "cluster". The cluster consists of M processors and a shared switching network that provides communication among the processors and access to a shared local memory. At the next higher level, the clusters are replicated. A clustered "system" consists of N clusters interconnected through a global network that allows communication among the clusters and access to a global memory that is shared among the cluster. The purpose of clustering is to reduce conflicts in accessing shared resources, whether these resources are the communication network or the storage system. Typically, a clustered architecture is a "hyperstar", that is, a star of stars. The cluster is a "star of processors" sharing a local communication network and a local memory. The system is then a "star of clusters" in which the shared resources are the global network and the global memory.