Four Decades of Cluster Computing

Abstract

During the latter half of the 1970s high performance computers (HPC) were constructed using specially designed and manufactured hardware. The preferred architectures were vector or array processors, as these allowed for high speed processing of a large class of scientific/engineering applications. Due to the high cost of the development and construction of such HPC systems, the number of available installations was limited. Researchers often had to apply for compute time on such systems and wait for weeks before being allowed access. Cheaper and more accessible HPC systems were thus in great need. The concept to construct high performance parallel computers with distributed Multiple Instruction Multiple Data (MIMD) architectures using standard off-the-shelf hardware promised the construction of affordable supercomputers. Considerable scepticism existed at the time about the feasibility that MIMD systems could offer significant increases in processing speeds. The reasons for this were due to Amdahl’s Law, coupled with the overheads resulting from slow communication between nodes and the complex scheduling and synchronisation of parallel tasks. In order to investigate the potential of MIMD systems constructed with existing off-the-shelf hardware a first simple two processor system was constructed that finally became operational in 1979. In this paper aspects of this system and some of the results achieved are reviewed.