The system management of the ICL goldrush parallel database server

Abstract

The GOLDRUSH MegaSERVER is a very high performance parallel database server sold by ICL. It consists of up to 64 Nodes each of which can collaborate to speed-up the execution of large, Decision Support queries, or high transaction workloads. Most uses of GOLDRUSH are for business-critical applications which require very high levels of availability, integrity and manageability. The design of the System Management of GOLDRUSH is the focus of this paper. It differs from conventional System Management as it must support both the underlying parallel machine and the business critical applications which run on it. We describe the requirements placed on System Management by this class of machine, and how these were met in the design of the GOLDRUSH management system: we explain both the management architecture and the management tools themselves. Finally in the light of experience in the use of these tools we point to future directions for work in this area.