US sources say ICL Plc will shortly detail its ambitious parallel server system – codenamed Goldrush – which it is pitching as a relational database engine for commercial transaction processing, today’s issue of our sister paper Unigram.X reports. Due in 1994, the system can scale up to 256 superscalar Sparc RISC chips and runs Chorus Systemes SA’s Unix System V.4 microkernel. Goldrush – it will appear under another name – turns into a product work that has been going on since 1989 under the auspices of a European Community Esprit II project, the European Declarative System, EDS. Compagnie des Machines Bull SA, Siemens Nixdorf Informationssysteme AG, ICL and others all benefitted from Community cash. The European Declarative System will spawn four projects and under Community rules, each member can take the results and implement them commercially. ICL is optimising the server to run corporate relational databases at the high end of the commercial market where it will hook into host systems, Unix or proprietary. Goldrush uses up to 128 pairs of superscalar Sparcs – possibly Cypress Semiconductor Corp’s HyperSparc, which is now owned by ICL’s parent, Fujitsu Ltd – one CPU for the application, the other to drive the network, each with up to 64Mb memory. Key to the architecture is Delta, a Siemens Nixdorf-developed, 25M-bytes per second high-speed interconnect bus that enables database queries to be spread over multiple CPUs and file systems. The Oracle database has been optimised for the server, which will cope with SQL queries from a variety of sources including ICL OpenVME and DRS 6000 boxes via a distributed Extended SQL interface – an SQL extension developed under the Declarative Systems project to exploit parallelised databases. Until now, ICL says, commercial users have had to re-engineer applications to take advantage of parallel architectures, and have been unwilling to do so. The moment database vendors come out with versions that do it automatically, the rules change, it says. Software optimised to run on the system includes a database request manager and data manager – the request manager translates ESQL queries for parallel execution.
