We carried out a detailed analysis of the I/O subsystem performance, including a study of its scalability and of its impact on the overall performance of the parallel server. In addition, we implemented a number of parallel data query operations on top of Oracle Parallel Server v.6.2 to investigate the limits of the technology and to evaluate the potential of the upcoming Parallel Server v.7.1 which will support parallel execution of the decision support type queries.
The current configuration of the server includes a 64 processor array with 2GB of RAM, 32 I/O nodes with SCSI controllers forming a scalable parallel I/O subsystem, 0.2 TB on-line disk space, and additional I/O nodes for control and networking. The server runs Oracle RDBMS v. 7.0.13. The server is currently being integrated with the high bandwidth NPAC LAN and with the ATM-based New York State NYNET wide area network. A number of commercial and military/dual-use applications are currently under discussion with the InfoMall partners, including fraud detection, imagery on demand, real-time news video server, GIS applications, and real-time market analysis.
Further plans for development of the parallel database technology in NPAC include beta testing of the IBM SP1 version of the Oracle Parallel Server and integration of a number Oracle tools, including their text retrieval system, with the Parallel Server.
2) S. Costicoglou, M. Podgorny, A. Choudhary, "Performance Evaluation of the Oracle Parallel Server v. 6.2 on nCUBE2", NPAC Technical Report, SCCS 489, May 1993.
3) S. Costicoglou, M. Podgorny, A. Choudhary, "On Benchmarking the Oracle Parallel Server on nCUBE2", Proceedings of the IEEE Workshop on Advances in Parallel and Distributed Systems, Princeton, October 1993, IEEE Computer Society Press, in press.
4) S. Costicoglou, M. Podgorny, "Parallel Query Processing with Oracle v. 6.2 on nCUBE2", NPAC Technical Report, SCCS 504, August 1993.