When I was with my former company, we did do a spike test. It was a large billing system (J2EE/WebLogic) running against a 2 TB, 3-node, 9.2.0.4.0 RAC database (I have more specs if you're interested). We did see a nominal gain but it really paled in comparison to the gains from fixing bad SQL (no binds, indexing, etc.). What our SA told us was that if we changed any of the variables such as workload, the spiked version of the binary could be worse than the original. So you'd have to re-evaluate performance from time-to-time. I'm trying to contact the SA who did the spike with us in case you have more specific questions.
The 9i and 10g Administrator's Reference , Appendix D /F on Tru64 mentions "spike" as a performance optimization tool on Tru64. eg
http://download-west.oracle.com/docs/cd/B19306_01/server.102/b15658/appf_tru.htm
http://download-west.oracle.com/docs/cd/B19306_01/install.102/b14206/pretru.htm (even the 9iR2 and 10gR1 certification pages mention "spike"). Apparently, the last version release of spike was in 2002. I cannot find much information on it at HP's site. (http://h30097.www3.hp.com/dtk/spike_ov.html)
Has anyone used "spike" ? Under what sort of database/application load did you use "spike" ? What performance gains did you obtain ?
Hemant
Hemant K Chitale http://web.singnet.com.sg/~hkchital
-- //www.freelists.org/webpage/oracle-l