Performance estimates

Question

Hello all,
We are designing a new system to be the back end of a network application. The basic premises are:
- very few simultaneous connections (probably no more than 4)
- datamodel consisting of some 20 tables, most of them tiny (300 rows or so)
- a couple of tables with approx. 300K rows
- most tables have 20 fields at most, 5 or 6 on average
- almost all fields are fixed width
- all users run a couple of different (simple) queries against one of
the main tables, with half a dozen joins against the tiny ones
- queries can be accessed via stored procedures or optimized in any
approppiate way
- there will be between 10 and 40 queries per second under normal load,
and up to 300 per second during peak load (a few minutes)
- query response time needs to be very fast, ideally about 50 milliseconds
The idea is to deploy this on a 10g RAC, consisting of two nodes of 1 CPU each, possibly small POWER5 systems. The physical DB will be stored in a shared high-end storage (FAStT), where the nodes will access through FiberChannel links.
Do you think it is possible to achieve this performance level (50 msec) with the proposed configuration?

Answer 1

It's possible yet irrelervant.
select * from table where pk_column='ABC' gives you result possible under 50 ms
select * from table1, table2, table3 when they have 300k rows, it gives you result for sure greater than 50 ms