Re: Pinning Tables in Memory

The volume of data compared to the amount of memory tends to suggest that getting terrific performance shouldn't be an issue; and the tables and matching indexes are likely to be permanently buffered by the nature of the activity.

On the other hand, if there is a need to improve performance, it is possible that the 99% hit ratio is an indication of a problem rather than something to be pleased about - it is, after all, relatively easy to get arbitrarily high hit ratios by doing things the wrong way and wasting resources: see

http://www.jlcomp.demon.co.uk/myths.html

--
Jonathan Lewis
Yet another Oracle-related web site:  http://www.jlcomp.demon.co.uk

Practical Oracle 8i:  Building Efficient Databases
Publishers:  Addison-Wesley

Reviews at: http://www.jlcomp.demon.co.uk/book_rev.html



Howard J. Rogers wrote in message ...


>

>> Assume 1 Gig of RAM.

>>

>> 1. What performance boost could I get for pinning all tables in an

>> address database (static, read only for the most part) (streets table -

>> 20K rows, blocks table 300 K rows, units table 100K rows).  If the cache

>> hit ratio indicates > 99%, is pinning any tables a waste of time

>> (address reads frequent, tables indexed, queries use indexes)?

>

>If you've a 99% cache hit ratio on the buffer cache, you are doing

>extraordinarily well (the usual target recommended is around 85%).  It

>effectively means you have managed to cache everything you require in

 memory


>anyway -so I'd suggest further time spent flogging this particular horse

>would be a waste.

>