Conference ulysse::rdb_vms_competition

1036.0. "I NEED ORACLE EXPERTS FOR HELP...." by BLG03::BRAZZOLOTTO () Fri Dec 06 1991 11:04

	I need help from ORACLE experts.......!!!!!!

	I have to size a system for a customer that uses ORACLE.
	By the tests it follows that CPU time is very close to
	elapsed time, even if there are a lot of select, update,
	insert and delete.
	
	I think may be significant an example of the tests I made.

	The database is about 2Gb large, with, among others, 
	table A small, table B 15 Mb, table C 4 Mb, table D 34 Mb,
	table E 140 Mb. It is split up over four disks.
        The SGA is 10 Mb. The system test is a VAX 6420.

	A batch job reads about 8000 rows in table A , inserts and reads 
        20000 rows in other tables, inserts 30000 rows in table B and 
	table C, updates 20000 rows in table D an table E, deletes 10000
        rows. It navigates in these tables via index, and ends with these 
	accounting informations :

	 buffered I/O count:   1098      peak working set : 20025
         direct I/O count:    13280	 peak page file :   31201
	 page faults:	      57135	
         charged cpu time: 0 00:48:21.44 elapsed time: 0 00:54:54.60

	This suggest me that I have no possibility to improve performance
	and response time, unless upgrading CPU.

        Is this a normal situation with ORACLE ?

	Is it reasonable expect to have the same ratio (between CPU and
        elapsed) as the database increases ? ( final size of DB will be
        15 Gb ).	
        
	 Thanks for any help 
	
				Patrizia.