Re: Get Oracle8 to simulate

sascha.roser_at_ubs.com wrote in message <7h99ph$umt$1_at_nnrp1.deja.com>...
>Is there possibility to get the oracle8.0.4 on a HP-UX maschine
>to simulate "high data volumes" and/or "large numbers of users"
>in order to test an extraction/presentation tool

IMHO high data volumes are more related to the number of rows than to the size of the of the database. After all, physical size is addressed by the disk controllers, onboard disk caches and stuff like that. So I think you can make a fairly safe assumption that the speed of physical disk access is not a function of Oracle - and whether a row is 180 bytes or 1MB, the time for Oracle to physical retrieve that row is a constant of hardware used.

What I have done to play with high volumes is to create large numbers of small rows - for example 100 million rows of transaction data containing very basic data such as invoice numbers, customer sequence numbers, product sequence numbers, date of sales and prices. The flat file containing the data was less than 1.8GB - with the Oracle overheads, the resulting table should not be more than 2.5GB. And there you have a VLT for stress testing the database and hardware.

Large number of users - well, decide what the typical user is going to do. Come up with a small set of SQL statements (update, deletes and queries). Then store these in a file and use a UNIX/NT script to play the role of a user using SQL*Plus. In this way you can simulate a whole bunch of clients. Of course, you should not run these "clients" on the server platform.

To create random access SQLs (for updating, deleting and selecting data) should be just as simple. As sequence numbers are used as the keys in these tables, all you need to do is use a random generator function to return a random number and use the random number to access unique rows for updating and deleting.

Hope I'm making some sense here. :-)

regards,
Billy Received on Wed May 12 1999 - 06:21:34 CDT