DSXchange

Hi- I have a simple job to load data from tables in one database to tables in other database.
The same job is parameterized and will run 20 times to load data which ranges from 1M - 10M

Oracle Connector-> Transformer -> Oracle Connector(Bulk Load)

I am achieving speed of 5000 rec/sec by designing the job with below properties.

4 node config file is being used.

Oracle connector properties (Source) :
Partition Reads method- Rowid Hash
Record Count- 1000000
Array Size - 500000

Maximum memory Buffer Size - 20971520
Buffer free run(percent) - 100
Queue upper bound size - 20971520
Dis write Increments - 13981000

To increase the speed of the job can I still increase the limits?
Are the above limits too much?
Does the buffer size utilize Scratch disk?
If I run similar jobs in parallel will it occupy my scratch disk and jobs will fail?

I suggest that you read the descriptions of buffering in the Parallel Job Advanced Developer's Guide.

Buffering is not a "more is more" situation. You may get less speed with larger buffers, if they're too big already and you don't have enough memory to support your settings.
Why have you set free run to 100%?

There are other factors at play here too. The buffers to which you refer are associated with the links in the job design. What, for example, are you using as the array size when selecting from Oracle? Why that particular value?

Thanks for the reply. As suggested I had gone through the guide.

But can you plz also suggest the settings I need to use to get maximum speed?
I am reading 10M records from table in one database and loading to other dbase. What values should be defined for before properties?

Oracle connector properties (Source) :
Partition Reads method- Rowid Hash
Record Count- ?
Array Size - ?

Maximum memory Buffer Size - ?
Buffer free run(percent) - ?
Queue upper bound size - ?
Disk write Increments - ?

My scratch disk space is 90 Gigs.

Row width?
Number of nodes?