universe and oracle on the same hardware
universe and oracle on the same hardware
This is a topic for an orphaned message.
There is a MAJOR performance benefit to be had when UniVerse and Oracle are on the same hardware, and that is to use what Oracle calls "bequeath" drivers. These use named pipes as the conduit to TNS listeners. Contact your Oracle support folk for more info, and push the issue if they plead ignorance. Bequeath drivers DO EXIST (at least for Oracle version 7). At one DataStage site we changed from "ordinary" Oracle drivers to bequeath drivers and got a large load down from 8 hours to 40 minutes. This is a genuine improvement - there were no other changes made.
The only other caveat is sizing message queue, shared memory and semaphore parameters in the UNIX kernel. Make sure theres enough to go around!
-------------------------------------------
Sara Burns wrote:
Roger asked if any third-party software is required for universe BCI to retrieve data from Oracle tables.
You need to have the ODBC drivers available for Oracle to allow this to come together. We use a product from OpenLink with good success. We both write to Oracle and also read from Oracle. The major difference is that you need to write SQL statements not use normal Read and Write. Once you get comfortable with these it is quite easy, just "wordy".
My experiments recently reading extensive data from Oracle is a major recommendation to use what Oracle calls Bind variables. This makes you very popular with the Oracle DBA and gives major performance enhancements. See the calls in the manual of SQLBindParameter(), SQLPrepare and SQLExecute. I think this gave me performance benefits of between 2 and 5 times speed improvement. If you do not do this your statements fill up the Oracle Shared Pool as each is parsed separately. Using bind variables the same statement is reused, just substituting the correct data each time.
We are running both Universe 9.5.2.2 and Oracle 7.3.4 on the same Sequent NUMA-Q running dynix 4.4. Sequent can only use OpenLink drivers but other unix flavours have more drivers available.
Hope this helps
Sara Burns (SEB)
Project Leader
Public Trust
Phone: +64 (04) 474-3841 (DDI)
or (04) 474-3770 Ext. 720
The only other caveat is sizing message queue, shared memory and semaphore parameters in the UNIX kernel. Make sure theres enough to go around!
-------------------------------------------
Sara Burns wrote:
Roger asked if any third-party software is required for universe BCI to retrieve data from Oracle tables.
You need to have the ODBC drivers available for Oracle to allow this to come together. We use a product from OpenLink with good success. We both write to Oracle and also read from Oracle. The major difference is that you need to write SQL statements not use normal Read and Write. Once you get comfortable with these it is quite easy, just "wordy".
My experiments recently reading extensive data from Oracle is a major recommendation to use what Oracle calls Bind variables. This makes you very popular with the Oracle DBA and gives major performance enhancements. See the calls in the manual of SQLBindParameter(), SQLPrepare and SQLExecute. I think this gave me performance benefits of between 2 and 5 times speed improvement. If you do not do this your statements fill up the Oracle Shared Pool as each is parsed separately. Using bind variables the same statement is reused, just substituting the correct data each time.
We are running both Universe 9.5.2.2 and Oracle 7.3.4 on the same Sequent NUMA-Q running dynix 4.4. Sequent can only use OpenLink drivers but other unix flavours have more drivers available.
Hope this helps
Sara Burns (SEB)
Project Leader
Public Trust
Phone: +64 (04) 474-3841 (DDI)
or (04) 474-3770 Ext. 720