Oracle BIWA Summit 2007

This wraps up my trip to Reston, VA.  I am heading back to Philadelphia.  It has been a very worthwhile experience.

1:40pm

Hanming and I had the pleasure to sit down with Jean-Pierre Dijcks and Antonio Romero at lunch for over an hour.  We started off with giving them (well-deserved) props for a solid product, as well as the time and cost savings we already realized with only 5 months in to production.

Topics we discussed:

• Lineage and impact analysis fail when case statements or pivot/un-pivot operators exist in a map
• Business drivers and problems can be solved by incorporating specification authoring into OWB; or, even better, combining the two tasks into one unified GUI: specifications authoring and map creation
• The uphill battle of data quality/standardization we fight everyday, especially lab test labels and unit conversion.  What can we do to first make it a linear problem, then experience complexity reduction over time (i.e., present possible matches through a scoring mechanism, save the manual matches into a library, then reuse what it learns)?
• How will the new flashback archive feature in 11g how us reduce the maintenance overhead for audit trail for ViewPoint and other applications?

They are committed to come on-site, possibly in December.  We talked visiting their HQs the Friday of OpenWorld.  JP suspected some of our challenges are solvable using Experts (i.e., OMB+, coupled with Java).  So, he hopes Dave Allen can help us out as well.   We will send them some of our problems and go from there.

10.2.0.3 porting is still underway, citing resource challenges.  Though, the upgrade path from 10.2 to 11.1 will be very straightforward, as simple as applying a patch.

Let's go straight to the first keynote speech.

8:00am

Rich Niemiec, CEO, TUSC

Title: "How Oracle Came to Rule the Database World and Why They Will Rule the BI World"

• Everything began with an ACM paper by E. F. Codd in 1968: "A relational model of data for large shared data banks" URL: http://portal.acm.org/citation.cfm?id=362685
• INGRES: Interactive Graphics Retrieval System, by Michael Stonebraker in 1972
• Sybase was license to Microsoft in 1992
• 1977: Oracle began as SDL (Software Development Laboratories)
• Ed Miner and Larry Ellison worked together at Ampex; Oracle was a project code name for CIA
• There was never a version 1; Larry Edison thought nobody would buy a version 1 of any software
• 1982: RSI changed its name to Oracle
• 1983: Version 3 was rewritten from PDP11 assembly language to C to create a portable code base.  It is also the first 32-bit RDBMS
• 1984: Version 4 released; ported to PC to compete with DBase; first RDBMS with read consistency
• 1985: Version 5 released; support rollback; last version before Oracle's IPO
• 1986: Oracle IPO went on March 12; Microsoft went IPO on March 13; Sun went IPO on March 8
  
• 1987: Oracle Applications group started
• 1988: Version 6 released with total rewrite for transaction processing; PL/SQL is introduced; row-level locking
• 1992: Version 7 released with parallel query, triggers, stored procedures, and security features
• 1995: Oracle began to focus on OLAP
• 1997: Version 8i released with browser-based GUI; OWB introduced
• 2000: Version 9i released
• 2004: Version 10g released with grid technology; flashback everything (database, table, drop)

I was completely exhausted after 12 hours of "meet n' greet," keynote speeches, workshops, and blogging.  Yet, I managed to venture out to the Reston town center last night.  There were a lot of people out on the street.  Tuesday was their "restaurant/entertainment night."

A cup of Starbucks joe wrapped up the day. 

Title: "Driving Your Data Integration Through Business Rules with OWB 11g Release 1" 

Speaker: Antonio Romero, Sr. Product Manager, OWB, Oracle

OWB 11g options:

• Enterprise ETL
• Connectors for non-Oracle data sources
• Data Quality

Difference between business rules and data rules. Example:

• Business rules: Eligibility applies to people who were born in Massachusettes
• Data rules: SSN begins with '010-30' or state of home address is 'MA'
  

Business rules are abstract, whereas data rules are actionable in a query. Often enough, data rules are more direct and clear than at least the initial version of the business rules. (note: SSN was not part of the original evaluation criterion in the above example)

Title: "Advanced Reporting with BI Publisher"

Speaker: Kanichiro Nishida, Sr. Principle Consultant, Oracle

I have to admit that I am not familiar with BI Publisher at all. However, I think it would be nice to have the product knowledge in case we need to create nicely presented reports. So, let me capture several the high points where I see relevant.

• Supports XML and XSL as report format and template, among other - can it be used for define.xml?
• Supports optional data export as XML - can it be used as CDISC ODM output for ViewPoint FUSE?
• Customizable RTF template
• Through JSP/Java servlet, user can directly import Microsoft Excel data into BI EE without needing to import the raw data into an Oracle table.

My head is spinning toward the end of the session. My impression is it is a very complicated too since Kanichiro kept toggling between application screens. Dizzy. Also, it is 5:30pm already.

4:30pm

There was a sentiment that users wanted better metadata management.  One of the PMs said "Information Grid" exists on the roadmap as a long-term goal.

I asked why there wasn't much focus on life science, yet quite a few presentations illustrated Oracle's desire to compete with SAS (even went to SPSS as a partner).  So, the same PM said Oracle has an industry business unit (IBU) for the life science industry.

Nevertheless, Oracle consistently attacks SAS pricing structure.

Finally, we met Jean-Pierre Dijcks.  We secured a lunch meeting for tomorrow.  Hopefully, Antonio Romero will be able to join us.  We extended our invitation again for their on-site visit.  I am in transit to the "Meet the Product Managers" session.

Just a quick thought!  I wonder if Oracle Text is a good (free) answer to our lab test unit standardization challenge in DIS.  The varying spelling, almost unstructured nature of lab test seem to fit the context of what Oracle Text is to solve.  I specially like the scoring/ranking feature in the result.

Title: "Getting More Value From Unstructured Data"

Product: Oracle Text, a free option (not installed by default)

Context index types that enables text searching:

• context
• ctxcat
• ctxrule

After text index is created, these keywords can be used:

• where contains
• where catsearch
• where matches

Oracle Text supports stop words (e.g., "the", "on", "a") and they are customizable.

Example:

• > create table foobar (text varchar2(4000));
• > insert into foobar values ('the cat sat on the mat');
• > create index idx_text on foobar(text) indextype is ctxsys.context;
• > select text from foobar where text contains 'cat';

Important: Text index does not synchronize automatically when new data are inserted. However, it can be instructed to synchronize on commit, which may incur a performance hit. Different DML types has different behaviors (RTF[ine]M!).

Grant CTXAPP role to the user to be able to create indices, synonyms, etc.