Hotsos Symposium Speaker – Gary Propeck

Gary Propeck is a returning Hotsos Symposium presenter.

Biography

Gary PropeckGary has a Ph.D. in Physical Organic Chemistry from the University of South Florida in 1984. In 1990, while working at a nuclear weapons facility, he began working with the Oracle Database in a chemistry laboratory environment. In 1995, Gary went to work for Oracle Education (now Oracle University) as an instructor. Gary taught the OU curricula for Database Administration, Tools (Forms, Reports, etc.), Oracle Application Server, and Languages (SQL, PL/SQL, and Java). During that time, Gary made numerous presentations at events such as Open World and RMOUG Training Days. In 2002, Gary went to work for Corporate Express as a DBA, gaining more experience working with the Oracle Database. In July 2004, Gary decided to go back to teaching as an independent contractor, teaching much of the same curricula as before, while adding new courses such as OEM and Apex to his repertoire. 2009 brought Gary to Hotsos, as he began teaching as a contractor. In August 2010, Gary started teaching as a Hotsos employee where he teaches the Performance Tuning curriculum. Gary presented the paper "The CBO's Look at Clusters, IOTs, Partitions, and other Objects" at the 2011 Hotsos Symposium.

Presentation Titles

The Oracle Cost-Based Optimizer and Partitioning

Abstracts

The default Oracle storage structure for persisting data is the heap table. As one's data grows, heap tables become very inefficient for queries to return data. Indexes can be used on predicate columns to improve performance. Yet, the indexes may also grow very large. Partitions improve performance by segregating the data into smaller components. Partitioning our data has several benefits: Partition Pruning, Partition Maintenance and Independence, and Better Parallel Execution among them. In addition, to convert an existing heap table into a partitioned object does not require any modification to the application code. This presentation demonstrates the statistical evidence that partitions improve performance of queries through the use of the 10053 trace data. Range partitions are contrasted to the same data stored in heap tables. The presentation will show that the CBO treats the partitions as though they were smaller individual tables in the calculation of the costs. The presentation details the cost calculations from the traces to give us an understanding of why the optimizer picks the row source operations that it does, basically the partition scan or the full table scan. The presentation also discusses the manner in which the CBO treats index partitions.

Presentation Materials

Presentation materials are available online to attendees only.

Schedule

The speaker schedule is as follows: