Hotsos Symposium Speaker – Maria Colgan

Maria Colgan is a returning Hotsos Symposium presenter.

Biography

Maria ColganMaria Colgan is a Senior Principal Product Manager at Oracle Corporation and has been with the company since version 7.3 was released in 1996. Maria's core responsibilities are Data Warehouse Best Practices and the Oracle optimizer. Based on Maria's extensive experience in Oracle's Server Technology Performance Group - where she conducted competitive customer benchmarks and tuning sessions - Maria creates and lectures on Data Warehouse best practices and dealing with the Optimizer.

Presentation Title

Session 1: Inside the 11g Optimizer: Removing the Mystery

Session 2: Tips to Prevent Suboptimal Execution Plans

Session 3 (co-presented with Bryn Llewellyn): Implement Extensible In-Database Analytics by Programming MapReduce in SQL and PL/SQL

Abstract

Session 1: Inside the 11g Optimizer: Removing the Mystery — Since its introduction in Oracle 7, people have been fascinated and intimidated by the Cost-Based Optimizer and the statistics it relies on. It has long been felt that the internals of the Optimizer were shrouded in mystery and that a degree in wizardry is required to be able to make it do the right thing. This presentation will explain the fundamentals of the cost based Optimizer and how it has been extended in Oracle Database 11g to address known limitations. For each of the new feature discussed, a detailed explanation will be given on what limitation it will address that and how you should take advantage of it in your Oracle environment. The session will also include a demonstration of some of the new features.

Session 2: Tips to Prevent Suboptimal Execution Plans — You have gathered the perfect set of statistics, but you still get a suboptimal execution plan. Why? In this presentation, we will discuss the other factors that influence the Optimizer's decisions and can prevent you from getting an optimal execution plan. This session will answer questions like, why an index wasn't used, why partition eliminations did not occur, why statistics were ignored and many more. With clear how-to examples, you will learn to identify and quickly resolve these issues without the necessity of Optimizer hints or initialization parameters.

Session 3 (co-presented with Bryn Llewellyn): Implement Extensible In-Database Analytics by Programming MapReduce in SQL and PL/SQL — The MapReduce programming model lets developers without experience with parallel and distributed systems utilize the resources of a large, multi-CPU system. Hadoop clusters can be used to implement this model; but Oracle Database also provides mechanisms to support the same model — and with less programming. SQL's analytic capabilities are more powerful than many experienced programmers suspect; but detecting complex patterns within suitably characterized, independent subsets of the complete data set sometimes needs user-written procedural code. Enter in-database MapReduce: pipelined PL/SQL table functions declaratively specify, with a few keywords, how parallel query should do the "map"; then the procedural body does the "reduce".

Bryn and Maria address ways to implement an application to solve a specific problem where, because of the typically huge volumes of data that must be distilled down to provide interesting information, performance matters. A person whose job it is to work with an installed system, won't be able to use what he explains in his talk to tune its performance.

Presentation Materials

Presentation materials are available online to attendees only.

Schedule

The speaker schedule is as follows:

Free Hit Counter