Tim Gorman, Delphix
There is a lot of talk about big data, and it is often associated with Hadoop and other NoSQL databases. But SQL and Big Data are not mutually exclusive. There are petabyte-sized databases on Oracle Database, and this is how they happen.
This presentation is distilled from 15+ years of extensive experience implementing multi-terabyte databases to support data warehouses on the Oracle RDBMS. One of the most important technical decisions is how to implement partitioning within the Oracle RDBMS to optimize query performance for response time even as data volumes increase, and also to optimize data loading to maximize availability during data loading, eliminating downtime and minimizing impact on the system. Partitioning is also implemented to optimize storage by enabling compression and taking advantage of "tiered" storage strategies and READ ONLY tablespaces.
Partitioning is the key to unlock the full power of numerous advanced Oracle features for data warehousing in a way that is not explicitly documented by Oracle. This interlocked dependency between features is described by the presenter as the "virtuous cycle." Not starting with partitioning on a data warehouse project can lead to a deadlock of unfortunate circumstances that can doom the project to failure, which the instructor characterizes as the "death spiral." In fact, one of the most important query optimizations for data warehouses in Oracle known as "star transformations" is effectively unavailable without using properly implemented partitioning.
Become a member of ODTUG to gain access to more than 12,500 files in our technical database.
- Not a member? Click through to the topic of interest to browse a list of available presentations.
- Already a member? Log-in here to access the full database
Click here to see the Full Techincal Resource Database