Chapter 12 will focus on Oracle Parallel Query and show different methods that can be used to invoke parallelism in cases where full table scans are being performed against large tables. We begin by showing how to find large-table full table scans and then discuss the different methods that can be used for parallelizing these kinds of queries. We'll also show how to set the degree of parallelism in order to maximize the throughput and execution time of the individual query, based on the number of CPUs and the number of disks for the table's file.

We will also discuss parallel Data Manipulation Language (DML) and how operations such as parallel index rebuilds can improve the speed of database maintenance activities using SQL updates, inserts and deletes. Chapter 12 also examines how submitting concurrent requests to Oracle can parallelize many large Oracle tasks.

Tuning Real Application Clusters (RAC)

Chapter 13 will discuss Oracle Parallel Server (Oracle7 through Oracle8i) and Real Application Clusters environment (for Oracle9i).  We will also describe the characteristics of application systems that will benefit from using an Oracle Parallel Server (OPS) or Real Application Clusters (RAC) environment. We will discuss the configurations necessary for clustered hardware and also take a look at application partitioning techniques that can be used to maximize throughput in an OPS or RAC environment. This chapter will also take a look at determining the optimal number of instances for OPS and RAC, and look at the Integrated Distributed Lock Manager (IDLM) for OPS. We will take a close look at the IDLM parameters, and show how STATSPACK can be extended in order to collect information on the behavior of the IDLM in a production environment. We'll also take a close look at pinging between instances, and understand how pinging can be minimized by partitioning application structures.

Now that we have covered the overall approach to tuning, we need to look at how STATSPACK is used to facilitate the tuning effort.