Describe the Oracle Optimizer Functions

The query optimizer is the key to one of the great advantages of all relational databases. Before we jump into the features of the Oracle optimizer, we should review the important reasons why a query optimizer is necessary, as well as the basics of Oracle's own query optimizer.

• Purpose of Optimization
  One of the foundations of a relational database is that users will not have to specify an access path to retrieve data.
  Question: What if there are multiple ways to access the data needed to satisfy a particular query?
  Virtually all relational databases include some form of query optimizer. As the name implies, the query optimizer is a piece of software that chooses the optimal path to access the data needed for a query. Choosing the right query access path can dramatically affect the performance of a query. For instance, if the data requested in a query can be selected and retrieved by using an index, it is usually much faster to use the index to access the data than to go directly to the database row. But what if multiple indexes could be used? Are there times when using an index will not be the fastest way to access the data? The job of the query optimizer is to provide the best path for any query.

Oracle has had a query optimizer for a long time. In the early days, Oracle offered a rule-based optimizer. As the name implies, this optimizer was based on a fixed set of rules. These rules tried to use the syntax of the Structured Query Language (SQL) query to determine the optimal access path, referred to as an execution plan ^[1]. For instance, if a query contained a reference to an index, the query optimizer would usually use that index to create the execution plan. There were two disadvantages of the rule-based optimizer:

1. It had somewhat limited complexity, because there were not many rules.
2. The rules, by themselves, did not always lead to the right decision.

To correct these deficiencies, Oracle7 introduced a cost-based optimizer. As the name implies, the cost-based optimizer includes calculations that estimate the cost of using a particular access path in the execution plan, which in turn results in a more accurate plan. Oracle has endorsed the cost-based optimizer, which means that new features, such as partitioning, are no longer considered by the rule-based optimizer. By default, Oracle uses the cost-based optimizer. The OPTIMZER_MODE setting tells Oracle which type of optimizer to use from among the following:

1. RULE for the rule-based optimizer
2. COST for the cost-based optimizer
3. CHOOSE so that Oracle can choose which optimizer to use, based on the presence or absence of statistical information or stored outlines, both of which are described later in this course

The entire basis of a cost-based optimizer is the simple fact that different access paths can require dramatically different (I/O) input/output costs. The nomenclature BIGTABLE and LITTLETABLE is specific to this example. To understand this better, assume that you have a simple query that joins two tables together

1. BIGTABLE with 1,000 rows and
2. LITTLETABLE with 10 rows.

For the purpose of this illustration, we can assume that there is only one row in BIGTABLE that can be joined to each row in LITTLETABLE. If you access BIGTABLE first, the retrieval will first read a row in BIGTABLE and then read the associated row in LITTLETABLE. To clarify, that is one access for the BIGTABLE row and one access to retrieve the value of the join column in BIGTABLE with the value of the same column in LITTLE TABLE. Oracle will perform this action 1,000 times, resulting in 2,000 accesses overall.
If you access LITTLETABLE first, the retrieval will read a row from LITTLETABLE and then the associated row in BIGTABLE, for a total of 20 accesses overall, or 100 times fewer rows than starting the query with BIGTABLE. Although the caching mechanism in Oracle will help diminish the effect of this dramatic difference, it is easy to see that the cost of starting the query access with BIGTABLE is much greater than the cost of using LITTLETABLE first.

To a rule-based optimizer, an index is an index is an index. The Oracle rule-based optimizer will choose an index based on a somewhat arbitrary rule, such as which index is mentioned first. The inflexible nature of rule-based optimization means that the way a query is written could result in significantly different performance characteristics. Or, even worse, that the same query could deliver significantly different performance characteristics as the actual composition of the database changes.

When you upgrade to Oracle Database 11g, optimizer statistics are collected for dictionary tables that lack statistics. This statistics collection can be time-consuming for databases with a large number of dictionary tables, but statistics gathering only occurs for those tables that lack statistics or are significantly changed during the upgrade. To decrease the amount of downtime incurred when collecting statistics, you can collect statistics prior to performing the actual database upgrade. As of Oracle Database 10g, Oracle recommends that you use the DBMS_STATS.GATHER_DICTIONARY_STATS procedure to gather these statistics.
In the next lesson, you will learn about the specific enhancements to Oracle8's query optimizer.

[1] Execution Plan: A description of the steps the Oracle database will take to retrieve and select the data requested by a query.