Click here to view and discuss this page in DocCommentXchange. In the future, you will be sent there automatically.

SQL Anywhere 11.0.0 » SQL Anywhere Server - SQL Usage » Query Optimizer » Query optimization and execution


How the optimizer works

The role of the optimizer is to devise an efficient way to execute SQL statements. To do this, the optimizer must determine an execution plan for a query. This includes decisions about the access order for tables referenced in the query, the join operators and access methods used for each table, and whether materialized views that are not referenced in the query can be used to compute parts of the query. The optimizer attempts to pick the best plan for executing the query during the join enumeration phase, when possible access plans for a query are generated and costed. The best access plan is the one that the optimizer estimates will return the desired result set in the shortest period of time, with the least cost. The optimizer determines the cost of each enumerated strategy by estimating the number of disk reads and writes required.

In Interactive SQL, you can view the best access plan used to execute a query by clicking the Plan tab in the Results pane. To change the degree of detail that is displayed, change the setting on the Plan tab of the Options window (available from the Tools menu). See Reading graphical plans, and Reading execution plans.

Cost based

The optimizer uses a generic disk access cost model to differentiate the relative performance differences between random and sequential retrieval on the database file. It is possible to calibrate a database for a particular hardware configuration using an ALTER DATABASE statement. See ALTER DATABASE statement.

By default, query processing is optimized towards returning the complete result set. You can change the default behavior using the optimization_goal option, to minimize the cost of returning the first row quickly. Note that when the option is set to First-row, the optimizer favors an access plan that is intended to reduce the time to fetch the first row of the query's result, likely at the expense of total retrieval time. See optimization_goal option [database].

Syntax independent

Most commands can be expressed in many different ways using the SQL language. These expressions are semantically equivalent in that they accomplish the same task, but may differ substantially in syntax. With few exceptions, the optimizer devises a suitable access plan based only on the semantics of each statement.

Syntactic differences, although they may appear to be substantial, usually have no effect. For example, differences in the order of predicates, tables, and attributes in the query syntax have no effect on the choice of access plan. Neither is the optimizer affected by whether or not a query contains a non-materialized view.

A good plan, not necessarily the best plan

Ideally, the optimizer would identify the most efficient access plan possible, but this goal is often impractical. Given a complicated query, a great number of possibilities may exist.

However efficient the optimizer, analyzing each option takes time and resources. The optimizer compares the cost of further optimization with the cost of executing the best plan it has found so far. If a plan has been devised that has a relatively low cost, the optimizer stops and allows execution of that plan to proceed. Further optimization might consume more resources than would execution of an access plan already found. You can control the amount of effort made by the optimizer by setting a high value for the optimization_level option. See optimization_level option [database].

In the case of expensive and complicated queries, or when the optimization level is set high, the optimizer works longer. In the case of very expensive queries, it may run long enough to cause a discernible delay.

Optimizer estimates and column statistics
Automatic performance tuning
Underlying assumptions
Predicate analysis
Improving performance with materialized views
Optimization for MIN and MAX functions
Plan caching