What's New in Oracle 12c
Posted: Mon Jan 27, 2025 9:04 am
What changes in this context with Oracle 12c? Oracle 12.1 introduces a lot of new features in the CBO area, which interestingly are partially being deactivated again with the new version 12.2, since some of the new features obviously did not have as positive an effect in real operation as Oracle had hoped.
With regard to the examples mentioned above, the new "Adaptive Joins" feature is particularly noteworthy, as it allows Oracle to adapt the join method at runtime when the bosnia and herzegovina telegram screening quantity estimates are not certain. This means that Oracle can decide between Nested Loop Join ("small jobs" strategy) and Hash Join ("one large job" strategy) at runtime, depending on how many rows are actually created in operations before the join. This feature proves to be very useful in some cases, as we will see shortly, but it cannot solve some problems, for example if an inefficient join order was chosen due to incorrect quantity estimates, which results in more data being created and processed than necessary. The join order cannot therefore be adapted adaptively, only the join method (Nested Loop or Hash Join).
In addition, Oracle 12c can also remember when such miscalculations of the quantities occur and persist corresponding "directives" that cause Oracle to maintain extended statistics and spend more time on so-called "dynamic statistics" at plan creation time. This means that Oracle executes part of the query while the execution plan is being created in order to determine which data volumes are actually generated. Since only part of the data is read here ("sampling"), this information can also be misleading - but in many cases it is actually helpful. Since these additional activities ("directives" and "dynamic statistics") can have a significant impact on the time that Oracle needs to create execution plans, and many customers have observed a sometimes extreme increase in runtimes and instability due to these additional activities, this part is deactivated again by default in 12.2 (and Oracle recommends doing the same for Oracle 12.1).
With regard to the examples mentioned above, the new "Adaptive Joins" feature is particularly noteworthy, as it allows Oracle to adapt the join method at runtime when the bosnia and herzegovina telegram screening quantity estimates are not certain. This means that Oracle can decide between Nested Loop Join ("small jobs" strategy) and Hash Join ("one large job" strategy) at runtime, depending on how many rows are actually created in operations before the join. This feature proves to be very useful in some cases, as we will see shortly, but it cannot solve some problems, for example if an inefficient join order was chosen due to incorrect quantity estimates, which results in more data being created and processed than necessary. The join order cannot therefore be adapted adaptively, only the join method (Nested Loop or Hash Join).
In addition, Oracle 12c can also remember when such miscalculations of the quantities occur and persist corresponding "directives" that cause Oracle to maintain extended statistics and spend more time on so-called "dynamic statistics" at plan creation time. This means that Oracle executes part of the query while the execution plan is being created in order to determine which data volumes are actually generated. Since only part of the data is read here ("sampling"), this information can also be misleading - but in many cases it is actually helpful. Since these additional activities ("directives" and "dynamic statistics") can have a significant impact on the time that Oracle needs to create execution plans, and many customers have observed a sometimes extreme increase in runtimes and instability due to these additional activities, this part is deactivated again by default in 12.2 (and Oracle recommends doing the same for Oracle 12.1).