I've recently came across an interesting observation I've not seen documented yet, so I'm publishing a simple example here to demonstrate the issue.
In principle it looks like that the efficiency of Bloom Filter operations are dependent on the cardinality estimates. This means that in particular cardinality under-estimates of the optimizer can make a dramatic difference how efficient a corresponding Bloom Filter operation based on such a cardinality estimate will work at runtime. Since Bloom Filters are crucial for efficient processing in particular when using Exadata or In Memory column store this can have significant impact on the performance of affected operations.
Chinar Aliyev has recently started to pick up on several of my blog posts regarding Parallel Execution and the corresponding new features introduced in Oracle 12c.
It is good to see that obviously Oracle has since then improved some of these and added new ones as well.
Here are some links to the corresponding posts:
New automatic Parallel Outer Join Null Handling in 18c
Improvements regarding automatic parallel distribution skew handling in 18c
Continuing from the previous part - which was about the Temp Table Transformation and join cardinality estimates - using the same simple table setup here is a slight variation of the previously used query to demonstrate the potential impact on single table cardinality estimates:
explain plan for
with
cte as (
select /* inline */ id from t1 t
where 1 = 1
)
select /*+
no_merge(a) no_merge(b)
*/ * from cte a, cte b
where a.id = b.id
and a.id > 990 and b.id > 990
;
-- 11.2.0.x Plan with TEMP transformation
--------------------------------------------------------------------------------
Having published recently two notes about the Temp Table Transformation highlighting the heuristics based decision and other weaknesses, for example regarding the projection of columns, it's time to publish some more notes about it.The transformation can also have significant impact on cardinality estimates, both join and single table cardinality.Looking at the difference in the join cardinality estimates of following simple example:
create table t1
as
select
rownum as id
, mod(rownum, 10) + 1 as id2
, rpad('x', 100) as filler
from
dual
connect by
level <= 1000
;
When performing aggregate GROUP BY operations an additional filter on the aggregates can be applied using the HAVING clause.Usually aggregates are one of the last steps executed before the final result set is returned to the client.However there are various reasons, why a GROUP BY operation might be somewhere in the middle of the execution plan operation, for example it might be part of a view that cannot be merged (or was hinted not to be merged using the NO_MERGE hint), or in the more recent releases (11g+) the optimizer decided to use the GROUP BY PLACEMENT transformation that deliberately can move the GROUP BY operation to a different execution step of the plan.In such cases, when the GROUP BY operation will be input to some other operation, it becomes essential for the overall efficiency of the execution plan preferred by the optimizer that the cardinality estimates are in the right ballpark, as it will influe
So you have that application that cannot be changed but makes use of some weird expressions that screw up the cardinality estimates of the optimizer.
Consider this simple example:
Oracle 11g added Extended Statistics support for column groups in order to detect correlated columns for filter predicates using an equal comparison.
Note that Oracle 11g also added the ability to use the number of distinct keys of a composite index as an upper limit for the cardinality estimates for matching column predicates, which means that the optimizer is now capable of detecting correlated columns without the explicit addition of Extended Statistics / Column Groups.
In a recent OTN thread I've been reminded of two facts about Dynamic Sampling that I already knew but had forgotten in the meantime:
1. The table level dynamic sampling hint uses a different number of blocks for sampling than the session / cursor level dynamic sampling. So even if for both for example level 5 gets used the number of sampled blocks will be different for most of the 10 levels available (obviously level 0 and 10 are exceptions)
2. The Dynamic Sampling code uses a different approach for partitioned objects if it is faced with the situation that there are more partitions than blocks to sample according to the level (and type table/cursor/session) of Dynamic Sampling
Note that all this here applies to the case where no statistics have been gathered for the table - I don't cover the case when Dynamic Sampling gets used on top of existing statistics.
If you consider the usage of Table Functions then you should be aware of some limitations to the optimizer calculations, in particular when considering a join between a Table Function and other row sources.
As outlined in one of my previous posts you can and should help the optimizer to arrive at a reasonable cardinality estimate when dealing with table functions, however doing so doesn't provide all necessary inputs to the join cardinality calculation that are useful and available from the statistics when dealing with regular tables.
Therefore even when following the recommended practice regarding the cardinality estimates it is possible to end up with some inaccuracies. This post will explain why.
Join Cardinality Basics
If you want to make use of Oracle's cunning Star Transformation feature then you need to be aware of the fact that the star transformation logic - as the name implies - assumes that you are using a proper star schema.
Here is a nice example of what can happen if you attempt to use star transformation but your model obviously doesn't really correspond to what Oracle expects:
purge table d;
drop table t;
purge table t;
create table t
as
select
rownum as id
, mod(rownum, 100) + 1 as fk1
, 1000 + mod(rownum, 10) + 1 as fk2
, 2000 + mod(rownum, 100) + 1 as fk3
, rpad('x', 100) as filler
from
dual
connect by
level <= 1000000
;
exec dbms_stats.gather_table_stats(null, 't')
create bitmap index t_fk1 on t (fk1);
Recent comments
2 years 51 weeks ago
3 years 11 weeks ago
3 years 15 weeks ago
3 years 16 weeks ago
3 years 20 weeks ago
3 years 41 weeks ago
4 years 10 weeks ago
4 years 39 weeks ago
5 years 24 weeks ago
5 years 24 weeks ago