Currently browsing category


“alter session force parallel query”, and indexes

This post is a brief discussion about the advantages of activating parallelism by altering the session environment instead of using the alternative ways (hints, DDL). The latter ways are the most popular in my experience, but I have noticed that their popularity is actually due, quite frequently, more to imperfect understanding rather than informed decision – and that’s a pity since “alter session force parallel query” can really save everyone a lot of tedious work and improve maintainability a great deal.…

Xplan: now with “self” measures for row source operations

One of the most useful information that the Oracle kernel attaches to plans in the library cache are measures of various resource consumption figures, such as elapsed time, consistent and current gets, disk reads, etcetera. These can be made available for each plan line (aka “row source operation”).

These figures are always cumulative, that is, include both the resource consumed by the line itself and all of its progeny.…

CBO: NewDensity for Frequency Histograms,11g- (densities part IV)

As we have seen in the previous posts of this series, in 11g a new figure named “NewDensity” has been introduced as a replacement for the “density” column statistic for columns whose histogram has been collected; this change has been backported in also.

In the previous post we discussed how NewDensity influences the CBO cardinality estimate for Height-Balanced histograms; in this one we are going to investigate the same for Frequency Histograms.…

CBO: “NewDensity” replaces “density” in 11g, (densities part III)

In this post we are going to explore and explain the rationale for the formula used by the CBO to compute the “NewDensity” figure that replaces, from onwards, the “density” column statistic in the cardinality estimation formulae for columns with height-balanced (HB) histograms defined.

In a previous post, we already discussed the pre- scenario: we saw how and when the “density” column statistic is used in the cardinality formula for equality filter predicates, we explained its statistical rationale and defining formula, introduced the concept of the NPS (Not Popular Subtable), and built a test case.…

CBO: the formula for the “density” column statistic (densities part II)

In this post we are going to explore and explain the rationale for the formula used by dbms_stats to compute the “density” column statistic, used by the CBO in versions less than to estimate the cardinality of a class of SQL statements. In the next post, we will speak about its replacement, named “NewDensity” in 10053 trace files.

We will consider only the non-trivial case of Height-Balanced histograms, since for Frequency Histograms density is a constant (0.5 / num_rows) and for columns without histogram, it is simply 1/num_distinct.…

CBO: about the statistical definition of “cardinality” (densities part I)

Let’s explore the concept of cardinality from the point of view of the statistician; this is both to get a clearer vision of the matter (i.e. for fun) and to path the way for understanding the rationale for the “density” statistics as calculated by dbms_stats (the topic of an upcoming post).

Let’s consider a statement with input parameters (bind variables), and consider the most fundamental of them all, the one with a filter predicate:

select ...

CBO: the “non-empty result set” assumption

The CBO assumes that SELECT statements are always going to retrieve at least one row – even if this is not necessarily the case, of course. Understanding why this is done is both useful and fascinating.

We must start from the very beginning and remember that one of the most important tasks of the CBO is estimating the statement cardinality, that is, to make a guess about the number of rows that will be fetched.…