db-derby-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Bryan Pendleton (JIRA)" <j...@apache.org>
Subject [jira] Commented: (DERBY-1260) Investigate impact of infinite cost estimates on arithmetic operations/comparisons in the optimizer.
Date Sun, 26 Nov 2006 16:46:21 GMT
    [ http://issues.apache.org/jira/browse/DERBY-1260?page=comments#action_12453442 ] 
Bryan Pendleton commented on DERBY-1260:

Another example of the problems of handling infinite cost estimates arises during
pushing and pulling of intermediate cost estimates as part of pushing and pulling
predicates. Specifically, there is the following code in OptimizerImpl:

                double newCost = currentCost.getEstimatedCost();
                double pullCost = 0.0;
                CostEstimate pullCostEstimate =
                if (pullCostEstimate != null)
                    pullCost = pullCostEstimate.getEstimatedCost();

                    newCost -= pullCost;

                    ** It's possible for newCost to go negative here due to
                    ** loss of precision.
                    if (newCost < 0.0)
                        newCost = 0.0;

Here are at least some of the interesting situations that can arise with this code:
 - If newCost is infinite, and pullCost is also infinite, then newCost will become NaN.
 - If newCost is infinite, but pullCost is finite, then newCost will remain infinite.

For more discussion about some of the possible behaviors of this
code, see 

> Investigate impact of infinite cost estimates on arithmetic operations/comparisons in
the optimizer.
> ----------------------------------------------------------------------------------------------------
>                 Key: DERBY-1260
>                 URL: http://issues.apache.org/jira/browse/DERBY-1260
>             Project: Derby
>          Issue Type: Task
>          Components: Performance
>    Affects Versions:
>            Reporter: A B
>            Priority: Minor
> For large, deeply nested queries and/or queries with a high number of FROM tables/expressions
with high row count estimates, the the resultant cost estimates can be multiplied--sometimes
many times over--throughout the optimization process, which means that the overall query estimate
can climb to a very large number very quickly.  If the query is big enough, this can actually
cause the optimizer to reach an estimated cost of INFINITY.
> That said, the current optimizer logic for costing and comparing access paths does not
expect--and therefore does not account for--infinite cost estimates.  As a result the optimizer
does comparisons of, and basic arithmetic with, cost estimates and row counts that, when applied
to estimtes of Infinity, can give unexpected results.  One specific example of this is DERBY-1259.
> I'm filing this issue for the task of investigating the optimizer code more closely to
see what other issues might exist when infinite cost estimates come into play.

This message is automatically generated by JIRA.
If you think it was sent incorrectly contact one of the administrators: http://issues.apache.org/jira/secure/Administrators.jspa
For more information on JIRA, see: http://www.atlassian.com/software/jira


View raw message