[ https://issues.jboss.org/browse/TEIIDDES-1618?page=com.atlassian.jira.plu... ] Steven Hawkins commented on TEIIDDES-1618: ------------------------------------------ There are multiple considerations here. The first is that if its true that most (if not all queries) hit a single partition, then there could be an option to pull representative statistics for just a single partition (which would require a change to the stats gathering). If that's still not representative, and the source partitions can be directly referenced and list partitioning is used then the source table could be represented as a partitioned view with each partition having it's individual statistics (which may require a change to the stats gathering) - this could be a designer assisted option. And or, with additional metadata, such as specifying the partitioning value lists via a metadata extension (which would be needed in the relational metamodel) and provide the individual stats via an extension as well, then Teiid could internally handle the more appropriate costing based upon the metadata. If range or other partitioning is in use then that would require additional work for Teiid to recognize additional partitioning schemes. So more input would be helpful from the owner as to the details of the partitioning scheme, etc. to determine the appropriate actions. -- This message is automatically generated by JIRA. If you think it was sent incorrectly, please contact your JIRA administrators For more information on JIRA, see: http://www.atlassian.com/software/jira