Databases Reference
In-Depth Information
<Target>object_reference</Target>
</MergePartitions >
Analysis Services merges partition in several steps. In the first stage, Analysis Services
copies the data segments from all partitions into the Target partition. Analysis Services
copies data from partial segments (segments that have fewer than 64K records) into a
separate buffer and creates new segments based on the data in this buffer. In this way,
Analysis Services avoids segmentation and fragmentation of data in partitions. Still, the
order of data differs from the order of records you would get if you processed the partition
using the ProcessFull option, because the data is not clustered together.
In the next stage, Analysis Services decides whether to merge partition aggregations. If
partitions have the same aggregation design and aggregation exist in every Source parti-
tion, Analysis Services starts merging aggregation data segments. If a Source partition is
missing some aggregations, the Target partition won't have those aggregations either.
After it completes merging aggregations, Analysis Services tries to merge partition indexes.
Only if all source partitions have indexes does Analysis Services merge them. Otherwise, it
rebuilds all the indexes for the Target partition. The same logic applies for the aggregation
indexes—only if they exist in all Source partitions does Analysis Services merge aggrega-
tion indexes.
NOTE
Analysis Services keeps data in the segments sorted according to the DISTINCT_ COUNT
measure. As a result, during a merge, Analysis Services merges records from all parti-
tions for the measure group containing DISTINCT_COUNT measure according to the sort
order. This causes rebuilding of all data segments of the new partition; therefore, the
merge operation rebuilds aggregations and indexes of such partitions.
After the Merge command finishes, your partition might not have all aggregations and
indexes built, so you can send the Process command with the ProcessIndex option to
build missing aggregations and indexes. If you have enabled lazy processing, Analysis
Services builds missing indexes and aggregations for you in the background.
Progress Reporting and Error Configuration
While processing objects, Analysis Services produces progress reports. It generates the
progress reports as the trace events that enable you to see the stage of processing
currently executing. You can use the SQL Server Profiler application to subscribe to the
 
Search WWH ::




Custom Search