SAS Safety Scalable Performance Data Server 4.5 Manual Do Utilizador

•

The minimum number of segments that any value is in.

•

The maximum number of segments that any value is in.

•

The average number of segments that all values are in.

•

The average percentage of segments that all values are in.

The distribution statistics can be used to determine the effectiveness of the index. The index
will perform better if the distribution of

the index values is clustered in a minimum number of segments. In general, the lower the
average percentage of segments that all values are in, the better the index will perform.

Use the 

 option to reorganize the index to consolidate segment lists and retrieve

unused disk space:

> ixutil -reorg x -dsn test -libpath /spds
SAS Scalable Performance Data Server 
   4.5(TS M0) Build(Feb 26 2009, 11:50:08)
Index File Utility
Copyright (c) 1996-2009 by SAS Institute Inc, Cary NC 27513 USA

Reorg for Index x:
Reorg successfully completed
Ixutil completed successfully

Running the index utility program again to get the statistics shows that the segment lists
for all of the values have been aggregated (the avg_chunks_per_list is 1.0) and the unused
disk space has been freed (the idx_garbage_bytes is 0), resulting in a proportional decrease
in the size of the index file.

Aggregating the segment lists and compacting the index file minimizes the reads on the
index for a query. It will also increase the locality of segment data for an index key. The
combination of these will give the best query performance for the index.

Use the 

 option once more to review the index and segment list data, in order to

view the improved performance statistics.

> ixutil -stats x -dsn test -libpath /spds
SAS Scalable Performance Data Server 
   4.5(TS M0) Build(Feb 26 2009, 11:50:08)
Index File Utility
Copyright (c) 1996-2009 by SAS Institute Inc, Cary NC 27513 USA

Statistics for Index X:
-----------------------------------------
+--segment_size        = 8192
+--n_segments_in_tbl   = 5
+--n_values_in_index   = 2
+--n_vdeleted_values   = 0
+--percent_vdeleted    = 0.00
+--n_seglist_values    = 2