How to check the partitioning of an index?
There may be doubts what partitioning the system is using at a given point in time. This can be checked in the TREXAdmin Standalone Tool.
If you have no easy access to this tool, you may also use transaction TREXADMIN..
Type of partitioning
Go to “Index” -> “Landscape” and select the InfoProviders indexes.
In this example we see the fact index of InfoProvider FAB_BIG, split in two parts, and also the second dimension index of this Provider, also split in 2 parts. Do a right mouse click on the logical index of a split index (not a physical/part index). Select “Split/Merge Index” from the context menu. In the popup window you see if and how the index is partitioned.
In this screen-shot we see that indeed the fact-index of Provider FAB_BIG (q99_bic:ffab_big)is partitioned by its second dimension. “key_fab_big2” is the name if this dimension-key in the attributes of the fact index. The "HASH 2" shows that there are two parts.
If you find the entry “ROUNDROBIN”, then the fact index is split, but not partitioned by any
attribute. This means that records are randomly distributed over the parts (just ensuring an even load balancing).
This screen-shot shows another fact-index (Provider FAB_BIG2). It is split but not partitioned by any attribute.
Value distribution in the parts
You can directly view the distribution of attribute values of a fact-index and the dimension index that is used for its partitioning.
In the tab “Index” -> “Mining” you can select a part-index of a partitioned fact-index (not the logical index). You get a list of all attributes of this index. Mark the line of the attribute that was used for partitioning. On the marked line, right-click on the attributes name (=on the second column!), chose “Show all values”. A bar-chart gets displayed, right-click on it and chose “Top”. Now the chart shows the first values of this attribute in this part of the index.
Example fact index: The values of the attribute of dimension 2 of the Provider FAB_BIG; starting with the smallest.
Now choose the dimension index that is used for the partitioning. Select the same part as previously for the fact-index. (The part on the same blade!) Choose the relevant attribute: “dimid” in a dimension index and do the same steps: right-click on the attribute name, chose "Show all values", etc…
Now the displayed list of values should be the same. All values from this part of the dimension index can find their relevant entries in the fact index on the same blade. This means the fact index is partitioned by this dimension.
Example dimension index: The dimid (=Dimension ID) values are the same as in the fact index on the same blade.