On the Edit Cache Parameters screen, you can change the settings for the OLAP cache.
The cache parameters are set centrally during implementation of the BW system.
For more details, see the Implementation Guide (IMG) for
.You can set the following parameters:
Local Cache Only
If you set this parameter, the database cache is deactivated centrally, and the local cache used instead. Note that this option impairs the performance of all queries throughout the system.
Activate Statistics
You can stipulate whether or not cache statistics are recorded.
Statistics are recorded for each cache element every time a query is executed. These include:
Creating statistics involves additional database access and can therefore impair performance. It is therefore possible to deactivate creation of statistics.
The standard setting is for this to be activated. In the standard setting, the statistics are buffered initially and written to the database in their entirety at the end of a query navigation step. System performance is then considerably better than when writing each individual data record to the database.
Database Cache Limit
With the database cache limit, you can set the maximum number of entries allowed in the database cache. Set this limit high enough so that it will not be exceeded during normal operation. This limit serves to intercept outlier batches. This allows you to prevent the directory becoming too large, and performance being impaired.
We recommend setting the database cache limit very high in the production system for a given period (for a week for example). Once this period has elapsed, you can assess how high the requirement is in regular operation and can set the database cache limit accordingly.
If the database cache limit is exceeded, the system deletes data automatically. It does not delete all entries however. Instead, it deletes only until the cache contains a certain number of entries below the database cache limit again.
You have set the database cache limit to 500 entries. The system carries on deleting until the cache contains 300 entries.
The following rules apply for deletion:
Automatic Adjustment
If you set this parameter, the database cache limit is automatically optimized. The system does this by analyzing the deletion as described above at runtime and attempting to set the database limit in such a way that the deletion runs once every three days.
This setting is selected by default.
Automatic Limit
If this parameter is set, the system compares the estimated search time with the estimated calculation time needed to recalculate the queried data while searching for a suitable cache object. If the estimated calculation time is shorter, the system aborts the search and recalculates the queried data. In order to make future search queries faster, the cache objects that have not been used for the longest period are also removed from the list of valid cache objects.
This setting is selected by default.
Additional Cache Settings for Query
The Cache Settings for Query screen appears.
Here, you can make cache compression settings (Threshold, Size, Dynamic) for every query in the system.
We only advise this course of action if problems occur when executing the query with cache mode Database Cache.
The system lists all queries that have already been executed.
In order to make cache compression settings for a query before it is executed for the first time, choose Create. A dialog box appears where you can select a query.
The following table provides an overview of the various cache compression settings:
Cach Compression Setting | Description |
---|---|
Threshold | If the number of rows with data in the cache object exceeds the number set here, the size of the cache object is restricted. The setting for the cache compression size is then used. The compression threshold should generally not be higher than 5000. This helps to avoid impairment of performance that would be caused by memory-sapping calculations of the size of most cache objects. As long as the query performance does not result in a dump, you should set this number as high as possible. |
Size | If the size of the cache object (in bytes) exceeds the number set here, the system compresses block by block. The compression size should be around 2 GB, though enough memory space should be left for the object's metadata. A lower compression size results in a larger number of blocks and longer runtimes. |
Dynamic | The cache object can be highly heterogenous. It is therefore not always possible to split it into identical blocks of a fixed size. Only use the dynamic method in cases where there are no runtime dumps, and set the compression size to under 100 MB. The dynamic method results in longer runtimes. |
More information: OLAP Cache Monitor.