Use of Database Parameters

Tamino has about 30 database parameters, some of which influence overall performance. In the following table, you will find a short description of the parameters that influence performance and recommendations about how to set them. For a general description of database parameters, see Tamino Manager, Creating a Database.

Name Description/Performance
Server Properties
Buffer pool size This is the size of the buffer pool of the Tamino Server. The higher this value is set, the shorter the I/O waiting times. The maximum should not exceed the size of physical memory in order to avoid swapping (keep other memory consuming resources in mind!). If Tamino starts receiving regular page faults, the value is probably too high compared to the size of the physical memory.

Tip:
In general, enlarging the buffer pool will lead to less I/Os and thus to better performance, as long as the system is not swapping and as long as the buffer pool is fully used (this can be checked using the Database Activity display of System Management Hub). A good practice is increasing the buffer pool by a constant value (for example 200 MB) and to simultaneously check the buffer pool hit rate within the Database Activity display. As long as the hit rate increases, performance improvements are to be expected. As soon as the buffer pool hit rate does not increase any longer when increasing the buffer pool size the optimum setting has been surpassed. Obviously the buffer pool can't be exceeded beyond the memory available.

Dynamic pool size This is the size of the internal dynamic working pool of the Tamino server. It is used to keep result sets and for sorting. If this parameter is set too low, information will be written to temporary working space which leads to disk I/O.
Request log file maximum size XML request logging may degrade performance to a small extent. Hence, performance data gained from XML request logging should be handled with care.
XML Properties
XML work threads (XWT) This is the number of threads used to process XML commands, which can be processed in parallel. If no XWT is available (too many parallel requests), the request will be kept in a queue. The thread is busy until the result is sent or an error occurs. If the value is too high, the threads may impede each other. On the other hand, if the value is too low, applications doing updates may become too slow due to locking.
Word fragment index

This specifies whether Tamino should create a word fragment index when the search type is set for text indexing. This accelerates the resolution of queries involving text strings such as name~="*string*", which would find name nodes whose values contain "string". Specifying "yes" for the word fragment index thus enhances the performance of queries, but negatively impacts performance and space consumption when loading and indexing data.

The index can be switched on and off with the Tamino Manager. The change is effective from the next start of the Tamino server. Note that switching on this parameter for a database that already contains data will result in very long startup times for the first server start after the enabling of the index. If you switch the word fragment index on or off, we recommend that you shut down and start up the database and then take a backup before you make any content changes to the database.

XML maximum request duration

The maximum duration of each request is limited to the specified number of seconds. If a request takes longer than the limit that is specified with this parameter, the request will be aborted, and an error message is returned.

XQuery document cache size This parameter defines the capacity of the document cache that is used for XQuery processing for each request. Document caching improves the performance of certain queries like join queries or sort-by queries where the sorting is not done via an index. Also certain XQuery update statements can benefit from a big document cache. The capacity of the XQuery document cache is specified in MB. The default capacity is 20 MB. For applications with a high parallel query load it may be necessary to reduce the parameter.