How FTS indexing works
BMC Remedy AR System uses the AR System Server Java process to insert, update, or delete data in the FTS indexes. Threads from the Full Text Indexing queue perform the indexing. This queue has one thread by default, which is sufficient for very large indexing environments. You can configure more indexing threads for FTS. However, you must exercise care. For more information, see the "Defining queues and configuring threads" section in Setting ports and RPC numbers.
The indexing mechanism is based on an asynchronous queue, which is based in an internal system table (
ft_pending) in the database during the originating transaction. The originating transaction typically involves a create entry, set entry, merge entry, or delete entry operation on a form where a field indexed for full text search exists.
A full text dispatcher thread processes the indexing tasks in real time on a first-in, first-out basis, queuing them for indexing threads to process. As a result of this indexing model, the performance of the originating transaction is affected only marginally by inserting the indexing task record into the system table, and is not subject to delays associated with full text indexing. However, the data might not be immediately available for searching. The size of the delay depends on the size of the indexing queue and the availability of system resources to perform the indexing.
The Full Text Search (FTS) has smart processing for all the entries to be indexed. The smart process optimizes the processing of the redundant entries and results better indexing performance. It provides the following functionality:
- Avoids processing of duplicate indexing requests from JMS and ft_pending.
- Does not depend on the JMS messaging based communication for the FTS indexing. It uses only JMS as a wakeup signal and also adds a scheduler for any periodic missed entries.
- Avoids logically redundant request by cleaning the requests from ft_pending during the insertion or processing. System ensure that only the redundant entries are deleted from the system so that work required for these entries is completed by some other entry.
- Optimizes the synchronization of global or form and entry level re-indexing in such a way that the resources in the server (CPU and threads) are optimally used for indexing.
- BMC recommends that you perform global re-index or schema re-index during off-peak hours, such as during a maintenance window.
- Re-indexing of entire schema or form deletes the existing indexes and creates the indexes again. So the re-indexing of a form may be faster as compared to individually updating the index for more number of entries.
If the administrator has disabled indexing, indexing tasks are still recorded, preserving the changes for later inclusion when indexing is enabled.
This section contains information about: