Tuning the indexer

On high performance hardware with a well-tuned database, Teamcenter can index 50,000 objects to 150,000 objects per hour. Actual performance on a production system depends on the rate of change of structure and geometry, for example, occurrences, occurrence transforms, item revisions, bounding boxes (JT and CAD), and absolute occurrence transform overrides. You can use Oracle auditing and SQL scripts to measure the hourly or daily changes based on the last modified date of each object. From those measurements, you can estimate the number of changed items, occurrence threads, and APNs per hour you must index (the indexer queue only contains those objects). You can also identify the optimum mixture of queued objects for the best performance.

You should then measure the actual maximum indexing throughput, which comprises two considerations:

• Throughput of operations that do not alter indexes, which you can identify by running the qsearch_process_queue utility with the –queue_update option.
• Throughput of requests that cause index changes, which you can identify with the CAD tool or a custom script or utility.

Tip: Avoid accumulating a large number of entries in the indexing queue. The faster updates are processed, the sooner live updates are available to users. During the deployment process, you should monitor the size of the queue by periodically running the qsearch_process_queue utility with the -list_queue option selected. If the queue size is increasing steadily, you may have to reconsider your deployment strategy. In a test environment on high performance hardware, Teamcenter can process between 60,000 and 120,000 requests per hour without the queue building up.

If you have a relatively small product structure (a thousand or less occurrence threads), you may improve performance by switching to a top-down search. This configuration limits the scope of the search to the product only. However, Teamcenter must first traverse the unconfigured structure in the database to establish the search scope. This approach may be more time-consuming with large product structures; if so, it may be better to tune the default bottom-up search for best performance in such cases.

By default, Teamcenter does not include all possible indexes that you might need for attribute queries. Additional database indexes have an associated performance cost, and you should consider carefully whether you need them. Use the Oracle database tools in a live production environment to assist with these decisions.