Below are my responses to questions sent from a Windchill consultant who is with a PTC partner.
Q1: I understand that it is OK to increase number of DB connection above 15. What is a drawback? How should I tune database for more connections?
Setting the number of DBConnections above 15 is ok, particularly the more memory (2g+) that is allocated to the MethodServer. In general though while there might be spikes in the mid twenty range occasionally, most method server will likely start to struggle with many more than 15 concurrent connections open to Oracle for lengthy amount of time. A caveat to the 15 DBConnections guideline is for the background method server, particularly on systems with 'a lot' of active queues more than 15 DBConnections is usually required 30-50 is a typical configuration. Additionally in Windchill PDMLink R10 there were some new properties which were introduced to allow for a more dynamic expansion of connections. See the<Windchill>/codebase/properties.html for the explanations of the wt.pom.dbConnectionsHardLimit property.
Oracle needs to be configured so that the 'processes' parameter can handle the expected maximum number of Method and Background Method servers multiplied by maximum number of DBConnections. Usually the number is 600-800 range.
Q2: I am wondering what can be considered normal Windchill performance and what is a limit that I can tune the system.
How fast is fast enough? We setup the PSM to monitor for transactions which take more than 5 seconds, many/most operations in a Windchill system will complete in under this time however there are a number of operations often those which operate on a large volume of data or in a complex way which can take much longer and that’s ok, because the user understands that the operation will be lengthy because of what’s involved. There is a lot that goes into performance expectations, and lot of what’s acceptable depends on the circumstances, but monitoring for operations that take 60-90+ seconds is what I would try and do.
Q3: Customers have large servers and even with a relatively empty Windchill (just couple of thousands of CAD models) the speed of page rendering is very far from instantaneous. For example recently I had a case <case number> where Workspace scrolling refresh speed was around 3-4 seconds. And it was considered OK by support person. But if I am running Dynatrace it is showing all pages in red saying that speed above 0.2 sec is bad. And 3-4 seconds equals bad user experience.
Can you please suggest anything to speedup overall UI?
Servers should be sized in accordance with the hardware sizing guidelines. The R10 Microsoft server sizing Guidelines can be found here.
Scrolling performance is a function of the client machine and will largely independent of the server performance. For Workspace scrolling I would look at the following:
- Add registry keys per TPI 133412
- Check that IE preference for smooth scrolling is enabled (tools > internet options > advanced > Use smooth scrolling)
- Switch to a featured object list if in a workspace
- Remove Productview thumbnail from table display
- Minimize the number of columns in the table display
- Upgrade to the latest IE browser as performance is better with later releases (IE6 < IE7 < IE8 < IE9 x64 < IE9 x86)
- Use a browser other than IE as performance is better with different browsers (IE < FF < IE w/Chrome Frame or Chrome) See the link for more information on using Google Frames with Internet explorer.
