This is the first half of Ed Bueche’s annual performance session. I’m only going to be posting notes from this half though as I will be presenting during his second session, so if you are attending, send me your notes and I’ll share them with everyone.
- Many of these topics will be covered in the xCP Performance Tuning Guide coming to the EDN
- Suggests taking the Architecture courses to help design and implement the technologies properly. (Good idea for any full-time architects or wanna-be architects)
- Use Structured Data Type’s (SDT) instead of Simple process variables
- SDT=Container of related variables for a workflow process
- Process variables are data elements that are part of a workflow, definition of the variables impacts the underlying data model and the performance
- Simple Process Variables with different base-data types are modeled as separate object types
- Can’t put composite index on mixed repeating / non-repeating attributes, which means separate tables (Only real fix is a union DQL, one that hits the single att & one that hits repeating)
- Using different SDTs in one process can add limits on filtering and sorting on those different SDTs.
- Denormalizing document names into SDT if needed for task list. In general, denormalize for performance as xCP works with the SDTs quite well
- Make Priority Selective within a Single Queue to improve performance on large queue
- Make sure that the work queues are balanced
- Set undisplayed process variables as invisible in order to speed form rendering, don’t depend on it just not being on the form.
- In general, leverage your WDK and Documentum performance tuning instincts to help when tuning xCP, they will serve you well.
- WDK tips
- Disable drag and drop and column resizing in the app.xml file, 10-25% performance improvement
- Disable the preview panel if NOT needed
- Increase <cache-control> parameter value, reduces browser requests for static content
- In SQL Server, make certain the paramatization config item is set to force, similar to Oracle cursor_sharing
- For user response time>fewer workflow agents and higher polling intervals (30-60 seconds)
- For task throughputs, more agents and shorter intervals
- Watch out for contention on dmi_queue_item between Index Agent items and BPM task items as both are in the same table
- Disable full text indexing if not needed, don’t just turn off indexer
- Be proficient at creating indexers on SDTs
- xCP 1.5 and D6.6
- Task list and queue list enhancements as well as Task Form loading
- Global Pagination ensures only 100 rows are being sorted and returned in all scenarios
- DQL to support result windows
- A new DQL hint (RETURN_RANGE)
- DQL Left Outer Join
- Used instead of Union when joining SDT or simple process variable data
- Improves ability to handle multiple process variables
Off to my session on a Migration project at Baltimore Gas & Electric. Still need more coffee. Thinking more sleep tonight would be a good thing.
All information in this post was gathered from the presenters and presentation. It does not reflect my opinion unless clearly indicated (Italics in parenthesis). Any errors are most likely from my misunderstanding a statement or imperfectly recording the information. Updates to correct information are reflected in red, but will not be otherwise indicated.
All statements about the future of EMC products and strategy are subject to change at any time due to a large variety of factors.
2 thoughts on “EMC World 2010: EMC xCP / Documentum Performance, Scalability, and Sizing – Part 1”
thanks Pie – btw the xCP performance guide is already there on EDN
whoops – see you have the link already!
Comments are closed.