Opinionated Software

… we have opinions about everything!

Archive for the ‘Performance’ Category

PeopleTools and Parallelism on SQL Server —

During a recent discussion with another PS Admin running on SQL Server it became apparent that his efforts to improve application SQL performance through adding more CPUs (with the associated license costs) were based on a fundamentally wrong assumption: PeopleTools/Application SQL will go parallel if needed. This is simply not true. In fact, most SELECT […]

Re-Use in Application Engine —

Almost without exception whenever I am asked to review the performance of a PeopleSoft system I discover issues that lead back to locally developed Application Engine processes. In particular, high database SQL parsing rates invariably originate from SQL steps that should have the re-use flag set to Yes. What this flag does is replace any […]

More Hardware Vicar? —

I’m old school. I started in “Personal Computing” when 2K of RAM was the norm. When I started programming in COBOL on IBM mainframes, we had machines with 8, 16 or 32 Megabytes of RAM. Not Gigabytes. Processor speeds were in single digit MegaHertz – not GigaHertz. Which is why I often despair when developers […]

RC_BACKLOG_VW in PeopleSoft CRM 9.1 —

Seriously? Who thought it would be a good idea to create a view to return the DISTINCT list of BUSINESS_UNIT from RC_CASE? Talk about an O (n) performance disaster. What’s wrong with a SELECT BUSINESS_UNIT FROM PS_BUS_UNIT_TBL_RC WHERE EXISTS (… case sub-query….). Oh look … Two index seeks … shocker! It scales too …. wow […]

DataMover EXPORT Performance —

Some interesting statistics from some recent tests: PeopleTools 8.49.28 EXPORT of a table containing 1.6 million rows (approx 1.6 GB of data): 3 hours 15 minutes PeopleTools 8.53.20 IMPORT of the same data: 34 minutes PeopleTools 8.53.20 EXPORT of the same table containing the identical 1.6 million rows (approx 1.6 GB of data): 24 minutes […]

_WA_Sys Stats or Hand Crafted? —

SQL Server generates histogram statistics automatically whenever a column is referenced in a predicate and it is not already the leading column in an existing histogram statistic. By design, indexes create histogram statistics automatically so any column you have that is the leading column in an index will already have histogram statistics. The autogenerated statistics […]