Hi,
we have a cube, ~ 450GB and ~3000 Partitions. When we make a ProcessUpdate on a Dimension the following things happens:
1) Process the Dimension itself (take 5 min)
2) Process the partitions where the Dimension belongs to (50 minutes, dropping all the aggregations)
3) Commit Phase (~ 25 min!!!! where new connections are not available)
Now I make a ProcessIndex of the cube:
1) Building up the Aggreations (45 min)
2) Now in Profiler I see Progress Report Begin, 6-Commit and Lock Aquired commit Read on Instance Level.
=> no more records in Profiler for about 7 min, new connections possible, then records with Object created and finally Progress Report End, 6-Commit and Lock Released commit Read on Instance Level.
From my point of view I had an Lock for only about 20 seconds.
The Question is why I have an instance wide lock while ProcessUpdate for the dimension for about 25 min but only some seconds while processing the indexes for all partitions?
KR