Hi everyone!
I am trying to speed-up cube processing time. There are 2 servers - one for the data storage (srvr1) and one for OLAP (srvr2).
Currently, the biggest problem is with one Fact table (60 columns) which has around 200 million records and processing of this table takes around 2 hours. I set ColumnStore index on that Fact table but I think there is no problem with reading the data from the disk (from srvr1), because server state goes from RUNNING to state SUSPENDED (SYNC_NETWORK_IO) which means srvr1 has sent the data.
Now, I have to find where the problem is. Is it network or OLAP processing (aggregations, MDX's,...). The processing rate is currently between 50k and 90k rows per second which is very slow (unacceptable), because Fact table is currently limited to 200 million
records and real amount of data is more than 1 billion, which means that processing would take ~10 hours which is just too much, because we don't have so much time (night is not long enough :)).
Can you give me some advice? Where should I look for a problem? How could I test the network? Could I configure OLAP server differently?
Thanks in advance!
Miljan