Susan, thanks for this article.
Are up-to-date statistics also relevant for creating/rebuilding columnstore indexes for memory grant request estimate to avoid infamous 701 "There is insufficient system memory in resource pool 'default' to run this query." error? `This error occurs even if there is enough RAM available on the server and even if the resource governor setting REQUEST_MAX_MEMORY_GRANT_PERCENT is set to provide 50% of MAXMEM of 175GB (Memory grant request rule-of-thumb estimate is about 24GB for the index)?
Our server successfully rebuilt the index today in the morning using a MAXDOP of all 24 CPU cores. Then after truncating the 14 mio rows (200 cols) table and bulk inserting 14 mio rows (+ current month data of about 60,000 rows), the statistics were out of date, creating the columnstore index failed with all 24 cores, it failed with 12 cores but it finally succeded with 10 cores - with roughly the same data (app. 60,000 difference) as in the morning.
We are often suffering from this error and do not know why we would greatly appreciate your ideas!
Does any of this change with clustered columnstore index in SQL 2014?