What are better approaches to reduce Impact CPU (High) on collect stats (full stats) statements ? One of the options is to convert full stats to sampled stats provided that sampled stats are not impacting queries performance. Did anyone come across this situation ? What approaches were followed to reduce impact cpu ?
If you are on TD14.10 or higher, use AutoStats. See Carrie Ballinger's documents on this site to get started.
It will not be possible to change the impact CPU use for a full stats collection, it is what it is - alrady optimized by the database to get it done as quickly as possible. But with AutoStats, the frequency of stats and the use of sampled stats can be optimized to only be collected when necessary rather than on a forced schedule basis.