Reputation: 93
I'd like to ask why automatic clustering incurs relatively high costs when compared to manual clustering with dedicated big warehouse? We are currently paying almost 100 credits a month for automatic clustering of some tables but at the same time we are also maintaining clustering of selected tables manually, and that costs just a few credits.
These are not well comparable, because the tables are different, but we strongly believe that automatic clustering is more expensive in general. Is there an explanation for that phenomena?
Upvotes: 3
Views: 1662
Reputation: 1108
The cost of automatic reclustering has to do with how much incremental work the algorithm can find to reduce depth, whereas manual recluster just does what it can for the resources allocated.
So in your scenario, the algorithm must be finding that there are performance gains to be had by reclustering to warrant the cost of the reclustering work.
Automatic reclustering has the goal "Reduce Worst Clustering Depth below an acceptable threshold to get Predictable Query Performance" which is different than manual reclustering which just groups/sorts as much as is possible within the given warehouse.
Upvotes: 3