I have the need to create a denormalized, flattened view of multiple sobjects in Salesforce.com. In numbers this means 20 Lookup fields, around 100 text formula fields to extract data from those relationships and around 100 integer fields.
With that I still stay under the limits but how will this affect performance when I have 1 million or more records of that type?
Any good or bad experience on that would be highly appreciated?
Just in case you might ask: I am doing this denormalisation to reduce ten's of millions of rows with a few fields in favor of few millions with many fields. As every object record in Salesforce is just 2Kb in size and storage is so expensive that seems to be a thinkable improvement. Isn't it?!