The critical components were the most frequent access/update patterns versus the physical organization of the primary table(s) and their _primary_ indexes, or hashes. Do DBMs still use those things?
Prmary Indexes have a lot more to do with I/O than most people think. You really have to understand how a database stores the physical data on rows and columns. The biggest index killer is Clustered Indexes. Among other things those type of indexes take longer to execute. Why? Because the database takes and reorders all the data for Index 1 - Index Infinite. Regular indexes (standard) do not do so. Also Heape Indexes well, just so to speak slap the data to screen and just don't care what the order it is. Can add in Foreign Keys Relations also.
This can be verified by digging way deep into your db documentation. Then you can actually exec queries to pull the raw data Pages from the database.
Bill, you're the first one I have herd say this in some time.
Among other things is the DB running in memory? Is the DB making temp tables in memory and then dropping them.