> At around a billion records, you start getting into distributed/scale-out systems, and distributed indexes are performance death whether for OLTP or analytics

> At around a billion records, you start getting into distributed/scale-out systems, and distributed indexes are performance death whether for OLTP or analytics. Even before you cross into distributed query processing, the random access nature of indexes levies a heavy tax. I assume you mean once process/aggregate a billion records. Having a billion records in a single table on a single, non-distributed system is not an issue at all.