CUD tests, fetch and query tests
Their results are important, if you deal with pretty limited sets of data. But "pretty limited set" is actually big enough. You'll still feel the difference on millions of records. But likely, not on billions.
On the other hand, some of features these tests are exposing, are important in any case. For example, CUD sequence batching, that reduces chattiness between RDBMS and ORM.
Materialization test
The most important factor here is complexity per each row produced by database server. If it is constant (quite frequent case), result of this test is important on database of any size. For example, if query is returning a part of the index, low-level RDBMS performance here is constant and quite high (~ up to 20M of index entries per second by our tests - it's easy to measure this, just calculate an aggregate). In such cases materialization will likely be the main factor limiting the performance.
But if per-row complexity depends on database size, materialization is much less important, if important at all. For example, if each produced row is actually a result of join of two big tables, or there was grouping and so on, likely, you shouldn't care about materialization at all in this case.
Kind regards,
Alex Yakunin
