I can now create an index over an entire collection faster. The bottleneck is still disk accesses, but I still think this was a good challenge.