[OT] Efficient file structure for very large lookup tables?

[tn]

On Tuesday, 17 December 2013 at 19:09:49 UTC, H. S. Teoh wrote:
> However,
> they do exhibit good spatial locality in higher-dimensional 
> space (i.e.,
> if entry X is accessed first, then the next entry Y is quite 
> likely to
> be close to X in that space).  Does anybody know of a good data
> structure that can take advantage of this fact to minimize disk
> accesses?

Perhaps some kind of k-d tree? Maybe even some sort of cache 
oblivious k-d tree to make it more suitable for storage on the 
disk?