Re: range remove in BTree, or other table storage

On Feb 23, 9:18 am, "Sampo Syreeni" <d..._at_iki.fi> wrote:
> On Feb 21, 9:15 pm, laurento..._at_gmail.com wrote:
>
> > Is anybody aware of literature about algorithms to remove a range of
> > records from a B-Tree, and doing bulk insertion, or better of a table
> > storage implementation which provides that sort of things?
>
> Try at least Carey, DeWitt, Richardson and Shekita: Object and File
> Management in the EXODUS Extensible
> Database System; and Stonebraker, Stettner, Lynn, Kalash and Guttman:
> Document Processing in a Relational Database System. The interesting
> part is where they use a variant B-tree to implement efficient
> insertions into the middle of the key space of a sequentially indexed
> flatfile -- that sort of thing calls for efficient bulk inserts as
> well.
>
> As for buffering, the neatest concept I've come across to date is
> probably Arge: The Buffer Tree: A New Technique for Optimal I/O
> Algorithms.
The most efficient way to do bulk inserts is interset fusion. Any procedurally inspired algorhythmics is at best a hack to reach the objective. As B-trees are themselves hacks because they do not separate logical and physical, the chances you will be able to build efficient bulk insert through b-trees are NULL. Sorry to discourage you but been there done that. Received on Fri Feb 23 2007 - 14:58:33 CET