- From: William Waites <ww@styx.org>
- Date: Tue, 10 May 2011 21:46:56 +0200
- To: Gregory Williams <greg@evilfunhouse.com>
- Cc: Semantic Web <semantic-web@w3.org>
* [2011-05-10 15:30:31 -0400] Gregory Williams <greg@evilfunhouse.com> écrit: ] This will also apply to other index orderings, not just (g,s,p,o). ] For example, a (p,o,g,s) index can share common (p,o) pairs and store ] lists of (g,s). Right, hence the 25% - things become triples in the best case from this optimisation rather than quads. There are also obvious savings because I think you store what is effectively a trie of the concatenation from (t_1, t_2, t_3) for each permutation, which only really matters if the distinct number of t_1 is small, whic his the case if t_1 is p or maybe if it is g in some sets of data - but here the question is, how much is gained from this in practice? Mind that the question here is actually very practical - if one wants to publish a big dataset and put a SPARQL endpoint in front of it, how much RAM does one need to buy? -w -- William Waites <mailto:ww@styx.org> http://river.styx.org/ww/ <sip:ww@styx.org> F4B3 39BF E775 CF42 0BAB 3DF0 BE40 A6DF B06F FD45
Received on Tuesday, 10 May 2011 19:47:20 UTC