Fuseki tdb indexing

["Chaudhuri, Rajiv" <ra...@pearson.com>]

Hi

Is there any command to index the fuseki tdb data?

Will  this indexing (if we have) results in performance improvement while
querying the triple store? Please advice.

Re: Fuseki tdb indexing

[Rob Vesse <rv...@dotnetrdf.org>]

No there is no command to directly index data as such, TDB automatically
creates and maintains appropriate indexes where you create/modify a TDB
database.

However there is a tdbstats tool which will generate statistics for your
data that can help TDB schedule your query better and potentially improve
performance (though there is no guarantee of this).  See
http://jena.apache.org/documentation/tdb/optimizer.html

Rob

On 22/12/2013 06:34, "Chaudhuri, Rajiv" <ra...@pearson.com>
wrote:

>Hi
>
>Is there any command to index the fuseki tdb data?
>
>Will  this indexing (if we have) results in performance improvement while
>querying the triple store? Please advice.

Re: Fuseki tdb indexing

[Andy Seaborne <an...@apache.org>]

On 22/12/13 06:34, Chaudhuri, Rajiv wrote:
> Hi
>
> Is there any command to index the fuseki tdb data?

If you look in the TDB directory, you will see files like SPO.idn and 
SPO.dat.  That's the SPO clustered index.  For triples the indexes are 
SPO, POS and OSP.  More indexes for quads (named graphs).

TDB has total coverage (every triple order can be answered by an index).

There is no triple table per se - it's all indexes.

>
> Will  this indexing (if we have) results in performance improvement while
> querying the triple store? Please advice.

Performance questions are hard to deal with without knowing about the 
system architecture, the size of the data and the queries being asked.

 From your other email, there are glimpses of this not not details. e.g. 
If you have an under provisioned database server, and are paging data 
across the network (e.g. NFS), then it can be slow especially if the NFS 
server itself is shared.

	Andy