Changes between Initial Version and Version 1 of minutes_11-27-07


Ignore:
Timestamp:
11/29/07 16:11:18 (10 years ago)
Author:
joshuadf
Comment:

--

Legend:

Unmodified
Added
Removed
Modified
  • minutes_11-27-07

    v1 v1  
     1= Meeting minutes from 11/27/2007 = 
     2 
     3=== Summary: === 
     4 
     5Discussion was mainly about the merits of different 
     6approaches to materializing large queries, such as  
     7full transitive closure of the entire FMA. 
     8 
     9Marianne presented some of her thoughts from attempts  
     10to "explode" transitive closure at the database level. 
     11 
     12She first with a postgreSQL stored procedure which  
     13took approximately 8 hours but resulted in a 100GB+ 
     14database size.  
     15 
     16Marianne also experimented with SQL99 recursive queries 
     17in MS SQL Server. Unfortunately computing transitive 
     18closure over the entire FMA took days (we theorized  
     19that the DBMS was running out of memory and so was  
     20frequently forced to swap to disk). 
     21 
     22Neither approach currently provide any performance  
     23advantages vs Todd's transitive closure implemented  
     24as high-level Java code (an extension to Jena). 
     25However, Dan Suciu pointed out that it is guaranteed 
     26to fail in the general case (for example, a query  
     27like "Give me the FMA plus some other nodes") due to  
     28memory constraints, while a DBMS will at least swap  
     29to disk. Additionally, we need to support large  
     30queries for the practical reason of providing fully  
     31materialized ontologies to third parties.