On 4/17/13 3:35 PM, Michael Brunnbauer wrote:
Hello Kingsley,On Wed, Apr 17, 2013 at 02:58:18PM -0400, Kingsley Idehen wrote:This is all about scalability ultimately being out of the hands of SPARQL and placed back into the hands of computing resources (which includes: the SPARQL processor, host operating system memory, and hardware components such as CPU, disk, interconnects etc..).But those resources only scale by adding more computers. If the problem cannot be effectively parallelized, you are lost at some point.
Yes, but what is it that cannot be effectively parallelized in the context of a useful SPARQL query?
For example, given a corpus of 50 billion+ 3-tuple based relations, is disambiguating patterns such as "China", "New York", "Paris Hilton" , "Database", "Gene" , "Protein" useful enough to demonstrate the concept of "anytime queries" that leverages a share-nothing cluster, key compression, vectorized query processing, and an expanding query timeout?
Live links from a live instance (note the footer for resource utilization and query timing etc..):
1. http://bit.ly/11et1s8 -- New York 2. http://bit.ly/1747mFc -- Protein 3. http://bit.ly/14x0Edj -- Gene 4. http://bit.ly/11isrs6 -- Database 5. http://bit.ly/11xnzQT -- Paris 6. http://bit.ly/12SIvWc -- China
Or did you prove that NC = P (Seehttp://en.wikipedia.org/wiki/P-complete) ? Regards, Michael Brunnbauer
-- Regards, Kingsley Idehen Founder & CEO OpenLink Software Company Web:http://www.openlinksw.com Personal Weblog:http://www.openlinksw.com/blog/~kidehen Twitter/Identi.ca handle: @kidehen Google+ Profile:https://plus.google.com/112399767740508618350/about LinkedIn Profile:http://www.linkedin.com/in/kidehen
smime.p7s
Description: S/MIME Cryptographic Signature
