[Bigdata-developers] Optimization for many inserts/deletes?

[Jim M. <mcc...@gm...>]

I'm using bigdata as the SPARQL endpoint for a nanopublication-based
knowledge graph framework, but part of what we are doing with it involves
many inserts and deletes as the knowledge is updated. For instance, a URL
can be mapped to a DOI, triggering a linked data import (as a
nanopublication, qualifying where the data came from, etc.), which then
might trigger an autonomous agent to find the PDF and extract its text,
which would trigger another agent to perform entity extraction, and so on.
This generates a large number of concurrent inserts as each agent adds its
bit of knowledge, and that seems to sometimes cause blazegraph to wedge.
Are there some example namespace configurations I can use to try to
optimize for frequent inserts, deletes, and while maintaining reasonable
read performance?

Thanks,
Jim
-- 
James P. McCusker III, Ph.D.
http://tw.rpi.edu/web/person/JamesMcCusker