Virtuoso performance/Questions for openlink


 * How do I read the explain output?
 * Why do all the indices have all keys
 * Could we run an RDF_TRIPLE table (i.e. skip graph altogether)
 * How should we think about segment size
 * Could we benefit form making P, G 16 bit instead of 32 bit (since there aren't that many predicates or graphs)
 * How do we figure out how to properly set
 * NumberOfBuffers (and MaxDirtyBuffers)
 * MaxCheckpointRemap
 * Given config file, how much memory will be used: What's the formula?
 * How is the FDsPerFile parameter used? During query? During load? During simultaneous access?
 * What factor, over raw disk speed, should the table scan timing be.
 * What is a good test of write speed
 * Why is index creation CPU limited. Is there some way to make it parallelized. Can indices built in parallel
 * What can be done in isql that can't be done over dbi? Why not? Can I load a specific file via dbi.
 * Can I give SPARUL a file:: url to load. Why doesn't SPARUL Load work on http://sparql.neurocommons.org/ ?
 * Does SPARUL load handle redirects properly?
 * Possible to save just one table? (e.g. serialize as transaction log)
 * http://docs.openlinksw.com/virtuoso/fn_rdf_load_rdfxml_mt.html Why not just N/k - 1?
 * How can I selectively turn off bpel package for virtuoso (--disable-bpel-vad doesn't seem to work)
 * How do we make a "bare" build of virtuoso - ONLY RDF ?
 * Why is my installation reading almost 2G on startup?
 * Which bits of virtuoso state, e.g. trx,
 * Is it a good idea, if we are writing once, reading mostly, to backup/restore or run compact?
 * What is this error: "GPF: insert.c:678 unmoved row lock" and how can we prevent it?
 * Is there some pragma that lets you choose rdf_quad table
 * Run arbitrary sql inside sparql?
 * Reasoning implementation
 * hook for sparql pre/post processing.
 * ODirect
 * How to use statistic.