[Biojava-l] Elapsed time of feature filtering

Thomas Down td2 at sanger.ac.uk
Tue Jun 10 12:50:45 EDT 2003


On Tue, Jun 10, 2003 at 10:15:31AM +0100, Y D Sun wrote:
> 
> How long it takes to add a sequence to the database in your case? For
> me, it takes 10 hours to insert a BA000040 sequence to DB. 

Okay, I think I may have found the problem...

Could you try again without the accelerators file loaded?
For this particular case, they actually seem to be making
things (significantly) slower rather than faster :-(.

I have a suspicion as to why this might be the case: adding
a large sequence to the database involves performing a huge
number of INSERT statements in a single transaction.  I think
using stored proceedures in the context of such a large
transaction is overloading some part of the transaction isolation
code in the postgres backend.

It may be possible to fix this by tweaking server settings, and
I'll look into this possibility.  But for now, try without the
accelerators and perform VACUUM and ANALYZE queries when you're finished
(this should be done regularly on PostgreSQL installations anyway)
and you should be fine.

      Thomas.


More information about the Biojava-l mailing list