[Biojava-l] Different implementation of Sequence?

[Thomas Down]

On Fri, Jun 06, 2003 at 10:33:28AM +0100, Y D Sun wrote:
> 
> > -----Original Message-----
> > From: Thomas Down [mailto:thomas at derkholm.net] 
> > Sent: 05 June 2003 22:58
> > To: smh1008 at cus.cam.ac.uk
> > Cc: Y D Sun; Thomas Down; biojava-l at biojava.org
> > Subject: Re: [Biojava-l] Different implementation of Sequence?
> > 
> > 
> > Once upon a time, David Huen wrote:
> > > On Thursday 05 Jun 2003 6:07 pm, Y D Sun wrote:
> > > > Having created the indices as following and restarted postmaster, 
> > > > the performance of feature filtering is even worse. Maybe 
> > MySQL is a 
> > > > better choice than PostgreSQL. Does anyone have the similar 
> > > > experience?
> > > >
> > > Was the access code exactly as you depicted it? ie. only 
> > filtering on 
> > > "CDS".
> > > Also what was the dataset you searched?  was it the same 
> > dataset in both 
> > > EMBL flat file and BioSQL?  What is your version of 
> > postgresql and what was 
> > > the platform?
> > > 
> 
> Yes, that is the exact code I use to filter CDS on one sequence. The
> same code is used for the same sequence loaded from Embl file and
> PostgreSQL database. The execution times (for one sequence only) in two
> cases are highly diverse .
> 
> I installed PostgreSQL 7.3.2 on Linux 2.4.20. 
> 
> The database contains 10 complete bacterial sequences with length from
> 2M to 9M (Embl file size 9M to 18M).  

Hmmm...

That amount of data definitely ought to be handled without
any big problems.  On the other hand, it's enough that if
a lot of the database accesses we're going for indices, it
could plausibly take a minute or so.

I've attached a new schema file which is compatible with the
one on the website, but has some extra CREATE INDEX statements.
Could you try again with that.  If that doesn't help, it might
be worth trying MySQL to get a different datapoint.

Finally, could you point me to one of the EMBL files you use
(preferably the biggest one), and I'll do some testing at
some point.

    Thomas.