[Bioclusters] Disk optimization for blast

Ami Klein bioclusters@bioinformatics.org
Fri, 2 May 2003 13:13:51 +0200


On Friday 02 May 2003 12:49, Borries Demeler wrote:
> > Dear All,
> > We would like to know if there are parameters in the disks setting th=
at
> > will accelerate our blast processing.
> > We have dual CPU Pentium4 machine with 2 IDE Hard disks 120G.
> > Our O.S. is Suse8.1
> > the output of the dhparm is:
> > hdparm -v /dev/hda
> > /dev/hda:
> >  multcount    =3D 16 (on)
> >  IO_support   =3D  1 (32-bit)
> >  unmaskirq    =3D  0 (off)
> >  using_dma    =3D  1 (on)
> >  keepsettings =3D  0 (off)
> >  readonly     =3D  0 (off)
> >  readahead    =3D  8 (on)
> >  geometry     =3D 15017/255/63, sectors =3D 241254720, start =3D 0.
>
> you can pretty much double your IDE performance by running your two
> drives in RAID-0 mode using Linux's software raid. In my experience
> this works best if you put each drive on a separate IDE port of your
> motherboard and also don't hook up IDE CDROMs or zip drives to the same
> channel. If you don't have enough IDE ports, you can buy a low cost IDE
> RAID card to get some more IDE ports. Hope that helps
>
> -Borries
Thanks,
I have the promise FastTrak100 Raid controlled build in on my motherboard=
 but=20
I have experienced some instabilities with this controller so I went back=
 to=20
the simple IDE.
I thought that the 'readahead' parameter for the hard disk might help.
--=20
Bye,
   Ami.