Subject: Re: [linux-audio-dev] disk bandwidth
From: Robert Jonsson (ddskrjo_AT_dataductus.se)
Date: Tue Jun 13 2000 - 08:20:36 EEST
Although strictly vaporware, I cannot resist, can't even remember where I read
it.
If you like reeeeeeeally great harddisk performance go for the new solidstate
drives.
100MB/s sustained throughput, virtually no seektimes.
They are supposed to be available at up to 8GB size. At which the price will
probably be astronomic. Never the less :)
Anyone knows who makes these babies? I sure can't remember... :(
/Robert
Benno Senoner wrote:
> On Fri, 09 Jun 2000, Paul Barton-Davis wrote:
>
> > >
> > >all too often, lawyers have made companies anal-retentive about specs.
> > >in the 75GXP doc, I can't find whether it supports a queue depth >0.
> > >for Linux, that's not a big deal: disconnect is the real win, since
> > >that lets us talk to the other disk on the chain. (and Linux already
> > >does a better job of head scheduling with a queue of 256 requests.)
> > >
> > >37 MB/s typical sustained bandwidth (>35 as measured by bonnie,
> > >through ext2). basically A Very Nice Disk.
> >
> > Wow! 37MB/s from one disk ? This is almost enough to make me abandon
> > SCSI :)
>
> hehe amazing how fast people can change their mind ...
> :-)
>
> Anyway what we need is better IDE disk drivers which fully support the
> tagged queueing to achieve SCSI-like performance.
>
> But if this is true ( 35MB/sec) , then
> immagine putting 2-3 of these disks in a soft-RAID0 configuration:
> you have so much disk bandwidth , that the CPU/bus becomes a Problem.
>
> One more reason to buy SMP hardware.
> Let's see when the first 4way or 8way SMP linux boxes will make their inroad
> in the studio.
> :-)
>
> Benno.
>
This archive was generated by hypermail 2b28 : Tue Jun 13 2000 - 09:14:57 EEST