> It's routine these days that faster machines have multiple drives in
> RAID configs. By pipelining, a RAID config can effectively multiply the
> bandwidth of one drive by the number of drives. At work we have two
> machines, one with a 4 drive RAID, the other with 8 drives (these are
> SAS drives). The difference in speed between the two machines is quite
> noticeable, even on jobs that don't read the files linearly (i.e. HDL
> synthesis).

I think it is not linear, however, just installed Ubuntu 8.10 on my new
laptop with RAID 0 and working on large files are definitely faster. The
laptop has two SATA HD inside and as I am using virtual machine to run Win +
MPLAB I can see the difference on both booting up the Windows on that
virtual machine and also saving and restoring the machine state. Also doing
some video rendering from my camcorder and to exporting the project into DVD
format is way faster than it was before with the single hard drive.

With USB hard drive I have a 400GB of Maxtor something, and just installed
dm-crypt to protect my backups. As part of the process of crypting I just
filled up the drive with pseudo random data. For this I was using the
/dev/urandom to copy the data from there to the drive - used dd command for
that. After two hours copy it finished only with 9GB ... dunno if that was
about the USB and the block size dd was using but I thought it was strange.
Copying 300GB worth of data was finished withing couple of hours though so I
suppose it was closer to the real speed of USB2 bulk transfer. But found it
still very slow, I would rather use external SATA but then that is not
compatible with every environment - like laptops typically do not have
external SATA connector.

BTW I have a 1394 port but have no drive for that to make some speed
comparision.

Tamas


On Thu, Nov 27, 2008 at 4:09 PM, Herbert Graf <mailinglist4@farcite.net>wrote:

> On Thu, 2008-11-27 at 09:34 +0000, Alan B. Pearce wrote:
> > > Why? Perdon my ignorance but, if current limitation is HD
> > > speed (not USB2 speed) how a mass storage application can
> > > increase speed (I mean in practice, not in comercial
> > > -greatest-and-latest-news-vendor-approach).
> >
> > I have seen SCSI drive systems using standard drives, but with extremely
> > high speed performance, because the drive control electronics had
> something
> > like 256MB of buffer cache, allowing the hardware to do look-ahead reads,
> > and buffer writes, making a hard drive assembly look like a memory drive.
> > This was used on a mainframe, where database access could really thrash a
> > drive, but the amount of buffering cache meant that many of the accesses
> > could be done from the cache instead of the HD itself.
>
> Wow, that's one big buffer!
>
> It's routine these days that faster machines have multiple drives in
> RAID configs. By pipelining, a RAID config can effectively multiply the
> bandwidth of one drive by the number of drives. At work we have two
> machines, one with a 4 drive RAID, the other with 8 drives (these are
> SAS drives). The difference in speed between the two machines is quite
> noticeable, even on jobs that don't read the files linearly (i.e. HDL
> synthesis).
>
> TTYL
>
> --
> http://www.piclist.com PIC/SX FAQ & list archive
> View/change your membership options at
> http://mailman.mit.edu/mailman/listinfo/piclist
>



-- 
Rudonix DoubleSaver
http://www.rudonix.com
-- 
http://www.piclist.com PIC/SX FAQ & list archive
View/change your membership options at
http://mailman.mit.edu/mailman/listinfo/piclist